CN111666957B - Image authenticity recognition method and device - Google Patents

Abstract

The invention provides a method and device for identifying image authenticity, comprising: extracting features of an image to be identified, obtaining deep learning features, ORB features and hash features; Classified images similar to the image; if there is no classified image similar to the image to be recognized, then based on the deep learning feature, determine the candidate similar image similar to the image to be recognized in the historical image; based on the ORB feature and hash feature, in the candidate similar image Determine the similar image of the image to be recognized; determine the authenticity of the image to be recognized according to the user corresponding to the image to be recognized and the user corresponding to the similar image. The method of the present invention combines a variety of different feature extraction methods, which greatly improves the accuracy of image recognition, and is more intelligent through machine automatic recognition, improves the efficiency of image recognition, and reduces labor costs.

Description

Image authenticity recognition method and device

technical field

The invention relates to the technical field of image recognition, in particular to an image authenticity recognition method and device.

Background technique

In recent years, with the popularization of the Internet and mobile Internet, the credit business has developed rapidly, and the business has become more and more diversified and convenient. However, while the credit business provides a wealth of financial services for the majority of users, it also brings new risks. In recent years, there have been more and more cases of using loopholes or necessary technical means to carry out illegal transactions, seriously damaging the property safety of banks and customers. With the professionalization of the criminal process and the upgrading of technical means, the traditional risk control model based on rules + verification has been difficult to meet the current risk control needs.

In the credit business, users will submit different types of images in multiple scenarios. When manually checking the authenticity of the images, it is impossible to check all of them, and the requirements for human memory are high. Not only is the speed slow, but the accuracy rate is not high.

To sum up, the existing identification methods for image authenticity have the technical problems of slow speed and low accuracy.

Contents of the invention

The object of the present invention is to provide an image authenticity identification method and device, so as to alleviate the technical problems of the existing image authenticity identification method which are slow in speed and low in accuracy.

In a first aspect, an embodiment of the present invention provides a method for identifying the authenticity of an image, including:

Carry out feature extraction on the image to be identified, obtain the deep learning feature, ORB feature and hash feature of the image to be identified;

Based on the deep learning features of the image to be recognized, determine whether there is a classified image similar to the image to be recognized in the current classification template library, so as to determine the authenticity of the image to be recognized according to the classified image similar to the image to be recognized sex;

If there is no classified image similar to the image to be recognized in the current classification template library, then based on the deep learning features of the image to be recognized, determine a candidate similar image similar to the image to be recognized in the historical image;

Calculate the distance between the image to be recognized and the candidate similar image based on the ORB feature of the image to be recognized and the hash feature of the image to be recognized, and determine the candidate similar image in the candidate similar image based on the distance Identify similar images of an image;

The authenticity of the image to be identified is determined according to the user corresponding to the image to be identified and the user corresponding to the similar image.

Further, based on the deep learning features of the image to be recognized, determine whether there is a classification image similar to the image to be recognized in the current classification template library, so as to determine the classification image to be recognized according to the classification image similar to the image to be recognized Authenticity of images includes:

Calculating the first cosine distance between the deep learning features of the image to be identified and the deep learning features of each classified image in the current classification template library;

determining a first target cosine distance among the first cosine distances based on a first preset distance, wherein the first target cosine distance is greater than the first preset distance;

Using the target classification image corresponding to the first target cosine distance as a classification image similar to the image to be recognized;

The authenticity of the image to be recognized is determined according to the category to which the target classification image belongs.

Further, based on the deep learning features of the image to be identified, determining a candidate similar image similar to the image to be identified in the historical image includes:

Calculating the second cosine distance between the deep learning features of the image to be recognized and the deep learning features of each historical image;

determining a second target cosine distance among the second cosine distances based on a second preset distance, wherein the second target cosine distance is greater than the second preset distance;

The target historical image corresponding to the second target cosine distance is used as a candidate similar image similar to the image to be recognized.

Further, calculating the distance between the image to be recognized and the candidate similar image based on the ORB feature of the image to be recognized and the hash feature of the image to be recognized includes:

calculating the first distance between the ORB feature of the image to be identified and the ORB feature of the candidate similar image;

calculating a second distance between the hash feature of the image to be identified and the hash feature of the candidate similar image;

Weighted calculation is performed on the first distance and the second distance based on weights to obtain a distance between the image to be recognized and the candidate similar image.

Further, based on the distance, determining a similar image of the image to be recognized among the candidate similar images includes:

determining a target distance among the distances based on a third preset distance, wherein the target distance is greater than the third preset distance;

The target candidate similar image corresponding to the target distance is used as the similar image of the image to be recognized.

Further, determining the authenticity of the image to be identified according to the user corresponding to the image to be identified and the user corresponding to the similar image includes:

If in the similar images, the user corresponding to the target similar image and the user corresponding to the image to be identified are not the same user, then determine that the image to be identified is a false image;

If, among the similar images, the users corresponding to all the similar images are the same user as the user corresponding to the image to be recognized, then it is determined that the image to be recognized is a real image.

Further, after determining the authenticity of the image to be identified according to the user corresponding to the image to be identified and the user corresponding to the similar image, the method further includes:

The first preset distance, the second preset distance, the third preset distance, and the weight are automatically updated according to the manual feedback result, wherein the manual feedback result is manually The recognition image is compared with the target similar image to obtain a result of whether the to-be-recognized image and the target similar image are similar images of different users.

Further, before performing feature extraction on the image to be recognized, the method also includes:

Obtain the current classification template library.

Further, obtaining the current classification template library includes:

Clustering images collected within a preset time range by using a clustering method to obtain images that appear frequently within the preset time range;

The authenticity of the frequently appearing images is judged manually, and the classification template library is updated according to the judgment result to obtain the current classification template library.

In the second aspect, the embodiment of the present invention also provides an image authenticity recognition device, including:

The feature extraction unit is used to perform feature extraction on the image to be identified, and obtain the deep learning feature, ORB feature and hash feature of the image to be identified;

The first determining unit is configured to determine whether there is a classified image similar to the image to be recognized in the current classification template library based on the deep learning features of the image to be recognized, so as to determine based on the classified image similar to the image to be recognized authenticity of the image to be identified;

The second determination unit, if there is no classified image similar to the image to be recognized in the current classification template library, based on the deep learning features of the image to be recognized, determine in the historical image that is similar to the image to be recognized Candidate similar images for ;

A computing unit, configured to calculate the distance between the image to be recognized and the candidate similar image based on the ORB feature of the image to be recognized and the hash feature of the image to be recognized, and based on the distance, in the candidate similar image determining similar images of the image to be identified;

The third determining unit is configured to determine the authenticity of the image to be identified according to the user corresponding to the image to be identified and the user corresponding to the similar image.

In an embodiment of the present invention, a method for identifying the authenticity of an image is provided. The method includes: first performing feature extraction on the image to be identified to obtain deep learning features, ORB features, and hash features of the image to be identified; then, based on the The deep learning features of the image determine whether there is a classified image similar to the image to be recognized in the current classification template library, so as to determine the authenticity of the image to be recognized based on the classified image similar to the image to be recognized; if there is no similar image in the current classification template library For classified images that are similar to the image to be recognized, based on the deep learning features of the image to be recognized, a candidate similar image similar to the image to be recognized is determined in the historical image; Identify the distance between the image and the candidate similar image, and determine the similar image of the image to be recognized among the candidate similar images based on the distance; finally, determine the authenticity of the image to be recognized according to the user corresponding to the image to be recognized and the user corresponding to the similar image. It can be seen from the above description that the image authenticity recognition method of the present invention combines a variety of different feature extraction methods, which greatly improves the accuracy of image recognition, and is more intelligent through machine automatic recognition, improving the efficiency of image recognition. The labor cost is reduced, and the technical problems of slow speed and low accuracy of the existing recognition method of image authenticity are alleviated.

Description of drawings

In order to more clearly illustrate the specific implementation of the present invention or the technical solutions in the prior art, the following will briefly introduce the accompanying drawings that need to be used in the specific implementation or description of the prior art. Obviously, the accompanying drawings in the following description The drawings show some implementations of the present invention, and those skilled in the art can obtain other drawings based on these drawings without any creative work.

FIG. 1 is a schematic flowchart of an image authenticity recognition method provided by an embodiment of the present invention;

FIG. 2 is a flow chart of a method for determining whether there is a classified image similar to the image to be recognized in the current classification template library provided by an embodiment of the present invention;

FIG. 3 is a flow chart of a method for determining a candidate similar image similar to an image to be recognized in a historical image provided by an embodiment of the present invention;

4 is a flowchart of a method for determining a similar image of an image to be recognized among candidate similar images provided by an embodiment of the present invention;

Fig. 5 is a schematic diagram of an image authenticity recognition device provided by an embodiment of the present invention.

Detailed ways

The technical solutions of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of them. Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without making creative efforts belong to the protection scope of the present invention.

To facilitate the understanding of this embodiment, a method for identifying the authenticity of an image disclosed in the embodiment of the present invention is firstly introduced in detail.

Embodiment one:

In order to facilitate the understanding of this embodiment, a method for identifying the authenticity of an image disclosed in the embodiment of the present invention is firstly introduced in detail, referring to the schematic flow chart of a method for identifying the authenticity of an image shown in Figure 1, which mainly includes the following steps :

Step S102, performing feature extraction on the image to be recognized to obtain deep learning features, ORB features and hash features of the image to be recognized;

In the embodiment of the present invention, the method for identifying the authenticity of an image can be applied to a server, and is an automatic and fast image identification method. The algorithms used for the above feature extraction include: ORB (Oriented Fast and Rotated Brief) algorithm, CNN convolutional neural network and hash algorithm.

Among them, ORB (Oriented Fast and Rotated Brief), can be used to quickly create feature vectors for key points in the image, and these feature vectors can be used to identify objects in the image. ORB first looks for special regions in the image, called keypoints. Keypoints are small prominent regions in an image, such as corners, that are characterized by sharp changes in pixel values from light to dark. ORB then computes the corresponding feature vector for each keypoint. The eigenvectors created by the ORB algorithm contain only 1s and 0s and are called binary eigenvectors. The order of 1s and 0s changes according to the particular keypoint and the pixel area around it. This vector represents the intensity pattern around the keypoint, so multiple feature vectors can be used to identify larger regions or even specific objects in the image. ORB is characterized by being super fast and somewhat immune to noise and image transformations such as rotation and scaling.

The basic structure of CNN (Convolutional Neural Network, Convolutional Neural Network) includes two special neuron layers, one of which is a convolutional layer, the input of each neuron is connected to the part of the previous layer, and the feature of the part is extracted ; The second is the pooling layer, which is used to calculate the local sensitivity and the calculation layer of the secondary feature extraction. This twice feature extraction structure reduces feature resolution and reduces the number of parameters to be optimized. The VGG16 network architecture (Very Deep Convolutional Networks for Large-Scale Image Recognition) is used in the present invention, which mainly allows the network to go deeper through the small convolution kernel size.

It should be noted that, in addition to the above three types of features, the hist feature may also be included.

Step S104, based on the deep learning features of the image to be recognized, determine whether there is a classified image similar to the image to be recognized in the current classification template library, so as to determine the authenticity of the image to be recognized according to the classified image similar to the image to be recognized;

In the embodiment of the present invention, the above-mentioned current classification template library includes two kinds of classification template libraries, one is a template library of real images, and the other is a template library of fake images. Wherein, the template library of the real image contains the real image template and the deep learning features of the real image template, and the template library of the fake image contains the fake image template and the deep learning features of the fake image template. The above-mentioned current classification template library is constructed based on clustering technology and manual methods, and the construction process of the template library will be described in detail below.

During implementation, if there is a classified image A similar to the image to be recognized in the current classification template library, and the category to which the classified image A belongs is a real image template, then it can be determined that the image to be recognized is a real image; if the category to which the classified image A belongs is a false image template, then it can be determined that the image to be recognized is a false image.

The process of determining whether there is a classified image similar to the image to be recognized in the current classification template library will be described in detail below, and will not be repeated here.

Step S106, if there is no classified image similar to the image to be recognized in the current classification template library, based on the deep learning features of the image to be recognized, determine a candidate similar image similar to the image to be recognized in the historical images;

In the embodiment of the present invention, a historical image database is stored in the server, and the historical image database includes historical images and image features of the historical images. Specifically, the image features of historical images include: deep learning features of historical images, ORB features of historical images, and hash features of historical images, each of which constitutes an image feature dictionary. For example, the deep learning features of historical images constitute the deep learning feature dictionary, the ORB features of historical images constitute the ORB feature dictionary, and the hash features of historical images constitute the hash feature dictionary.

The above image features are obtained by extracting historical images based on ORB features, hash features and deep learning feature vectors in opencv.

Step S108, calculating the distance between the image to be recognized and the candidate similar image based on the ORB feature of the image to be recognized and the hash feature of the image to be recognized, and based on the distance, determining a similar image of the image to be recognized among the candidate similar images;

The inventor considers that the recognition of similar images is easily affected by the shooting environment due to the complexity of the image itself, such as changes in illumination, scale, and perspective, and the recognition accuracy of different types of images is not high. There are many kinds of image feature extraction methods, and there is no one-size-fits-all feature that can solve the above-mentioned various situations. Based on this, the inventor has integrated the above-mentioned various feature extraction methods, which can take into account multiple types of images to improve similarity. recognition accuracy.

Step S110, determine the authenticity of the image to be identified according to the user corresponding to the image to be identified and the user corresponding to the similar image.

In an embodiment of the present invention, a method for identifying the authenticity of an image is provided. The method includes: first performing feature extraction on the image to be identified to obtain deep learning features, ORB features, and hash features of the image to be identified; then, based on the The deep learning features of the image determine whether there is a classified image similar to the image to be recognized in the current classification template library, so as to determine the authenticity of the image to be recognized based on the classified image similar to the image to be recognized; if there is no similar image in the current classification template library For classified images that are similar to the image to be recognized, based on the deep learning features of the image to be recognized, a candidate similar image similar to the image to be recognized is determined in the historical image; Identify the distance between the image and the candidate similar image, and determine the similar image of the image to be recognized among the candidate similar images based on the distance; finally, determine the authenticity of the image to be recognized according to the user corresponding to the image to be recognized and the user corresponding to the similar image. It can be seen from the above description that the image authenticity recognition method of the present invention combines a variety of different feature extraction methods, which greatly improves the accuracy of image recognition, and is more intelligent through machine automatic recognition, improving the efficiency of image recognition. The labor cost is reduced, and the technical problems of slow speed and low accuracy of the existing recognition method of image authenticity are alleviated.

The above content briefly introduces the image authenticity recognition method of the present invention, and the specific content involved will be described in detail below.

The inventor considers that there are tens of millions of historical images, and there are a lot of them every day, and in so many image submission records, only a few images may be false images submitted by illegal users, and the proportion of black and white samples in the total image data is very different . In such a case, a large number of people are arranged to directly construct the classification template library, which requires extremely high labor, and it is impossible to accurately, quickly and fully construct the classification template library.

Based on this, the embodiment of the present invention provides a method for constructing the current classification template library. In an optional embodiment of the present invention, obtaining the current classification template library includes the following steps (1)-(2):

(1) Using a clustering method to cluster the images collected within the preset time range to obtain images that appear frequently within the preset time range;

(2) Carry out authenticity judgment on frequently occurring images by artificial means, and update the classification template library according to the judgment result to obtain the current classification template library.

Specifically, within the preset time range (such as within the past 30 days), the images collected within the time range are clustered using the clustering method to obtain the cluster center image, that is, the image with high frequency, and then the high frequency The image pushes that appear are manually judged for their authenticity. During implementation, manually determine the authenticity of the above-mentioned frequently occurring images based on false image templates found in history (such as an image of a bank transfer recording PS amount) and real image templates (such as business approval emails and screenshots of system documents, etc.). Then update the classification template library to obtain the template library of the current real image and the template library of the current false image. The initial construction process of the classification template library can also be realized by referring to the above process.

The above-mentioned template library of real images is used to eliminate batches of real images without early warning and feedback; the above-mentioned template library of fake images is used to identify batches of fake images.

The invention combines the clustering technology to find images that appear frequently in the near future, automatically establishes a classification template library, and can assist manual judgment of whether the images of this type are general images of gang crimes or business approval.

In an optional embodiment of the present invention, referring to FIG. 2, step S104, based on the deep learning features of the image to be recognized, determines whether there is a classified image similar to the image to be recognized in the current classification template library, so as to Similar Classified Images The steps to determine the authenticity of the image to be recognized include:

Step S201, calculating the first cosine distance between the deep learning features of the image to be recognized and the deep learning features of each classified image in the current classification template library;

Specifically, the first cosine distance between the deep learning feature vector of the image to be recognized and the deep learning feature vector of each classified image in the current classification template library is calculated.

Step S202, determining a first target cosine distance in the first cosine distance based on the first preset distance, wherein the first target cosine distance is greater than the first preset distance;

Step S203, using the target classification image corresponding to the first target cosine distance as a classification image similar to the image to be recognized;

Step S204, determining the authenticity of the image to be recognized according to the category to which the target classification image belongs.

As can be seen from the above, the classification template library of the present invention can add false images or real images that appear frequently in history as templates, and use clustering technology to automatically find images that appear frequently to assist the establishment of the classification template library, reducing manual work In addition, the authenticity of the image to be recognized can be determined by matching the image to be recognized with the classification image in the classification template library, which reduces the amount of calculation and improves the efficiency of image recognition.

In an optional embodiment of the present invention, referring to FIG. 3, step S106, based on the deep learning features of the image to be recognized, the step of determining a candidate similar image similar to the image to be recognized in the historical image includes:

Step S301, calculating the second cosine distance between the deep learning features of the image to be recognized and the deep learning features of each historical image;

Step S302, determining a second target cosine distance in the second cosine distance based on the second preset distance, wherein the second target cosine distance is greater than the second preset distance;

Step S303, taking the target historical image corresponding to the second target cosine distance as a candidate similar image similar to the image to be recognized.

The process from step S301 to step S303 above completes the first round of comparison between the image to be recognized and the historical image, and a set of candidate similar images is obtained through preliminary screening.

In an optional embodiment of the present invention, referring to FIG. 4, step S108, the step of determining a similar image of the image to be recognized among candidate similar images includes:

Step S401, calculating the first distance between the ORB feature of the image to be recognized and the ORB feature of the candidate similar image;

Step S402, calculating the second distance between the hash feature of the image to be recognized and the hash feature of the candidate similar image;

Step S403, performing weighted calculation on the first distance and the second distance based on the weight, to obtain the distance between the image to be recognized and the candidate similar image;

Specifically, calculate the first distance (could be a cosine distance) between the ORB feature of the image to be recognized and the ORB feature of each candidate similar image, and calculate the hash feature between the image to be recognized and the hash feature of each candidate similar image The second distance (which may be the Hamming distance) is further calculated based on the weight of the first distance, the weight of the second distance, the first distance and the second distance to obtain the distance between the image to be recognized and each candidate similar image.

Step S404, determining a target distance among distances based on a third preset distance, wherein the target distance is greater than the third preset distance;

In step S405, the target candidate similar image corresponding to the target distance is used as a similar image of the image to be recognized.

The process from step S401 to step S405 above completes the second round of comparison between the image to be recognized and the candidate similar images, and the second round of fine screening obtains a set of similar images of the image to be recognized.

In an optional embodiment of the present invention, step S110, the step of determining the authenticity of the image to be recognized according to the user corresponding to the image to be recognized and the user corresponding to the similar image includes the following 1) and 2) processes:

1) If in the similar image, the user corresponding to the target similar image and the user corresponding to the image to be identified are not the same user, then determine that the image to be identified is a false image;

2) If in the similar images, the users corresponding to all the similar images and the user corresponding to the image to be recognized are the same user, then it is determined that the image to be recognized is a real image.

Specifically, when judging whether the user corresponding to the image to be recognized and the user corresponding to the similar image are the same user, it may be judged in the form of user business information. For example, if the certificate number of the contract corresponding to the image to be recognized is the same as the certificate number of the contract corresponding to the similar image, it is determined that the user corresponding to the image to be recognized is the same user as the user corresponding to the similar image, and vice versa.

Under normal circumstances, the user corresponding to the image to be recognized and its similar image should be the same user. If the user corresponding to the image to be recognized and its similar image is not the same user, it means that the user corresponding to the image to be recognized has stolen a similar image similar to the image to be recognized. image.

In an optional embodiment of the present invention, after determining the authenticity of the image to be identified according to the user corresponding to the image to be identified and the user corresponding to the similar image, the method further includes:

Automatically update the first preset distance, the second preset distance, the third preset distance and the weight according to the manual feedback result, wherein the manual feedback result is the image to be recognized obtained by comparing the image to be recognized with the similar image of the target in a manual way Whether the image and the target similar image are the result of similar images from different users.

After completing the recognition of the authenticity of the image to be recognized, similar images that are not the same user (i.e. target similar images) are sent to the manual, and the manual compares the image to be recognized with the target similar image to obtain the comparison result, and compares the comparison result Feedback to the image authenticity recognition algorithm of the present invention, so that the algorithm automatically updates the first preset distance, the second preset distance, the third preset distance and weight, so that the algorithm becomes a closed loop, which can better adapt to Variety.

The image authenticity recognition method of the present invention extracts image features through different algorithms, and after three rounds of image comparison of classification template library, preliminary screening and fine screening, can determine similar images of the image to be recognized more quickly and accurately. In addition, through image aggregation Class-assisted manual judgment of whether the image of this type is a fake image or a real image (that is, a general image of the business approval class) is used to build a template library, reducing labor costs.

The identification method of image authenticity of the present invention has the following advantages:

1. According to the characteristics of different feature extraction algorithms, multiple image feature extractions are performed on the same image, and after preliminary screening and fine screening (that is, a variety of different image feature extraction methods and similarity calculation methods are integrated), different images can be used. The characteristics of the type feature extraction algorithm, good accuracy;

2. For the classification template library, false images or real images that appear frequently in history can be added as templates, and clustering technology can be used to automatically find frequently occurring images to assist the establishment of the classification template library, reducing the manual workload;

3. The authenticity of the image to be recognized can be determined by matching the image to be recognized with the classified image in the classification template library, which reduces the amount of calculation and improves the efficiency of image recognition;

4. It can automatically update the distance threshold and weight according to the manual feedback results, making the algorithm a closed loop, which can better adapt to changes and improve the accuracy of the algorithm.

Embodiment two:

The embodiment of the present invention also provides an image authenticity identification device, the image authenticity identification device is mainly used to implement the image authenticity identification method provided in the above content of the embodiment of the present invention, the following provides the embodiment of the present invention The identification device for image authenticity will be introduced in detail.

Fig. 5 is a schematic diagram of an image authenticity recognition device according to an embodiment of the present invention. As shown in Fig. 5, the image authenticity recognition device mainly includes: a feature extraction unit 10, a first determination unit 20, and a second determination unit 30. The calculation unit 40 and the third determination unit 50, wherein:

The feature extraction unit is used for feature extraction of the image to be recognized, and obtains the deep learning feature, ORB feature and hash feature of the image to be recognized;

The first determination unit is configured to determine whether there is a classified image similar to the image to be recognized in the current classification template library based on the deep learning features of the image to be recognized, so as to determine the authenticity of the image to be recognized according to the classified image similar to the image to be recognized ;

The second determining unit, if there is no classified image similar to the image to be recognized in the current classification template library, then based on the deep learning features of the image to be recognized, determine a candidate similar image similar to the image to be recognized in the historical image;

A calculation unit for calculating the distance between the image to be recognized and the candidate similar image based on the ORB feature of the image to be recognized and the hash feature of the image to be recognized, and based on the distance, determining a similar image of the image to be recognized among the candidate similar images;

The third determination unit is configured to determine the authenticity of the image to be identified according to the user corresponding to the image to be identified and the user corresponding to the similar image.

In an embodiment of the present invention, an image authenticity identification device is provided, which includes: first performing feature extraction on the image to be identified to obtain deep learning features, ORB features, and hash features of the image to be identified; then, based on the image to be identified Deep learning features to determine whether there is a classified image similar to the image to be recognized in the current classification template library, so as to determine the authenticity of the image to be recognized based on the classified image similar to the image to be recognized; if there is no image similar to the image to be recognized in the current classification template library For classified images with similar images, based on the deep learning features of the image to be recognized, determine the candidate similar image similar to the image to be recognized in the historical image; and then calculate the image to be recognized based on the ORB feature of the image to be recognized and the hash feature of the image to be recognized The distance from the candidate similar image, and based on the distance, determine the similar image of the image to be recognized among the candidate similar images; finally, determine the authenticity of the image to be recognized according to the user corresponding to the image to be recognized and the user corresponding to the similar image. From the above description, it can be seen that the image authenticity recognition device of the present invention combines a variety of different feature extraction methods, which greatly improves the accuracy of image recognition, and is more intelligent through automatic machine recognition, improving the efficiency of image recognition. The labor cost is reduced, and the technical problems of slow speed and low accuracy of the existing recognition method of image authenticity are alleviated.

Optionally, the first determination unit is also used to: calculate the first cosine distance between the deep learning features of the image to be recognized and the deep learning features of each classified image in the current classification template library; Determine the first target cosine distance in a cosine distance, wherein the first target cosine distance is greater than the first preset distance; the target classification image corresponding to the first target cosine distance is used as a classification image similar to the image to be identified; according to the target classification The class to which an image belongs determines the authenticity of the image to be recognized.

Optionally, the second determination unit is also used to: calculate a second cosine distance between the deep learning features of the image to be recognized and the deep learning features of each historical image; based on the second preset distance in the second cosine distance Determining a second target cosine distance, wherein the second target cosine distance is greater than a second preset distance; using the target historical image corresponding to the second target cosine distance as a candidate similar image similar to the image to be recognized.

Optionally, the calculation unit is also used to: calculate the first distance between the ORB feature of the image to be recognized and the ORB feature of the candidate similar image; calculate the second distance between the hash feature of the image to be recognized and the hash feature of the candidate similar image Distance: weighted calculation is performed on the first distance and the second distance based on weights to obtain the distance between the image to be recognized and the candidate similar image.

Optionally, the computing unit is also used to: determine the target distance among the distances based on a third preset distance, wherein the target distance is greater than the third preset distance; use the target candidate similar image corresponding to the target distance as a similar image of the image to be recognized .

Optionally, the third determination unit is also used for: if in the similar image, the user corresponding to the target similar image and the user corresponding to the image to be identified are not the same user, then determine that the image to be identified is a false image; if in the similar image , the user corresponding to all the similar images is the same user as the user corresponding to the image to be recognized, then it is determined that the image to be recognized is a real image.

Optionally, the device is also used to: automatically update the first preset distance, the second preset distance, the third preset distance and the weight according to the manual feedback result, wherein the manual feedback result is the recognition image and The target similar images are compared, and the result of whether the obtained image to be recognized and the target similar image are similar images of different users.

Optionally, the device is also used to: acquire the current classification template library.

Optionally, the device is also used to: use a clustering method to cluster images collected within a preset time range to obtain images that appear frequently within a preset time range; The authenticity is judged, and the classification template library is updated according to the judgment result to obtain the current classification template library.

The implementation principles and technical effects of the device provided by the embodiment of the present invention are the same as those of the foregoing method embodiment. For brief description, for the parts not mentioned in the device embodiment, reference may be made to the corresponding content in the foregoing method embodiment.

Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, devices, and units can refer to the corresponding processes in the above-mentioned method embodiments, and will not be repeated here. The image authenticity identification device provided in the embodiment of the present application has the same technical features as the image authenticity identification method provided in the above embodiments, so it can also solve the same technical problem and achieve the same technical effect.

In addition, in the description of the embodiments of the present invention, unless otherwise specified and limited, the terms "installation", "connection" and "connection" should be understood in a broad sense, for example, it can be a fixed connection or a detachable connection , or integrally connected; it may be mechanically connected or electrically connected; it may be directly connected or indirectly connected through an intermediary, and it may be the internal communication of two components. Those of ordinary skill in the art can understand the specific meanings of the above terms in the present invention in specific situations.

If the functions described above are realized in the form of software function units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the essence of the technical solution of the present invention or the part that contributes to the prior art or the part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including Several instructions are used to make a computer device (which may be a personal computer, a server, or a network device, etc.) execute all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage media include: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), magnetic disk or optical disc, etc., which can store program codes. .

In the description of the present invention, it should be noted that the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer" etc. The indicated orientation or positional relationship is based on the orientation or positional relationship shown in the drawings, and is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the referred device or element must have a specific orientation, or in a specific orientation. construction and operation, therefore, should not be construed as limiting the invention. In addition, the terms "first", "second", and "third" are used for descriptive purposes only, and should not be construed as indicating or implying relative importance.

Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present invention, rather than limiting them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: It is still possible to modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the technical solutions of the various embodiments of the present invention. scope.

Claims (10)

1. A method for identifying image authenticity, comprising: Carry out feature extraction on the image to be identified, obtain the deep learning feature, ORB feature and hash feature of the image to be identified; Based on the deep learning features of the image to be recognized, determine whether there is a classified image similar to the image to be recognized in the current classification template library, so as to determine the authenticity of the image to be recognized according to the classified image similar to the image to be recognized sex; If there is no classified image similar to the image to be recognized in the current classification template library, then based on the deep learning features of the image to be recognized, determine a candidate similar image similar to the image to be recognized in the historical image; Calculate the distance between the image to be recognized and the candidate similar image based on the ORB feature of the image to be recognized and the hash feature of the image to be recognized, and determine the candidate similar image in the candidate similar image based on the distance Identify similar images of an image; The authenticity of the image to be identified is determined according to the user corresponding to the image to be identified and the user corresponding to the similar image. 2. The method according to claim 1, wherein, based on the deep learning features of the image to be recognized, determine whether there is a classified image similar to the image to be recognized in the current classification template library, so as to be based on the classification image similar to the image to be recognized Classifying images similar to the image to be identified and determining the authenticity of the image to be identified includes: Calculating the first cosine distance between the deep learning features of the image to be identified and the deep learning features of each classified image in the current classification template library; determining a first target cosine distance among the first cosine distances based on a first preset distance, wherein the first target cosine distance is greater than the first preset distance; Using the target classification image corresponding to the first target cosine distance as a classification image similar to the image to be recognized; The authenticity of the image to be recognized is determined according to the category to which the target classification image belongs. 3. The method according to claim 2, wherein, based on the deep learning features of the image to be identified, determining a candidate similar image similar to the image to be identified in the historical image comprises: Calculating the second cosine distance between the deep learning features of the image to be recognized and the deep learning features of each historical image; determining a second target cosine distance among the second cosine distances based on a second preset distance, wherein the second target cosine distance is greater than the second preset distance; The target historical image corresponding to the second target cosine distance is used as a candidate similar image similar to the image to be recognized. 4. The method according to claim 3, wherein calculating the distance between the image to be identified and the candidate similar image based on the ORB feature of the image to be identified and the hash feature of the image to be identified includes: calculating the first distance between the ORB feature of the image to be identified and the ORB feature of the candidate similar image; calculating a second distance between the hash feature of the image to be identified and the hash feature of the candidate similar image; Weighted calculation is performed on the first distance and the second distance based on weights to obtain a distance between the image to be recognized and the candidate similar image. 5. The method according to claim 4, wherein, based on the distance, determining a similar image of the image to be recognized in the candidate similar images comprises: determining a target distance among the distances based on a third preset distance, wherein the target distance is greater than the third preset distance; The target candidate similar image corresponding to the target distance is used as the similar image of the image to be recognized. 6. The method according to claim 5, wherein determining the authenticity of the image to be identified according to the user corresponding to the image to be identified and the user corresponding to the similar image comprises: If in the similar images, the user corresponding to the target similar image and the user corresponding to the image to be identified are not the same user, then determine that the image to be identified is a false image; If, among the similar images, the users corresponding to all the similar images are the same user as the user corresponding to the image to be recognized, then it is determined that the image to be recognized is a real image. 7. The method according to claim 6, wherein after determining the authenticity of the image to be identified according to the user corresponding to the image to be identified and the user corresponding to the similar image, the method further comprises: The first preset distance, the second preset distance, the third preset distance, and the weight are automatically updated according to the manual feedback result, wherein the manual feedback result is manually The recognition image is compared with the target similar image to obtain a result of whether the to-be-recognized image and the target similar image are similar images of different users. 8. The method according to claim 1, wherein, before performing feature extraction on the image to be recognized, the method further comprises: Obtain the current classification template library. 9. The method according to claim 8, wherein obtaining the current classification template library comprises: Clustering images collected within a preset time range by using a clustering method to obtain images that appear frequently within the preset time range; The authenticity of the frequently appearing images is judged manually, and the classification template library is updated according to the judgment result to obtain the current classification template library. 10. A recognition device for image authenticity, comprising: The feature extraction unit is used to perform feature extraction on the image to be identified, and obtain the deep learning feature, ORB feature and hash feature of the image to be identified; The first determining unit is configured to determine whether there is a classified image similar to the image to be recognized in the current classification template library based on the deep learning features of the image to be recognized, so as to determine based on the classified image similar to the image to be recognized authenticity of the image to be identified; The second determination unit, if there is no classified image similar to the image to be recognized in the current classification template library, based on the deep learning features of the image to be recognized, determine in the historical image that is similar to the image to be recognized Candidate similar images for ; A computing unit, configured to calculate the distance between the image to be recognized and the candidate similar image based on the ORB feature of the image to be recognized and the hash feature of the image to be recognized, and based on the distance, in the candidate similar image determining similar images of the image to be identified; The third determining unit is configured to determine the authenticity of the image to be identified according to the user corresponding to the image to be identified and the user corresponding to the similar image.

Images