CN114495204A - A dual-terminal passenger authentication method for autonomous vehicles - Google Patents

Abstract

The invention belongs to the field of automatic driving identity authentication, and specifically discloses a double-end passenger identity verification method for an automatic driving vehicle, comprising the steps of: verifying and preprocessing passenger face information, and the preprocessing process includes face information collection, detection and inspection; Save the original image of the passenger's face to the server; the face inference framework service will convert the newly uploaded passenger's face photo into face vector data in real time and return it to the database; The face vector data is pushed to the in-vehicle system of the corresponding order-taking vehicle; the identity verification of the passenger is carried out after the passenger gets on the bus. The present invention can accurately identify the real identity of the passenger through face recognition, prevent the taxi from being inconsistent with the passenger, the face recognition speed is fast, and the convenience of user operation and user experience are improved.

Description

A dual-terminal passenger authentication method for autonomous vehicles

technical field

The invention relates to the field of automatic driving identity authentication, in particular to a double-end passenger identity verification method for an automatic driving vehicle.

Background technique

Identity authentication technology has been developing continuously in recent years, especially the technology of verifying user identity based on the inherent biological characteristics of human beings has been widely used in various fields. A new ride-hailing model with safety officers and self-driving vehicles as an online car-hailing model is undergoing road tests in the experimental area. The identification of passengers is critical to vehicle safety.

At present, the self-driving online car-hailing is still in the research and development stage, and it cannot ensure the rapid identification of users on complex public transportation roads, that is, it cannot realize convenient, safe and efficient passenger authentication processing. The identity verification method can accurately complete the identity verification and identification processing, which not only ensures the efficiency, but also ensures the accuracy of the identification.

SUMMARY OF THE INVENTION

The purpose of the present invention is to provide a dual-terminal passenger identity verification method for an automatic driving vehicle, so as to solve the problems raised in the above-mentioned background art.

In order to achieve the above purpose, the present invention provides the following technical solutions: a double-end passenger identity verification method for an automatic driving vehicle, comprising the steps of:

S1: Verification and preprocessing of passenger face information, the preprocessing process includes face information collection, detection and inspection;

S2: Save the original image of the passenger's face to the server;

S3: Face inference framework service, which will convert the newly uploaded passenger face photos into face vector data in real time and save them to the database;

S4: After the passenger places an order successfully, the server pushes the order and the passenger's face vector data to the vehicle-machine system corresponding to the vehicle receiving the order;

S5: After the passenger gets on the bus, the identity verification of the passenger is performed.

Preferably, S1 specifically includes:

S101: Call the mobile phone camera to take a picture, and obtain multiple face photos of the passenger;

S102: Obtain a face photo and use a multi-task cascaded convolutional neural network algorithm for face detection;

S103: Based on the MINIFASNET in vivo silence algorithm, perform in vivo inspection on passengers to ensure the operation of real people.

Preferably, when S2 is sent to the server, a signature string sign is added to the message to increase security, and the signature algorithm adopts SHA256, which is transmitted to the server by converting the picture into a secondary byte stream, and saved to the database. .

Preferably, the data push in S4 adopts a signature algorithm based on SHA256.

Preferably, S5 specifically includes:

S501: Obtaining the face photos of passengers getting on the bus by taking pictures and collecting the camera and/or externally installed cameras;

S502: Use the multi-task cascaded convolutional neural network algorithm to detect the face of passengers getting on the bus;

S503: Perform liveness detection to ensure human operation;

S504: Extract the mobile face feature string based on the MobileFaceNets algorithm;

S505: Collect and extract the current feature vector of the passenger, compare and verify with the face collected by the mobile terminal and the feature vector calculated by the server, and complete the final verification and verification of the passenger's identity.

Compared with the prior art, the beneficial effects of the present invention are:

The present invention introduces face recognition into the field of identity recognition of automatic driving, and can accurately identify the real identity of passengers through face recognition, preventing inconsistency between taxis and passengers, and the speed of face recognition is fast, which improves the convenience of user operation. user experience; and the method logic of the present invention is that the original face image of the passenger is transmitted through the vector, which greatly enhances the security, and is not easy to leak the user's private information. Supervision and real-name rides.

Description of drawings

FIG. 1 is a flow logic block diagram of the method of the present invention.

Detailed ways

The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

Referring to FIG. 1, the present invention provides a technical solution: a method for dual-terminal passenger identity verification of an autonomous vehicle, comprising the steps of:

S1: Verification and preprocessing of passenger face information, the preprocessing process includes face information collection, detection and inspection;

S2: Save the original image of the passenger's face to the server;

S3: Face inference framework service, which will convert the newly uploaded passenger face photos into face vector data in real time and save them to the database;

S4: After the passenger places an order successfully, the server pushes the order and the passenger's face vector data to the vehicle-machine system corresponding to the vehicle receiving the order;

S5: After the passenger gets on the bus, the identity verification of the passenger is performed.

In this embodiment, S1 specifically includes:

S101: Call the mobile phone camera to take a picture, and obtain multiple face photos of the passenger;

S102: Obtain a face photo and use a multi-task cascaded convolutional neural network algorithm for face detection;

S103: Based on the MINIFASNET in vivo silence algorithm, perform in vivo inspection on passengers to ensure the operation of real people.

In this embodiment, when S2 is sent to the server, a signature string sign is added to the message to increase security, the signature algorithm adopts SHA256, and the image is converted into a secondary byte stream and transmitted to the server, and Save to database.

In this embodiment, the data push in S4 adopts a signature algorithm based on SHA256.

In this embodiment, S5 specifically includes:

S501: Obtaining the face photos of passengers getting on the bus by taking pictures and collecting the camera and/or externally installed cameras;

S502: Use the multi-task cascaded convolutional neural network algorithm to detect the face of passengers getting on the bus;

S503: Perform liveness detection to ensure human operation;

S504: Extract the mobile face feature string based on the MobileFaceNets algorithm;

S505: Collect and extract the current feature vector of the passenger, compare and verify with the face collected by the mobile terminal and the feature vector calculated by the server, and complete the final verification and verification of the passenger's identity.

In this embodiment, the MobileFaceNets used in S504 have the advantage of comparing with several other high-efficiency CNN architectures such as MobilenetV1, ShuffleNet, and MobileNetV2 that the extraction effect is more than 2 times faster than that of other models under the same environment. . When a single size is less than 5MB, the analysis accuracy of MobileFaceNets on LFW and AgeDB 30 is more than 95%, and the calculation time on the mobile phone can be about 20 milliseconds, and finally a face vector with 128 values is obtained.

In this embodiment, the face feature vector similarity comparison algorithm uses the cosine similarity algorithm to compare, and obtains a score after comparison. The higher the score, the higher the similarity, and it can be regarded as the same person, and the identity verification.

The cosine of the angle β between two vectors can be expressed as:

Cosβ=a×b/|a|×|b|

In this embodiment, assuming two vectors a=[2,2,3] and b=[3,3,4], the cosine similarity of the angle β between the two vectors is calculated as follows: Substitute b into the formula:

The numerical range of cosine similarity is also the range of cosine values, namely [-1,1]. The higher the value, the greater the similarity. Generally greater than 0.7 can be considered to be the same person. A vector of actual face alignments contains 128 items of data.

The embodiment of the present invention extracts MobileFaceNets through multi-task cascaded convolutional neural network MTCN+living detection MINIFASNET+feature string, and finally uses cosine similarity algorithm for comparison, forming the architecture system of the present invention for face recognition. At the same time, it is possible to collect faces from the mobile phone, place orders, and identify passengers on the unmanned vehicle terminal.

Although embodiments of the present invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, and substitutions can be made in these embodiments without departing from the principle and spirit of the invention and modifications, the scope of the present invention is defined by the appended claims and their equivalents.

Claims (5)

1. a self-driving vehicle double-end passenger identity verification method, is characterized in that, comprises the steps: S1: Verification and preprocessing of passenger face information, the preprocessing process includes face information collection, detection and inspection; S2: Save the original image of the passenger's face to the server; S3: Face inference framework service, which will convert the newly uploaded passenger face photos into face vector data in real time and save them to the database; S4: After the passenger places an order successfully, the server pushes the order and the passenger's face vector data to the vehicle-machine system corresponding to the vehicle receiving the order; S5: After the passenger gets on the bus, the identity verification of the passenger is performed. 2. a kind of self-driving vehicle double-end passenger identity verification method according to claim 1, is characterized in that, specifically comprises in described S1: S101: Call the mobile phone camera to take a picture, and obtain multiple face photos of the passenger; S102: Obtain a face photo and use a multi-task cascaded convolutional neural network algorithm for face detection; S103: Based on the MINIFASNET in vivo silence algorithm, perform in vivo inspection on passengers to ensure the operation of real people. 3. a kind of self-driving vehicle double-end passenger identity verification method according to claim 1 is characterized in that, when sending to server in described S2, in message, increase a signature string sign to increase security, and signature algorithm adopts SHA256, which is transmitted to the server by converting the image into a secondary byte stream, and saved to the database. 4. A kind of self-driving vehicle double-end passenger identity verification method according to claim 1, is characterized in that, the data push in described S4 adopts the signature algorithm based on SHA256. 5. a kind of self-driving vehicle double-end passenger identity verification method according to claim 1, is characterized in that, in described S5, specifically comprises: S501: Obtaining the face photos of passengers getting on the bus by taking pictures and collecting the camera and/or externally installed cameras; S502: Use the multi-task cascaded convolutional neural network algorithm to detect the face of passengers getting on the bus; S503: Perform liveness detection to ensure human operation; S504: Extract the mobile face feature string based on the MobileFaceNets algorithm; S505: Collect and extract the current feature vector of the passenger, compare and verify with the face collected by the mobile terminal and the feature vector calculated by the server, and complete the final verification and verification of the passenger's identity.

Images