Vehicle-mounted face snapshot method and system

Abstract

The invention relates to the technical field of vehicle-mounted face snapshot, and in particular relates to a vehicle-mounted face snapshot method and system. The method comprises the steps of S1 setting a face snapshot camera in a vehicle; S2 shooting a pedestrian face image through a face snapshot camera, and carrying out face detection and face tracking; S3 obtaining vehicle position information and vehicle state data in real time, and judging whether a vehicle is parked or not; S4 if it is judged that the vehicle is parked, further judging the last deceleration before the vehicle is parked, controlling the face snapshot camera to shoot towards the vehicle door within the time period from the beginning of the last deceleration to the opening of the front door of the vehicle, and executing the step S2; otherwise, controlling the face snapshot camera to shoot towards the front of the vehicle, and going to the step S2. Aiming at a vehicle-mounted application scene, the operation performance and the real-time performance are optimized, the cost of the face snapshot device is reduced, the snapshot device is linked with a bus, and the coverage rate of security arrangement is greatly improved through movement of the bus.

Description

technical field [0001] The invention relates to face recognition technology, in particular to a vehicle-mounted face capture method and system. Background technique [0002] With the rise of the field of artificial intelligence, in recent years, machine learning and deep learning technology have been greatly developed, and face recognition technology, one of the key technologies, has gradually matured, and face detection, tracking and recognition have also been in It is widely used in mobile phones, servers and other terminal devices. [0003] At this stage, due to security needs, it is necessary to install face capture cameras in major buses for face capture and data comparison to complete the security layout. However, the specific technical process of the existing vehicle-mounted face capture method and device is still not perfect, and no complete and available vehicle-mounted face capture method has been proposed. [0004] Existing face tracking methods usually include ...

AI Technical Summary

Problems solved by technology

However, the specific technical process of the existing vehicle-mounted face capture method and device is still not perfect, and no complete and usable vehicle-mounted face capture method has been proposed.

[0004] Existing face tracking methods usually include frame-by-frame detection face tracking methods, camshift and Kalman filter tracking methods, full convolutional neural network face recognition tracking methods, etc. If such tracking methods are directly applied to vehicle capture scenes The best performance, or unqualified, the reason is as follows: the frame-by-frame detection face tracking method updates the face position of each frame by repeatedly performing all the face detection operations on each frame image, and the calculation load is large. When multiple people appear at the same time, it consumes more computing performance; the tracking method using camshift and Kalman filter can have a faster convergence speed through position estimation and centroid iteration when the moving speed of the target is slow and the scale change is small. However, when the capture device is facing the front of the car, the moving speed of the face target is faster, and because the frame number of the capture device is generally not too high, the target deviation between two frames is relatively large, which leads to an increase in the number of iterations of the algorithm. When there are many pedestrians, it will undoubtedly greatly increase the load of the embedded system and reduce the real-time performance; and the face recognition and tracking method of the full convolutional neural network has been well applied in many scenes, but in order to make its calculation On the arm platform, it is necessary to reduce its network model and reduce its complexity, which means that more vehicle scene face samples are needed to improve its accuracy, that is, a larger amount of pre-scaled normalization and labeled training is required samples, and based on the current application environment, these samples are not easy to obtain with the existing face capture technology

[0005] In other words, the existing face capture and face tracking technologies cannot be fully adapted to vehicle-mounted application scenarios, and usually require higher-performance hardware to run, resulting in higher device costs, or the development of conditions that do not currently exist ; and only consider the independent operation of the camera, but not in conjunction with the vehicle application, resulting in a waste of capture device resources; also did not take into account the cost control of the equipment on the vehicle, and the direction of the linkage between the equipment and the vehicle to achieve coordinated applications

Images