Design Of Anti-collision Door Locking Protection System For Vehicle Network Vehicle Terminal

With the continuous upgrade of Internet of thingsand hardware platform,and the advent of a new generation of 5G communication technology,how to make the car more safe and reliable,how to make the response speed of the automobile safety system faster and more efficient,and how to fully improve the automatic obstacle avoidance ability have gradually become an important index to measure the overall performance of a vehicleon the basis of more and more stringent automobile safety performance norms.Moreover,with the development of science and technology,image processing technology is more and more widely used in various fields,such as medical treatment,transportation,military,agriculture and so on.The expansion of such application field and the diversification of the application scene make the image processing efficiency have strict requirements,and also promote the further development of image processing technology,hardware performance improvement combined with algorithm optimization thus becoming an effective method to meet the real-time and efficient requirements.This paper designs an anti-collision door locking protection system based on binocular stereo vision and dynamic image processing technology.The main purpose is to monitor the rear target of the vehicle in real time through the camera on the rearview mirror,and measure and predict the speed of the target so as to judge whether it is possible to cause a collision led by ignoring the rear vehicle and opening the door mistakenly.The system uses the method of ARM + FPGA as a whole.The combination of hardware and software algorithms improves the speed of image recognition and processing,thereby improving the overall real-time performance of the system and the safety performance of the collision avoidance system.An image preprocessing method based on OpenCL standard and FPGA acceleration is proposed,which improves the speed of traditional image grayscale processing by 47 times by taking advantage of FPGA's highly parallel feature.Based on the comparative experimental analysis of the single inter-frame difference method and the background difference method,a dynamic target recognition method based on the combination of inter-frame difference method and background difference method is proposed,which improves the integrity and clarity of the contour feature acquisition of the target object,and reduces the misjudgment rate of the system as a whole.According to the demand of real-time recognition,the appropriate binocular camera calibration algorithm and target ranging algorithm are adopted to identify and measure the simulated vehicle target and output the target coordinate and distance.Finally,the simulation target recognition and ranging experiments are carried out many times,and the measured target depth distance is compared with the actual distance,and the error rate is small,which achieves the expected design effect and meets the actual requirements.