Research On Key Technologies Of Vehicle Safety Anti-collision Warning Based On Blind Spot Images

With the increasing number of car ownership,traffic safety has become a problem that cannot be ignored.As an important means to solve this problem,advanced assisted driving technology has been widely studied by scholars at home and abroad.As a basic function of advanced assisted driving technology,the blind area vehicle safety anticollision warning is of great importance to realize applications such as lane change and collision avoidance.Therefore,based on the embedded platform with low price and low computing efficiency,this thesis studies the key technologies of safe collision avoidance warning of vehicles in blind areas.The main research work is as follows.This thesis presents the overall analysis and design of the system.Firstly,the overall requirements of this thesis are introduced,and then the requirements analysis is carried out for blind area vehicle detection and blind area dynamic obstacle detection respectively.Finally,for different application scenarios,this thesis uses the two-frame difference method to switch between the blind area vehicle detection algorithm and the blind area dynamic obstacle detection algorithm to realize the algorithm switching of the system in different scenarios.Aiming at the problem of poor real-time detection of traditional blind area vehicle detection algorithms on embedded platforms,this thesis studies and realizes an improved CAda Boost blind area vehicle detection algorithm.Aiming at the problem of the high dimension of Haar-like features extracted by CAda Boost algorithm,under the premise that the detection accuracy of the algorithm remains unchanged,this thesis uses feature compression to improve the detection speed of the algorithm.For night scenes,this thesis uses Gamma correction to enhance the image of the blind area at night.When tested on Raspberry Pi 4B,the detection speed of the improved CAda Boost algorithm proposed in this thesis is 2.5 times faster than that of the original CAda Boost algorithm,and the detection accuracy of the algorithm is more than 95%,which proves the validity and robustness of the algorithm in this thesis.Through the analysis of the traditional dynamic obstacle detection algorithm,for the special scene of drivers waiting for traffic lights,this thesis studies and implements a blind area dynamic obstacle detection algorithm based on the three-frame difference method.First,preprocess the blind area image of the vehicle,and then use the threeframe difference method to perform two differences on the preprocessed adjacent three frames of images,and calculate the logical AND operation of the two difference images to obtain the target image,and then perform the target image morphological postprocessing.When testing on the Raspberry Pi 4B,the detection accuracy of the blind area dynamic obstacle detection algorithm is more than 80%,and the detection speed basically meets the real-time requirements.This thesis integrates the blind area vehicle detection algorithm and the blind area dynamic obstacle detection algorithm into the system for testing.In the actual road scene test,the detection accuracy of the integrated blind area vehicle detection algorithm and the blind area dynamic obstacle detection algorithm is basically the same as that of the separate algorithm test.