Research On Technique Of Vehicle Assistant Driving Based On Infrared Video

In recent years,with the large increase number of motor vehicles being used,vehicle-assisted driving technology plays an increasingly important role in the field of traffic safety.A large number of statistical results show that most traffic accidents happen at night,so night-assisted driving technology plays a vital significance role on road traffic safety.Infrared thermal imaging technology break the limitations that visible light equipment can not be imaging at night due to dim.The maneuvering target detection and recognition technology in the visible image is improved to realize the recognition of maneuvering target in the infrared video stream.The imaging principle of infrared image and the characteristics of infrared image is studied and the image denoising and image enhancement algorithm based on the characteristics of infrared image is researched to improve the traditional median filter algorithm and the traditional histogram equalization algorithm.In the stage of target detection and localization,the algorithm of target detection based on gray-scale and the algorithm of target detection based on wavelet transform are studied,and a new method of combining gray-scale and wavelet theory is proposed to detect the moving target.Then,a target localization algorithm based on gray features and a target localization algorithm based on object center of gravity are studied.After a large number of experiments,the target localization algorithm based on the center of gravity of object is selected to realize the location of moving target.In the feature extraction stage,a feature of Haar-LBP that combined with Haar-like feature and LBP feature is proposed to extract the feature of maneuvering target.Finally,the Fisher discrimination is used to classify the features of the weak classifier,and a strong classifier model based on FDR-Ada Boost training mechanism is proposed,and the moving target(pedestrians and vehicles)can be identified in infrared video.A large number of experimental results show that the system can efficiently realize real-time recognition of moving target in infrared video.