Research On Vehicle Tracking And Recognition Algorithms Based On Active Infrared Surveillance Camera

Visible target detection method is ineffective in the night,while active infrared camera can still obtain the target imaging due to its own emitted infrared light reflected by the target imaging.With the increasing scale of vehicles,it is necessary to control vehicle targets in both military and civilian situations.The active infrared surveillance camera is convenient to carry and install.And the camera has fewer restrictions and high flexibility when selecting the installation location.This paper will study the vehicle tracking and recognition algorithm technology based on active infrared survelliance camera.This paper studies the detection,tracking and recognition technology of vehicle target.A particle filter algorithm based on image criterion is proposed in the detection and tracking stage.Firstly,the image is preprocessed and enhanced.Then the image criterion is introduced and the state equation of particle filter is improved by combining multi-frame information.Combine image information in the equation of state to improve the anti-jamming capability of the algorithm.In the resampling stage,the threshold value is set by the particle weight,and the new particle affected by the small weight particle and the large weight particle.It can not only restrain the concentration of particle weight,but also ensure the diversity of particles.Experimental results show that the proposed algorithm is more accurate and anti-interference than the traditional particle filter.In the stage of vehicle target recognition,firstly,a large number of images are collected by active infrared surveillance camera,and the samples are selected to construct the infrared sample library.The sample library constructed in this paper contains the vehicle target and the sample unrelated to the vehicle target.Secondly,the Hu moments and Fourier characteristics of the vehicle target are analyzed and the non-target is classified by SVM support vector machine.A Faster r-cnn method based on fast NMS was proposed to simplify the calculation process of non-maximum suppression.By changing the calculation method of anchor point offset,the speed of acquiring the final region of the target can be improved.Through the experimental analysis,the image classification process can be completed correctly,and can be applied to track the dynamic vehicle targets captured in the process.