Research And Application Of Target Tracking Method For Panoramic Tracking Vehicle

Mobile tracking vehicles are widely used in security monitoring,logistics and transportation.In order to improve the tracking performance of the mobile vehicle,a panoramic tracking vehicle based on multi-cameras is designed and the key issues are thoroughly studied in the thesis,the main contents are as follows:Firstly,an improved template matching method is proposed for the tracking failure caused by the traditional template matching method when the target rotates during the tracking process.The gray centroid and gray neighborhood difference centroid of the template image and unmatched images are calculated,the gray reference direction and the gray neighborhood difference reference direction are obtained.At the same time,the grayscale matching value and the corresponding grayscale neighborhood difference matching value of the template image and the unmatched images are calculated.The calculation results are weighted as a new matching value,and the weighting coefficients are updated according to the criterion that the smaller of matching value,the larger of weighting coefficient.Experimental results show that the improved method has good adaptability to solve the tracking problem when target is in rotation.Secondly,for the target occlusion problem,combing Camshift algorithm,a Kalman filter tracking strategy based on neural network is proposed.In the process of target tracking,the Bhattacharyya coefficient is used to determine whether the target is occluded.When occlusion occurs,the prediction position obtained by BP neural network is used as the measurement value,and Kalman filtering is used to give the optimal estimation of the target.When the target moves away from the occlusion area,Camshift algorithm is used to track the target.The simulation results show that the improved method can still accurately track the target even if the target is frequently occluded or blocked by fast motion.Finally,an improved ORB feature matching method is proposed based on the research of feature point matching algorithm.ORB algorithm only calculates the size coding of random points in the neighborhood of the feature points,in order to improve the matching accuracy,the amplitude comparison information of random point pairs are added on the basis of ORB algorithm.The Hamming distance is respectively used to calculate the size and amplitude distance of feature points,and the two distance values are weighted and fused as a new distance value for feature point matching.The experimental results show that the improved method has a good matching effect.The hardware construction of the panoramic tracking vehicle is completed in the thesis,and proposed improved methods are applied to the panoramic tracking vehicle for experimental testing.The effectiveness of the improved methods have been proved in the experiments.