Research On Target Tracking Methods Based On SURF Features

The application of target detection and tracking technology can reduce people's work intensity and has strong practicability in life.But in the process of tracking,there are many disturbing factors affecting the accuracy of tracking effect.In this paper,we improved the traditional SURF feature extraction algorithm and designed a target tracking algorithm that combined the traditional Camshift algorithm with the improved SURF feature.SURF algorithm was improved to obtain rough matching points through feature extraction and filter invalid feature points.Finally,the algorithm is combined with Kalman filtering prediction algorithm to achieve accurate and stable target tracking in a complex environment.The main work of this paper is as follows:(1)In this paper,traditional SURF algorithm was improved in three aspects: feature detection,feature description and feature matching.The SURF algorithm that integrated bilateral filtering and Canny edge detection improved the accuracy of feature point matching,and the use of DGHM method enhanced the stability of target detection.SURF algorithm was improved for target detection to obtain coarse feature points,and the normalized product correlation algorithm(NCC)was used to filter out useless feature points.(2)In this paper,we mainly studied a target tracking algorithm based on improved SURF feature.Meanshift algorithm was used to iterate every frame of image to realize tracking.In the process of tracking feature detection,a large number of feature points must be extracted at every moment.Combined with this algorithm,Kalman filter can track the target and only detect the target features in the prediction area,which greatly reduces the computation of feature detection in the tracking process.Moreover,Kalman filter can effectively solve the tracking problem when the target is blocked and ensure the continuous and stable tracking of the target.(3)In this paper,vehicle tracking is taken as a class.By comparing the tracking results of the traditional Camshift algorithm and the improved algorithm in a complex environment through experiments,it is proved that the algorithm proposed in this paper has a good feasibility in tracking.