Research On Moving Object Tracking Algorithm Based On TLD Framework

With the increase in vehicles and the fast pace of manufacture and life,traffic safety accidents caused by fatigue driving frequently occur,posing a huge threat to people’s lives and property safety.Therefore,research on a fast and accurate fatigue driving detection system is of great significance to improve driving safety.Aiming at the problems of low detection accuracy,poor real-time performance,and poor robustness in the existing detection methods,combined with the current popular computer vision technology,this paper studies the algorithm of fatigue driving state detection based on deep learning technology.The research content of this paper is as follows:(1)Research on driver’s facial area location detection algorithm.Aiming at the problem of real-time detection of face region location,a fast face region location detection algorithm with improved Retina Net algorithm is proposed.The algorithm eliminates simple negative anchor points in the shallow detection layer of the image feature pyramid to reduce the search space of the second step classification;adjusts the position and size of the anchor points in the high-level detection layer of the image feature pyramid to improve the second step regression bounding box.A lightweight enhanced receptive field structure based on the idea of Inception Net and Res Net’s short connection structure is designed to select the most effective face region position and effectively detect the face in extreme poses.(2)Research on the key information detection of driver’s face and face tracking algorithm.A lightweight feature extraction unit structure is designed.Based on this unit structure,a backbone network structure for face key information detection and face tracking is designed.The key face information detection includes face key point detection and head pose estimation.According to the idea of multi-task learning,the face key information detection task and the face tracking task are jointly learned using the same network,which improves the detection accuracy of the lower key information under the condition of posture change.The heat map generated from the coordinates of the key points of the face is shared as a feature in the face tracking branch and the head pose estimation branch,so that more accurate head pose estimation and higher face tracking can be obtained under the premise of ensuring the overall running speed of the model Accuracy.(3)Research on driver fatigue detection algorithm.Aiming at real-time driver fatigue recognition,this paper designs a fatigue recognition algorithm based on multi-feature spatiotemporal fatigue feature sequence.The algorithm extracts multiple facial fatigue features,including eye,mouth,and head posture,stitches facial fatigue feature vectors of multiple frames of images in the video to construct a spatiotemporal fatigue feature sequence,and designs a fatigue detection network based on the long-and short-term memory network structure to perform fatigue classification judgment of the state.It can be seen from experiments that the accuracy of the fatigue detection algorithm designed in this paper is as high as 92.1%,it can also meet the real-time requirements of fatigue driving detection.