Research On Target Tracking Algorithm Of Robot Bionic Eye Based On Deep Learning

In recent years,computer vision technology has the ability to simulate the biological eyes to perceive the external environment.Researchers apply this technology in the field of target tracking,which greatly improves the application value of computer vision technology.However,due to the interference of various environmental factors,as well as the slow speed of online tracking algorithm and the poor accuracy of offline tracking algorithm,higher requirements are put forward for target tracking algorithm.In order to improve the tracking algorithm's ability to deal with various complex scenes,a simple and fast deep learning object detection algorithm YOLOv4 is applied to the tracking algorithm,the main research contents are as follows:1)A tracking algorithm combining detector with tracking trajectory is designed.The detection set is generated by using the YOLOv4 detector,and the depth feature is extracted from the detection results.At the same time,the prediction ability of Kalman filter is used to generate the target trajectory,extract the trajectory features,and generate the candidate set of trajectory.2)Using the results of color feature and local feature extraction and trajectory confidence,a scoring function is designed to select the best detection result and trajectory.The optimal candidate is associated with the tracking trajectory by using the Reid algorithm to update the trajectory.3)Image enhancement is used to preprocess the image,deepen the feature extraction network,and extract the features of the processed image.A long-time tracking model is designed by using Reid,which combines the short-term tracking model based on yolov4 algorithm with the long-time tracking model to realize tracking.4)The optimized target detection and tracking algorithm is applied to the bionic eye system,and the IBVS system is used to control the bionic eye system.Based on two cameras,a camera set is built to realize the target tracking of the bionic eye in a wide area environment.Figure 41;Table 6;Reference 62...