Research On Target Recognition Feature Analysis And Tracking Algorithm In Complex Scenes

Target recognition tracking in infrared images faces great difficulties due to the influence of illumination and occlusion in the complex scenes.In the first part of this thesis,various features were extracted from typical targets and interference scenes.Feature selection methods such as F-score and PCA were used to select features and classification experiments were carried out,which can provide support for the design of the target detection and recognition algorithm.In the second part of this thesis,we proposed a target region extraction method based on optical flow algorithm and Superpixel segmentation,and a target region extraction method based on ViBe algorithm and frame difference method to improve the extraction accuracy of target region in the occlusion environment.In the third part of this thesis,an anti-occlusion moving target tracking algorithm based on correlation filters was proposed.First,the target region extraction method based on optical flow algorithm and Superpixel segmentation is used to extract the target region,which minimizes the interference of background points and improves the accuracy of target feature extraction.Secondly,we establish a feature expression template using the HOG feature and gray features to determine the target position.Then,when updating the tracking template,we update the correlation filtering template and gray histogram independently and set the occlusion judgment mechanism,which reduces the impact of the complexity and uncertainty of the actual scene on the target tracking.In order to verify the generalization performance of the proposed algorithm,we conducted tracking experiments on the actual infrared data collected in different weather and different scenes.The sample data contains many common interferences encountered by moving targets.On this dataset,we compared the tracking effect of the anti-occlusion algorithm and other classical tracking algorithms.The results show that the anti-occlusion moving target tracking algorithm has good performance in tracking accuracy and tracking success rate,and show good generalization performance.