Underwater Acoustic Target Tracking Method

Acoustic detection is one of the main underwater detection methods at present.Therefore,the target detection and tracking method based on acoustic and visual images is of great significance for underwater vehicles to complete tasks such as obstacle avoidance and tracking.Due to the complexity of the underwater environment and the limitation of the multi-beam forward-looking sonar imaging principle,the underwater acoustic visual image loses a lot of target feature details,and at the same time there is a lot of noise in the image,the current detectors inevitably generate false alarms and leaks.Therefore,the following exploratory work is carried out in this paper:1.Researched on the acoustic visual target detection method based on CFAR.The characteristics of the constant false alarm underwater acoustic and visual target detection method are discussed,and the latent acoustic visual target model is constructed,which improves the target detection ability of the constant false alarm detection method under the condition that the underwater acoustic and visual image background does not conform to the Rayleigh distribution,reduces the number of targets.False alarm targets in the context of non-Rayleigh distributions.And by combining the region growing method,the acoustic visual target detection results are obtained.2.Researched on a deep learning-based acoustic visual target detection method.Since the accuracy of the traditional method is greatly affected by the background noise,the application effect of the Mask-RCNN target detection method in underwater acoustic visual images is analyzed,and the reasons for the missed detection of the detector are analyzed.The method improves the accuracy of object detection and improves the recall.The influence of the depth of the backbone network on the detection results is studied,and the detection effect of the MaskRCNN detection method on underwater acoustic and visual targets is improved by improving the backbone network and the non-maximum suppression process.3.Researched on the SORT tracking method of underwater acoustic targets.The effect of the target detection method on underwater acoustic visual targets and the failure form of the SORT tracking method are discussed,and the influence of the detection effect on the underwater target tracking results is analyzed.Aiming at the high recall rate of the constant false alarm detection method and the high precision rate of the Mask-RCNN detection method,a target tracking method combining the two is proposed,and compared with the SORT tracking method,the effectiveness and accuracy.