Real-time target detection and tracking in underwater environments based on deep learning

To solve the problems of low precision, insufficient real-time and weak robustness of traditional detection and tracking methods in complex underwater environment, the underwater real-time target detection and tracking based on deep learning is proposed. In this paper, the passive detection model of underwater unmanned submersible is constructed by using the aerial passive sonar buoy array. The adaptive correlation cross location technology of multiple passive sonar buoys is adopted to achieve data pre-processing before concealment detection. The dual threshold random Hough transform tracking before detection algorithm is used to meet the requirement of low detectability of multiple targets under the condition of high underwater noise density. Experimental results show that the detection probability of our method is higher than that of DP-TBD. At the same time, the average running time of the three methods decreases with the increase of SNR. Due to the use of random Hough transform, the timeliness of the method in this paper is obviously better than that of DPTBD and HT-TBD methods.