| Under the overall promotion of deep learning technology,underwater image processing technology has been rapid development.As one of the underwater image processing methods,underwater target detection algorithm plays an increasingly important role in daily life.With the continuous pursuit of lightweight,low-cost and high-performance,it is of great significance to design a set of high-performance,low-cost,edge computing platform based underwater image processing system for underwater multi-target real-time detection.The main contents of this paper are as follows:First of all,analyze the actual system requirements,according to the requirements of the system overall structure design and system overall process design.An improved underwater image enhancement algorithm is proposed.According to the underwater image evaluation index,it is compared with other underwater image enhancement algorithms.Then,it optimizes the YOLO series of target detection algorithms from three aspects.According to the performance index of the target detection algorithm,it analyzes the performance of the algorithm after optimization from two aspects of individual optimization and overall optimization,and proves that the Yolo series optimization scheme is feasible.Next,compares and analyzes the results of tiny-YOLOV4,improved YOLOV3,YOLOV4,improved YOLOV4,YOLOV3.Combined with the actual results of comparison analysis.Finally,according to the system requirements,the edge computing platform is designed,and the underwater multi-objective detection algorithm and edge computing platform are combined to build an underwater image processing system based on edge computing platform.The performance results are analyzed to determine that the system meets the actual system requirements.The actual performance of the system based on edge computing platform and other performance equipment is analyzed and compared,and the actual performance analysis of underwater image processing system based on edge computing platform is obtained. |
PDF Full Text Request