Marine Target Segmentation Based on improved DeepLabv3+

Marine target segmentation is of great significance to improve the automation of maritime management. In order to improve the accuracy of marine target image segmentation, a marine target image segmentation algorithm based on improved DeepLabv3+ network is proposed. Firstly, aiming at the problems of image distortion, blur and low contrast caused by weather and motion, Gaussian filtering, brightness enhancement, sharpening and histogram equalization are used to improve the image quality. Then the enhanced marine target image is brought into the improved DeepLabv3+ network for training. The features are extracted by the optimized improved Xception network, and then the multi -scale feature information is captured by the optimal dense prediction cell. In the decoder part, the further fusion of low-level features and high-level features improves the accuracy of the segmentation boundary and achieves the resolution of the original image. The experimental results show that this method can segment the marine target more accurately, and can basically meet the needs of the marine target segmentation task.