A Lightweight Underwater Fish Body Image Segmentation Algorithm Based on Co-PSPNet
To solve the problems of image blurring and limited resources in underwater environments,a lightweight underwater fish segmentation algorithm based on the Co-PSPNet network is proposed.The main network employs MobileNetV2 to reduce compu-tational complexity and parameter count.The CoordConv module is introduced to enhance the network to the perception of detailed in-formation like fish boundaries.The features obtained after the global pooling are used as inputs for an attention mechanism network,enhancing the features with higher semantic information.Through the extensive experimental evaluation,the algorithm has a superior performance of publicly available underwater fish image datasets.In practical applications,this algorithm meets the engineering re-quirements of underwater fish image segmentation in the domains of underwater ecology research and underwater robotics.By solving the image blurring and resource limitations in underwater environments,the proposed algorithm provides an effective solution for im-age segmentation in underwater ecology research and underwater robotics applications.
万方数据
维普
