Remote Sensing Images Background Noise Processing Method for Ship Objects in Instance Segmentation

Detection and segmentation of ship targets in remote sensing images is a research hotspot in the field of computer vision. However, due to the large coverage area of sea surface remote sensing images, the complex and changeable environment of the ship target, such as cloud interference, coastal buildings, and navigation ripples, the ship causes low detection and segmentation effect. In this paper, we propose an attention module-based method for background noise processing in remote sensing images. To solve the problem of complex background features and noise interference in remote sensing images, this paper introduces an attention module to suppress noise and other interfering features in the complex background by using the channel attention mechanism and spatial attention mechanism, which can enhance the network’s ability to extract object features, and improve the detection and segmentation effect of the network on remote sensing images. Firstly, we introduce Group Convolution into the original Residual Network to enhance the feature representation capability of the model. Secondly, the Swish activation function with better performance in the deep networks is introduced to replace the ReLU activation function in the original Residual Network to improve the accuracy of ship detection and segmentation. Finally, in view of the complex environment of ships in remote sensing images and the problem of noise interference, we introduce an attention mechanism to suppress the interference area and highlight the characteristics of ship areas. The experimental results show that with the improved method, the average accuracy (AP) of ship detection and segmentation has increased from 70.7% and 62.0% to 76.8% and 66.4%, respectively.