Foggy water surface target detection model based on joint optimization

In haze scenes, the image quality of water surface images will suffer from varying degrees of contrast reduction and blurred edge information. Therefor, a Dehazing Target Detection Algorithm Based on Joint Optimization is designed in this paper. Firstly, based on the ability of Pyramid Squeeze Attention Block and Coordinate Attention in extracting multi-scale features and positional correlation information, Feature Fusion Attention Network is improved to enhance the multi-scale feature aggregation ability of the dehazing network. Secondly, a skip connection operation with residual edges is introduced in the feature aggregation module to enhance the feature recovery ability of the dehazing network. Then, Depthwise Separable Convolution is used to optimize the ordinary convolution of the backbone of the dehazing network to speed up the inference speed of the dehazing network. Finally, a 1x1 convolutional block is introduced before the Convolutional Block Attention Module to enhance the ability of the object detection network to detect small targets. The experimental data sets include the WSFOGVisDrone and WSFOGVOC data sets with artificially added haze. The average accuracy of our algorithm on WSFOGVisDrone reaches 41. 18%, which is 3% higher than the 38. 18% of the original detection algorithm. In addition, the average accuracy on WSFOGVOC reaches 89. 95%, which is 4. 11% higher than the 85. 84% of the original detection algorithm. The experimental results verify that the algorithm can effectively improve the detection accuracy of small targets in hazy scenes, and the algorithm has better applicability and robustness.