水电能源科学

为准确检测水下水利设施的裂痕，提出一种多注意力机制融合YOLOX的裂痕检测算法。首先，分析水下无人潜航器(UUV)携带的前视声呐成像原理，了解声呐采集水利设施图像的特点。然后，将注意力机制引入YOLOX算法，在Focus网络中添加CA注意力机制，在SPP模块中引入SE注意力机制，改进YOLOX算法裂痕检测性能。最后，基于双闭环控制策略实现UUV对水下水利设施的全覆盖观测。试验结果表明，UUV基于多注意力机制融合的YOLOX算法和双闭环控制策略能对水下水利设施进行全覆盖检测，且能够提高水下设施裂痕的检测精度。

To accurately detect cracks in underwater hydraulic structures, a crack detection algorithm that integrates a multi-attention mechanism with YOLOX(You Only Look Once X) is proposed. Firstly, the imaging principle of the forward-looking sonar carried by UUVs is analyzed to understand the characteristics of sonar images of hydraulic structures. Then, the attention mechanism is introduced into the YOLOX algorithm by adding a Coordinate Attention(CA) mechanism in the Focus network and incorporating a Squeeze-and-Excitation(SE) attention mechanism in the SPP module, improving the crack detection performance of the YOLOX algorithm. Finally, a double closed-loop control strategy is employed to achieve comprehensive observation of hydraulic structures by the UUV. The algorithm's performance is tested, and the results show that the UUV-based the YOLOX algorithm integrated with multi-attention mechanisms and the double closed-loop control strategy can comprehensively detect hydraulic structures and improve the accuracy of underwater facility crack detection.