Water level detection based on ghost convolution and CA attention mechanism

Water level monitoring is the key issue of water conservancy construction, timely grasp of water conditions, flood prevention is important for social development, for this reason we propose an intelligent image water level recognition algorithm: the introduction of ghost convolutional backbone network optimization, adding GA attention mechanism of yolov5's deep learning water level recognition algorithm, the training of the collected images, to achieve the automatic identification of water level. Experimental results show that the improved yolov5 deep learning water level recognition algorithm can effectively is to recognize the water level, the mAP value reaches 97.6%, compared with the unimproved version to improve 5.2%, in line with the engineering water level monitoring error requirements.