Abstract: In view of the problem of low detection accuracy of small targets in visible-light videos collected by network cameras and rotary-wing drones in complex backgrounds, a small target detection and recognition method based on improved YOLOv5 is proposed. By introducing strategies such as the SPD-Conv module, coordinate attention (CA) module, transposed convolution, optimized anchor boxes, and an improved Alpha - IoU loss function into the YOLOv5 architecture, the detection ability for small targets can be effectively enhanced. Experimental results demonstrate that the proposed improved YOLOv5 has the ability to detect small targets. The precision, recall rate, and mAP indicators can reach 82.14%, 75.96, and 82.41% respectively. The inference speed is less than 25 ms, and the number of parameters is less than 40 M. The experimental results verify the effectiveness and practicality of the proposed method, which has broad application prospects.