Multi-rotor UAV Detection Algorithm Based on the Improved Yolov5

An improved method for enhancing current multi-rotor UAV detection algorithms is presented here with the introduction of an upgraded Yolov5 multi-rotor UAV detection algorithm. Initially, a collaborative attention mechanism is integrated into the Yolov5 backbone to enhance the expression capability of the targeted object within the image and diminish any disruptive background noise information. Subsequently, adaptive feature fusion technology is included in the primary model's section to facilitate the integration of feature maps with varying spatial resolutions. Finally, the original bounding box regression loss function in Yolov5 has been fine-tuned by substituting it with the SIoU loss function. A new Angle loss parameter has been incorporated into the prior loss function metric to restrict the movement freedom of the predictive frame during training, thereby refining the accuracy of relocating the predictive frame.