Research on Lightning Target Detection in Complex Scenes Based on RCC-YOLOv7

Lightning is a common natural phenomenon, and monitoring lightning in an area plays a crucial role in analyzing the characteristics of lightning activity and conducting lightning hazard mitigation. The RCC-YOLOv7 algorithm is proposed to address the misdetection and omission problems of lightning weather video detection. First, the RepNCSPELAN4 feature extraction module is introduced to enhance the feature extraction of lightning dataset under strong noise, and the main convolution in RepNCSPELAN4 is improved into CoordConv to make the convolution operation location-sensitive, which incorporates the location information as an extra channel in the input feature map. Finally, the Neck layer replaces the CBS convolution with the CBAM attention mechanism, which reduces experimental misdetection omissions. The experiments show that the mAP50 value of the RCC-YOLOv7 algorithm reaches 77.1% on the homemade sampling dataset, which is an improvement of 3.2% compared with the traditional YOLOv7 algorithm, and achieves a better detection accuracy and effect.