Dangerous Vehicle Identification Based on RCM-YOLOv8

To address the issues of insufficient feature extraction, low accuracy, and poor real-time performance in existing object detection algorithms for dangerous vehicle detection, this paper proposes an improved RCM-YOLOv8 algorithm. Based on the YOLOv8 framework, this algorithm introduces the C2f_Restromer module to optimize the image restoration model, integrates the CMRF module to enhance feature representation through multi-channel redundant information, and employs the MDFM module to resolve the problem of overlooking discrepancies during feature fusion and to strengthen multi-scale features. Experimental results show that the RCM-YOLOv8 model outperforms the original YOLOv8 and other algorithms in terms of recall rate, precision, and detection performance, effectively enhancing the detection capability for dangerous vehicles.