To address the limitations of existing YOLOv7 models in handwashing action detection, including low detection accuracy, weak environmental interference resistance, and insufficient discrimination of similar actions, this paper proposes an enhanced CCL-YOLO object detection algorithm based on improved YOLOv7. The proposed algorithm introduces three key innovations: (1) An Enhanced Axial Local Attention mechanism is incorporated to strengthen the model′s capability in capturing long-range contextual dependencies; (2) The CARAFE operator replaces conventional nearest-neighbor interpolation for upsampling, enabling more effective content-aware feature reorganization without increasing model parameters; (3) Structural optimization from SPPCSPC to SPPFCSPC improves detection accuracy by 2.9% and frame rate by 10 while maintaining equivalent receptive fields. Additionally, a lightweight adaptive decoupled detection head is designed to replace traditional coupled detection heads, achieving a 7.6% recall improvement and 2% mAP@0.5 enhancement at the cost of only 2% precision reduction. Experimental results on a custom dataset demonstrate that the improved algorithm achieves 81.2% mAP@0.5, representing a 7.2% accuracy improvement over baseline YOLOv7, with precision and recall rates increased by 2.9% and 11% respectively. The proposed method effectively meets practical requirements for real-world handwashing action detection while maintaining computational efficiency.