Research on intrusion detection and intelligent early warning model of crane boom dangerous area based on deep learning

In this paper, an intrusion detection and intelligent early warning model of crane boom in dangerous area based on deep learning is proposed to solve the shortcomings of traditional monitoring methods in efficiency, cost, flexibility and generalization ability. The model adopts three-level structure design, including multi-modal input layer, feature fusion and target detection layer and adaptive early warning decision-making layer. Multi-modal input layer integrates video stream and sensor data to construct dynamic boundary model. In the feature fusion and target detection layer, based on the improved YOLOv7 architecture, a dual-path attention module (DPAM) is introduced to improve the accuracy of small target detection. The adaptive early warning decision-making layer evaluates the potential danger through four-dimensional feature space, and adopts dynamic threshold strategy to reduce the false alarm rate. The experimental results show that the model is superior to the traditional method and other benchmark models in detection accuracy ([email protected] reaches 98.7%), real-time performance (FPS is 62 frames /s), false alarm rate (0.13 times/hour) and response time (210ms), which shows good adaptability to dynamic scenes and small target detection ability, and provides efficient and effective safety monitoring for dangerous areas of crane jib.