Currently, the operational risk assessment of quay crane operators at ports relies on manual evaluations based on experience, but this method lacks objectivity and fairness. As port throughput continues to grow, the port accident rate has also increased, making it crucial to scientifically evaluate the risk behaviors of operators and improve their safety awareness. This paper proposes an automated evaluation method based on a Deep Q-Network (DQN) to assess the risk behaviors of quay crane operators in virtual scenarios. A risk simulation module has been added to the existing automated quay crane remote operation simulation system to simulate potential risks during operations. Based on the collected data, a DQN-based benchmark model reflecting the operational behaviors and decision-making processes of skilled operators has been developed. This model enables a quantitative evaluation of operators’ behaviors, ensuring the objectivity and accuracy of the assessment process. The experimental results show that, compared with traditional manual scoring methods, the proposed method is more stable and objective, effectively reducing subjective biases and providing a reliable alternative to conventional manual evaluations. Additionally, this method enhances operators’ safety awareness and their ability to handle risks, helping them identify and avoid risks during actual operations, thereby ensuring both operational safety and efficiency.

To address the high-precision measurement requirements for container pose on dual-trolley quayside crane-transfer platforms, this paper proposes a machine vision-based measurement method that resolves the challenges of multi-scale lockhole detection and precision demands caused by complex illumination and perspective deformation in port operational environments. A hardware system comprising fixed cameras and edge computing modules is established, integrated with an adaptive image-enhancement preprocessing algorithm to enhance feature robustness under complex illumination conditions. A multi-scale adaptive frequency object-detection framework is developed based on YOLO11, achieving improved detection accuracy for multi-scale lockhole keypoints in perspective-distortion scenarios (mAP@0.5 reaches 95.1%, 4.7% higher than baseline models) through dynamic balancing of high–low-frequency features and adaptive convolution kernel adjustments. An enhanced EPnP optimization algorithm incorporating lockhole coplanar constraints is proposed, establishing a 2D–3D coordinate transformation model that reduces pose-estimation errors to millimeter level (planar MAE-P = 0.024 m) and sub-angular level (MAE-$\theta$ = 0.11°). Experimental results demonstrate that the proposed method outperforms existing solutions in container pose-deviation-detection accuracy, efficiency, and stability, proving to be a feasible measurement approach.

As the core sensing method for port automation equipment, LiDAR is widely used in spatial distance measurement, three-dimensional positioning, and environmental perception. However, its ranging mechanism, parameter configuration, and role in complex port operations are relatively abstract, making traditional training difficult to achieve intuitive understanding and efficient application. To address this issue, this paper proposes a virtual LiDAR simulation training platform based on digital twin technology. The platform constructs a virtual environment highly consistent with real port operation processes, designs a virtual LiDAR module with adjustable horizontal/vertical field of view, resolution, channel count, and ranging distance, and achieves bidirectional synchronization of data and control signals between physical and twin spaces. Using a bulk cargo handling machine as the verification object, consistency comparison experiments were conducted on point cloud density, ranging accuracy, and geometric similarity. The results show that the virtual point clouds generated by the platform are highly consistent with real point clouds in key indicators, featuring low cost, high flexibility, and scalability, and can provide reliable support for operation training, algorithm testing, equipment deployment assessment, and operational safety analysis of port automation equipment.

With the continuous improvement of port automation, remote control of container gantry cranes has become mainstream. However, real machine training poses significant safety risks, where novice drivers (Trainee) are prone to equipment collisions and violent container swings during operations. To address this issue, this paper designs and implements a digital twin system integrated with an operation risk assessment module, used for safety reproduction and risk quantification of gantry crane operations. The system adopts a time series prediction method based on attention mechanism, establishing a safety benchmark by modeling skilled drivers' (Trainer) operational data, and predicts and compares novice drivers' operational behaviors to quantify risk deviations. Experimental results demonstrate that the system can effectively distinguish operation risks across different driver experience levels, showing excellent application value in avoiding real machine training hazards, improving training efficiency, and achieving personalized risk assessment.