Objective: Crane lifting is a critical process in steel and concrete structure construction, but safety accidents occur frequently, necessitating effective preventive measures. Traditional supervision relies on human experience and on-site judgment, which are vulnerable to operator fatigue and distraction. As information technology advances, it plays a vital role in crane safety risk monitoring. However, current research mainly focuses on monitoring lifted objects, with insufficient comprehensive studies on the cranes' overall workspaces. In terms of worker monitoring, existing wireless sensing methods perform poorly in steel structure construction scenarios due to significant interference, while visual methods mostly focus on tracking workers' locations in the image, lacking their accurate real-world coordinates. This study aims to propose a spatial collision monitoring method that integrates building information modeling, crane sensing, and computer vision technologies to enhance safety monitoring of crane-worker interactions in steel structure construction scenarios. Methods: The specific research framework and methods are illustrated as follows: First, a crane's workspace is categorized into low-risk, medium-risk, and high-risk areas, and equations for defining medium- and high-risk boundaries are established. Then, the crane's location and posture are monitored in real time using sensors. Simultaneously, using the YOLO11-OBB model and perspective transformation, the actual location of the workers on the floor is calculated based on precalibrated reference points and their location in the image. Finally, the on-site monitoring data are integrated into a 3-D management platform, which calculates and visualizes the spatial collision risks between the crane and the workers in real time. Results: A case study from a steel structure construction project in Xi'an was used to test the accuracy and feasibility of the proposed method. The test results showed that the monitoring error for the medium-risk and high-risk crane workspaces was ±2.600 and ±2.611 m, respectively. The accuracy of the YOLO11-OBB model for worker localization and recognition was 0.968, with a recall rate of 0.969, the mean average precision at intersection over union (IoU) 50% (mPA50) of 0.980, and the mean average precision at IoU 50%-95% (mPA50-95) of 0.864. The mean absolute error, mean relative error, and root mean square error for the calculation of the actual locations of the workers were 67.44 mm, 4.32%, and 86.16 mm, respectively. During the 9-month monitoring of the site, the frequency of workers entering high-risk areas showed a fluctuating decline, demonstrating the feasibility of the method in enhancing safety warnings on construction sites. Conclusions: This study presents a spatial collision monitoring method for crane and worker interaction in steel structure construction scenarios, including the definition and monitoring of crane hazardous workspaces and worker locations. A 3-D visualization platform is used to monitor the collision situation between cranes and workers. The case study indicates that using only regular cameras and without the need for workers to wear additional monitoring devices, the method can effectively ensure the accuracy of worker localization within a distance of 20 m from cameras. The method also enables differentiated responses to cranes operation in spaces with varying risk levels, allowing managers to intuitively view the spatial collision status of the crane and workers through the 3-D model on-site.