Real-Time Burned Arm Localization Using Stereo Vision for Clinical Decision Support

Burn assessment constitutes a pivotal component of clinical decision-making, particularly within the domain of global public health.In this study, we propose a real-time burn localization method that utilizes binocular vision and the semi-global block matching (SGBM) algorithm. This method employs three dimensional reconstruction techniques to achieve millimeter-level accuracy in the localization of the burned area. The system utilizes a binocular camera to capture high-quality skin images, employs the Zhang Zhengyou calibration method to calibrate the camera parameters, and utilizes the SGBM algorithm to fuse global and local information for stereo matching. This approach effectively reduces parallax map noise and generates highly accurate depth maps.Subsequently, the burned area was calibrated in the world coordinate system using three dimensional reconstruction technology to generate visualized medical images.These images support the tracking of the treatment process and optimize decision-making. The experimental findings demonstrate that the depth error of the system can be maintained within 2 mm (450-500 mm error less than 1 mm) within the working distance of 450-800 mm, and the processing time of a single frame is 0.42 seconds, which fulfills the real-time requirement. The integration of binocular vision with the SGBM algorithm provides a solution for burn localization that is both accurate and efficient, thus demonstrating its great potential for application in the clinical decision-making process.This advancement is expected to improve the quality of rehabilitation management for burn patients.