A Method for Nuclear Power Pipeline Weld Seam Recognition Based on Structured Light Multidimensional Information Fusion

This paper presents an automated method for precise pipeline weld identification and measurement using autonomous robots in nuclear power environments. It addresses challenges related to initial shooting exposure, camera angle, and distance from the environment by employing adaptive data collection techniques. These methods adjust exposure based on statistical brightness information from color images captured by a structured light camera. By detecting the pipeline model in threedimensional point cloud data, optimal camera positions and postures are determined for high-quality data acquisition. Geometric features are extracted from the point cloud data to create a geometric feature response map projected onto a twodimensional image, which is fused with color features to generate a grayscale fusion map. Image recognition algorithms identify weld areas, enabling precise three-dimensional measurements. Experimental results confirm the method’s effectiveness in improving data quality and achieving accurate weld seam identification and measurement.