Abstract: Low alloy steel welding cracks in steel structure bridges are often hidden inside or below the surface of the weld seam, and their complex and diverse shapes make it difficult to comprehensively and accurately identify and locate these hidden crack defects, thus making it impossible to achieve effective quantitative detection of welding crack defects. Therefore, a quantitative detection study on welding crack defects in low-alloy steel for steel structure bridges is proposed. Based on laser ultrasound technology combined with B-scan imaging technology, low-alloy steel welding crack images of steel structure bridges are obtained. Then, a multimodal feature fusion algorithm is used to extract the ROI region of crack defect images, in order to reduce false detections and hidden crack defect omissions; Based on the extracted crack defect target area, effective extraction of crack edges is carried out. According to the extracted crack edge curve, the depth, width, and other dimensions of the crack are obtained to achieve quantitative detection of crack defects. The results show that using this method for quantitative detection of welding crack defects in low-alloy steel has high detection accuracy and good results.