Rebar lashing process of (precast concrete, PC) members is mostly manual operation, which increases production cost and reduces production efficiency. To solve the above problems, image-driven recognition of crossing points in rebar lashing for PC component was proposed. It consists of shadow processing based on adaptive bilateral filtering and HLS- normalized saturation value difference index, and corner detection based on second-order generalized Gaussian directional derivative filter and probabilistic Hough transform and neighborhood point search group average method. The algorithm can accurately identify the location of the intersection for the steel bars. The experimental results show that this algorithm has high recognition accuracy, fast recognition speed, good robustness, accuracy of more than 85%, and algorithm running time no more than 0.15s, which has practical application value.