To address the problems of hollow recognition results, high noise, and poor crack continuity in pavement crack identification using the Sobel algorithm, this paper proposed a joint algorithm combining the Sobel algorithm and matched filter for three-dimensional crack recognition of asphalt pavement. A matched filter in the form of a sine function was designed, and the matched filter was used to extract crack features from the results of the Sobel algorithm. The matched filter results were subjected to threshold filtering and connected domain analysis to effectively solve the hollow recognition problem, while also reducing noise and improving crack continuity. The test results show that the Sobel + matched filter joint algorithm can achieve 84.72% accuracy, 95.32% recall, and 89.44% F-value, and its performance is better than that of the Sobel algorithm and the Sobel + morphological closing joint algorithm.