Purpose The detection of pylorus, a natural gastrointestinal (GI) anatomical structure, is one of the fundamental techniques that enables a high-autonomy digestive tract robot to move from the stomach to the duodenum. The pyloric center is the optimal position for passing pylorus from the soft-tissue protection standpoint. Thus, detection of the pylorus center should be investigated further in view of its indispensability for a high-autonomy GI robot. However, to the best of our knowledge, no result of pylorus center detection has been published thus far. Methods In this paper, we have developed a pylorus center detection method using CNN classification, Sobel and Laplace operators. The proposed algorithm's effectiveness is demonstrated by the precise center detection of six types of healthy pylori and six settings of diseased pylori. Results The average detection accuracy of the pylorus center is 22.33 pixels, which corresponds to a relative error of 2.33% when compared to 960 pixels, which corresponds to the diagonal length of an endoscopic image. A single image takes an average of 26.51 ms to process. Conclusion The clinical feasibility of the algorithm for real-time pylorus center tracking is established. The developed algorithm enables GI robots to autonomously locate and pass the pylorus.