Colonic diseases pose a significant global health challenge, emphasizing the need for efficient colonoscopy and analytical tools. Computer vision algorithms and techniques revealed promising solutions and results in the literature. However, several limitations, including low accuracy, prevent it from being implemented or integrated into current systems. This paper presents a holistic approach for automatic anatomical landmark localization in colonoscopy videos employing feature-based monocular visual odometry. Our system preprocesses the video to eliminate irrelevant frames and further tracks the camera’s motion through optical flow, estimating the colon’s trajectory in a robust manner. Dynamic Time Warping aligns the obtained trajectory with a reference trajectory that is obtained from the medical literature, which will allow us to pinpoint eight major anatomical landmarks accurately. Benchmarking results on the datasets demonstrate superior performance compared to several image classification methods and reach an F1-score of 0.8464. This robust method is promising for enhancing colonoscopy navigation, automating reporting, and improving patient care in gastrointestinal diagnostics.