With the rapid development of high-speed rail technology, foreign object intrusion on railway tracks has become a critical factor affecting train operational safety. There is an urgent need for rapid identification of foreign objects on tracks to ensure safe train operation. Traditional foreign object detection methods suffer from issues such as insufficient detection accuracy and low efficiency. Based on a comprehensive analysis of domestic and international research on railway foreign object intrusion, this paper proposes a novel foreign object detection method for railway tracks. It completes key technical tasks such as image preprocessing, track region extraction, and detection model design, thereby improving the speed and accuracy of foreign object detection on tracks. The main contributions of this paper are as follows: To address issues such as detail loss and feature blurring in images or videos caused by noise, fog occlusion, and other factors, this paper conducts image preprocessing work, including denoising with a bilateral filtering algorithm that integrates multi-scale σ values, defogging with an improved ω-value dark channel prior algorithm, and image enhancement with a histogram equalization algorithm. Experimental comparisons with different algorithms demonstrate that the improved algorithms effectively enhance image quality. Regarding the problem of track region extraction, a multi-algorithm fusion-based track region extraction algorithm is proposed. This algorithm integrates grayscale transformation, Canny edge detection, optimized Hough transform line detection, and ROI (Region of Interest) track region selection algorithms, achieving precise segmentation and extraction of track regions. Its segmentation and extraction effectiveness were validated through experiments. To address