Curve extraction plays a pivotal role in various fields such as medical imaging, road mapping etc., involving the identification of curves from images. Our research focuses on Railway Engineering Scale Plan (ESP) images, aiming to automatically categorise railway lines as tracks and points within these images. Employing a curve extraction algorithm, we initiate the process by selecting a single curve and tracing it throughout the image to capture all edges effectively. The algorithm is designed to classify curves that intersect each other, disappear, and reappear, with the goal of producing results that closely resembling human identification of railway lines. Railway lines normally meet at points but crossing may also happen. Initially, identifying railway tracks as curves or edges in images is achieved using computer vision techniques. Preprocessing techniques like smoothing, edge detection, and thresholding enhance image features and reduce noise. Object detection algorithms are then utilized to identify railway tracks within the pre-processed images. Subsequently, the curve-tracing algorithm is applied to follow the paths of the identified railway tracks throughout the image. This algorithm accounts for the curvature of the tracks and the presence of other curves in the image. Once the curves have been identified, the curve-tracing algorithm can be applied to follow their path throughout the image using Steger’s Curve tracing algorithm. Unlike existing methods, we continue tracing the existing curve while handling intersecting curves as separate tracks, as railway lines intersection should be traced differently than standard intersecting curves extraction. We also incorporate the concept of looking a distance ahead, to continue tracing curves for disappearing and reappearing curves in images. Moreover, railway lines have specific properties such as being long, smooth, and continuous with minimal abrupt changes in direction. To address this, we have done a filtration based the characteristics of the railway tracks. We developed a method for filtering curves based on the analysed railway tracks and points properties, ensuring that only curves supporting railway characteristics are retained. While working with images of railway plan we may have very large image plans to work with which usually have issues with memory storage and complex computations. We address this by processing images tile by tile, effectively capturing small track details with minute features. Additionally, we bridge gaps in rail lines by identifying similar characteristic curves in image and joining them to form single tracks, which are then classified and labelled. Furthermore, we imple- mented a piece-wise spline approximation of lines and points to obtain smoother curves. These curves can be incorporated into Signal Interlocking Plans (SIP) automatically for signalling and interlocking purposes, contributing to railway infrastructure management and automation