Rail structured light strip image segmentation method based on density ratio scaling clustering

During high-speed driving of vehicles outdoors, complex image noise can be introduced due to the impact of environmental foreign object splashes and strong reflected light, which can alter the shape and distribution of the rail light band and affect the measurement accuracy of the rail profile. This paper proposes a rail light band segmentation method based on density ratio scaling clustering. Firstly, density estimation is performed based on the rail ROI histogram; then, the data within the rail ROI is re-scaled according to the cumulative distribution function of each point's neighborhood; finally, rail light band segmentation is achieved using a single threshold based on density clustering method. This method helps to improve the accuracy and robustness of rail profile extraction and is suitable for complex track structure scenarios in high-dynamic detection processes.