An Efficient Contour Detection Approach for Extracting Rim from Wheel Images

In the automobile industry, the wheel alignment process plays a major role in ensuring that the wheels are placed at proper angles, and pointing straight for proper road contact and safe drive. Identifying the location of the rim in the wheel plays a major role in aligning wheels. There are many state of art image processing techniques available for rim detection from wheel images which includes circle and ellipse detection algorithms. When the captured wheel images are skewed (angulary/translationally), then the existing techniques fail to detect the rim. Hence, we propose a contour-based rim detection method which can detect the position of the rim from the images of automobile wheels even when they are angularly or translationally skewed. The rim detection is done as a two step process which encompasses detection of contours on the image and filtering the irrelevant contours. If the rim is detected with an anomalous contour, our proposed methodology will correct it using K-means clustering algorithm by considering the distance from center of the contour. The performance of the proposed methodology has been analyzed using 50 images having various viewpoint variations, and it is observed that it is extracting the rim perfectly.