Detection of Fault Features in Remanufacturing of Automotive Components Using Image Processing and Computer Vision Techniques

Remanufacturing of automotive components involves restoring cores from end-of-life vehicles (ELV) to a condition where the component can be used again. Implementing image processing technique in the visual inspection stage is used to reduce cost and time while increasing accuracy. This research aims to perform real-time surface visual defect detection that is simpler and less costly which commonly uses machine learning and/or deep learning techniques. The methods used in this research are a combination of image processing and computer vision techniques to great effect. The techniques used are a pixel-based contour detection and image subtraction. For this research, connecting will be the sample component to test the algorithm. This approach will take less time to train since it compares the non-defective connecting rod image with the defective connecting rod. The defects that are tested in this research are cracks and buckles, which commonly occur in connecting rod failure due to high load and stress. There are different settings of threshold and illumination that were tested, such as different thresholds of 50, 60, and 70, also illumination colour temperatures such as white, natural, and warm. After 270 trials with different settings and defects, the proposed algorithm achieved a 93% accuracy using a 70 per cent threshold and white light (3000–3500 K). The findings suggest that the implementation of a pixel-based contour detection through image subtraction holds promise as a cost-effective alternative to utilizing machine learning and/or deep learning techniques.