Presenting a Method for Gluing Error Detection Using Image Processing

The application of artificial intelligence and image processing for automatic fault detection in industrial settings holds substantial significance. However, image processing systems encounter challenges in achieving accurate automatic detection, including variable lighting conditions, low image quality, and the complexity of observed objects. This study specifically addresses the detection of gluing errors in automotive parts. An improved template matching algorithm is employed to identify boundary errors. The proposed method incorporates SIFT (Scale-Invariant Feature Transform) for feature extraction, allowing for the identification of key local features in images regardless of changes in scale, rotation, or lighting conditions. To address variations in scale and alignment, affine transformations are employed to standardize the images, ensuring consistent dimensions and orientations. After this preprocessing step, brute-force matching techniques are utilized to accurately match the extracted features from the modified images with a target reference, refining the output and enhancing the detection accuracy of improperly glued areas. Results from executing the algorithm on input images demonstrate the high performance of the proposed method in gluing error detection.