Part Flatness based on Shi-Tomasi Corner Detection and CNN Multimodal Fusion

The flatness detection of parts in industrial inspection is very important for quality control, and the traditional single modal method has information limitations. In this study, a detection method based on Shi-Tomasi corner detection and CNN multi-modal fusion is proposed: Shi-Tomasi is used to extract image edge features (18.7 % redundant corner points are reduced compared with Harris), laser ranging is used to obtain height data, and features are extracted by CNN respectively and then spliced and fused. Experiments show that the detection accuracy of the multimodal model is 94.8 %, which is 7.2 % higher than that of the single mode. The mean square error is 0.032 mm2, the average absolute error is 0.11 mm, and the error fluctuation is less than 5 % under illumination change and noise interference. This method provides an effective solution for industrial high-precision detection, and can optimize data processing to adapt to complex scenarios in the future.