Quantifying Optical Edge Detection Bias in Automated Seam Opening Tensile Analysis
Optical edge detection overestimates seam opening force by 8% to 23% due to yarn fuzzing, requiring physical feeler gauge verification for accurate batch compliance.
Canny Edge Detection
Detection Algorithm
Multi-stage computational procedures identify sharp changes in image intensity to locate the boundaries of objects. Modern textile quality systems use canny edge detection to isolate yarn defects or weave irregularities within a digital frame. This method identifies local maxima of the image gradient through a rigorous mathematical filter.
Noise Suppression
Gaussian filtering removes high-frequency interference that would otherwise produce false positives during the inspection of textured fabrics. Because canny edge detection relies on derivatives, a pre-processing step smooths the pixel values to prevent minute fibre fly or surface fuzz from being recorded as structural breaks.
Gradient Magnitude
Directional operators calculate the intensity slope across the horizontal and vertical axes to determine the strength of a potential edge. When canny edge detection processes a fabric scan, it produces a map where the intensity of each point corresponds to the sharpness of the transition. These values define the primary shape of a knitting fault or a seam line.
Hysteresis Thresholding
Dual limit values decide which edge candidates to keep and which to discard based on their connectivity to strong features. A weak pixel survives in canny edge detection only if it touches a pixel above the high threshold. This logic ensures that broken yarn paths appear as continuous lines rather than scattered points.
If the low threshold is set too high, the system loses subtle but important details such as fine pilling or micro-tears. Proper calibration allows the algorithm to ignore surface glare while retaining the structural geometry of the weave.