In response to the high difficulty of size detection after the assembly of backlight panels in the electronics industry, backlight panel size and contour detection software has been developed. The process begins with preprocessing operations such as tilt correction, feature area extraction, and character grayscale value modification. Then, by using an adaptive threshold brightness method based on grayscale density distribution and grayscale difference, the feature area is traversed with sub-images, merging defect sub-images with overlapping areas and filtering out regions without obvious defects. Subsequently, a support vector machine multi-class classification method is used to extract the pixel distribution patterns of binary images as training features to identify the types of backlight panel defects.