Improved-canny-operator-based part image edge extraction method

The invention discloses an improved-canny-operator-based part image edge extraction method. A part color image is obtained and the obtained part color image is converted into a grey-scale map; frequency-domain filtering is carried out on the grey-scale map by using Fourier transform; with prewitt operator templates in four directions, convolution sharpening is carried out on the filtered grey-scale map to obtain a maximum gradient graph of the image; a grey level histogram of the maximum gradient graph is calculated, a threshold T is obtained by using a lowest point method, the grey-scale map is divided into two regions by using the threshold T, the two regions are processed by using an OTSU method to calculate a high threshold and a low threshold; and then binarization of the grey-scale map is carried out based on the high threshold and the low threshold, with a binary image generated by the high threshold as an edge base, edge base connection is carried out by using a binary image generated by the low thres