An Improved BM3D-Canny-Zernike Algorithm for Micro-Size Detection of Electronic Connectors

Abstract

To detect the micro-size injection molded parts of electronic connectors, this paper establishes a complete size detection system based on machine vision, and measures the size through image acquisition and processing, according to the features of the injection molded parts. The proposed system is called the improved BM3D-Canny-Zernike algorithm. Specifically, the traditional block matching and three-dimensional filtering (BM3D) image denoising algorithm was improved to optimize the peak signal-to-noise ratio (PSNR) and reduce the mean squared error (MSE). Then, the Canny algorithm was improved for pixel-level edge detection, and the Zernike moment is improved for detecting edges on the subpixel-level more effectively and reducing the calculation amount. Finally, the least squares method was employed to fit the edge to be measured. The exact pixel length was obtained by solving the function of different edges, thereby realizing size measurement. Experimental results show that the mean error percentage of our algorithm was 8.73%, which meets the needs of industrial detection.