Electrical colored wires inspection algorithm for automatic connector producing machines

This paper proposes a method for inspecting the order of electrical colored wires in industrial connector cables manufacture. The key idea of the algorithm are threefold: First, the rectangle connector parts of sample image and test image of product are analyzed and detected by using shape and region properties. The region of connectors are then extracted and rectified to horizontal direction. Second, the pair images are aligned using features based method. The robust features are detected and the matching process is combined with outlier removal using Random Sample Consensus (RANSAC) algorithm. This step is really necessary when the products appear with arbitrary geometric shapes in the images. Third, the region of wires under connector parts of both image are determined. The difficult problem is the displacement and deformation of wires. To detect the positions of each wires exactly, a histogram of a horizontal scanline of the region under connector part is built. To find the exact position of wires, the peaks of horizontal histogram of the scanlines are detected. When the central of wires are detected, a square patch surround the central of each wire is extracted. The Color Structural Similarity Index Metric (CSSIM) is applied to compare the similarity of sample patches and test patches. The simulation results will demonstrate efficiency and effectiveness of the proposed method for color wires inspection using vision system.