Detection of ball grid array solder joints based on adaptive template matching

Abstract

This paper aims to achieve the accurate detection of ball grid array (BGA) solder joints. To this end, the author presented an adaptive template matching method for BGA solder joints based on shape detection. First, the region of interest (ROI) was selected from the X-ray image of the printed circuit board (PCB). Then, an edge template was generated through ROI extraction and threshold segmentation, and the direction vector f the edge template was taken as the prior knowledge. After that, the global traversal search was performed on the image pyramid in the top-down manner, aiming to obtain the potential matching targets. Finally, the edges were adjusted by the least squares method to yield the optimal matching results. The proposed method was proved robust, rapid and accurate through an experiment. The research findings shed new light on the BGA solder joint detection and extraction in various conditions.

Optimized detection method of automotive floor solder joints based on digital-analog information

“…[6].LI Yong -Jing proposed a detection method based on shape template matching for automatic detection of shape defects of stamping parts on assembly line [7].J Ma designed a reflection component estimation method based on simulated annealing algorithm to identify surface solder joints according to the shape distribution [8]. Hao X proposedan adaptive template matching method for BGA solder joints based on shape detection, and the ed ges were adjusted by the least squares method to yield the optimal matching results [9].…”

“…G(x, y, σ) = 1 2πσ 2 e − (x 2 +y 2 ) 2σ 2 (9) L i = G i − Up(G i+1 )⨂κ 5×5 (10) Wherein, L i represents the Laplacian image of layer i, G i , G i+1 represents the Gaussian image of layer i and i+1, Up represents the up sampling function, ⨂ represents the convolution, κ 5×5 represents the 5 * 5 kernel.…”

Soldering defect detection in automatic optical inspection

“…Thus, for example, [4] uses the normalized cross-correlation as the measure of similarity, which is directly calculated based on pixel intensities of PCB images. [5] utilizes template matching in IC images and extracts the edge features as the input by taking the inconsistent shape and non-uniform color distribution into account. However, template matching approaches are very sensitive to small differences in the appearance of the target structure, which cannot be avoided when manufacturing solder joints as they are tiny and the manufacturing process is not entirely consistent.…”

Image Segmentation of Fiducial Marks with Complex Backgrounds Based on the mARU-Net

“…Crispin et al [13] proposed a template matching localization method for PCB component inspection that used gray-model fitting to generate common templates for a set of components, and considers normalized cross-correlation (NCC) as a measure of similarity. Hao et al [14] selected an edge template from the X-ray image of a PCB through object extraction and threshold segmentation, and used the direction vector of the edge template as prior knowledge. However, these kinds of approaches are not robust enough due to their high sensitivity to small differences in the texture of target structures.…”