SMT patch defect detection algorithm based on prior information of image

Abstract

Patch testing is an important part in circuit board testing after SMT melted welding. Domestic patch recognition methods exist many shortcomings, such as the cumbersome algorithm thought, large calculation, the high requirement of hardware, poor robustness. In this paper, aiming at detecting the defects of rosin joint, lack of tin, excess solder, surface defect, wrong parts, tombstone, offset, a thought that using priori information for image positioning was used. In combination with target area color analysis, area connected domain algorithm thought, rapid detection of the patch was achieved in three steps: patch image positioning and size calculation, patch regional color analysis, patch type identification. The experiment has proved that this method had many advantages: convenience of feature extraction, the low interference of the shooting condition, well robustness, the convenience of threshold setting, high defect recognition rate, robust algorithm. In MATLAB platform, the correct recognized rate of of the patches in this study was 94%∼95%, false alarmrate was 4%∼5%, false negative alarm was 1%, the averaging time consuming was 41ms.