Registration Algorithm For New-type Surface Mount Components

Surface mounting is an accurate manufacturing technology which assembles electroniccomponents onto Printed Circuit Board (PCB). In this paper, it mainly analyzes the detectingalgorithm of new surface-mount components’ angle and position offset.The proposed algorithm is based on the principle of registration, including the mutualinformation algorithm based on global gray, the Energy Minimization SURF and FREAKalgorithms based on local features. The details are as follows:1. As some component images in the precision electronics manufacturing have richinformation of gray level, the registration algorithm based on mutual information (MI) isintroduced in this paper, to obtain the component’s angle by seeking for maxim mutualinformation value. The experimental results showed that the MI algorithm can meet thepractical requirements in the accuracy but not in the velocity due to its complications.2. As the components in the precision electronics manufacturing is various and differentin characteristics, this paper employed the SURF algorithm based on feature point registration.The algorithm registered images based on fast feature extraction which is suitable for avariety of components. Due to SURF’s accuracy is not satisfied, the Minimum-EnergyMethod is used for accurate locating. The experimental results showed that the EnergyMinimization SURF algorithm can basically meet the practical requirements of componentpositioning in precision electronics manufacturing.3. To improve the speed of registration, a fast binary descriptor called FREAK isproposed in this paper. FREAK is excellent in features generating and matching, but haslower precision than SURF. So RANSAC is used to enhance the matching precision with theadvantages of calculation speed. Experiment demonstrated that the detecting algorithm basedon FREAK can meet the demand of practical requirements, and is more stable and quickerthan SURF.