Research On The Image Matching Algorithm Of LED Chip Check And Sorting

Automatic alignment technique on the basis of machine vision is one of the most important techniques in the check and sorting of LED chip. By getting the transformation parameters between the template image of the chip and the image of the chip on the wafer, the location of the electric poles of the LED chip can be known,then the electronic parameters of the LED chip can be tested,the parameters can be used in the LED chip sorting. This paper bases on the alignment of the LED chip, and does an intensive research on the image filter, feature extraction and image matching .The main contents are:This paper first did a research of the existing image matching methods, including NGC (Normalized Grayscale Correlation) Method, SSDA (Sequential Similarity Detection Algori-thm), FFT (Frequency Fourier Transform)Method and Feature Based Methods,after analyzed the advantage and disadvantage of these algorithms, the paper chose a Point Pattern Matching Methods as the matching method,which uses corner points as the feature .Analyzed the disadvantage of the Median Filter, presented a Self-adaptive Median Filter algorithm. Then the two algorithms were implemented in VC .With the programs, a case study on the noise images have been done, the results show that the Self-adaptive Median Filter Method protects the corners better than the Median Filter.The algorithms of corner point extraction had been analyzed, including Harris and SUSAN. The two algorithms have been improved and implemented in VC.The two programs both extract the corner points from the images of LED chip's assignment mark accurately.At last, the paper chose SUSAN because it has lower computation complexity and stronger noise resistance.A quick PPM (Point Pattern Method) algorithm had been reaserched, which integrates Local matching and Global matching and increases the matching speed dramatically.Then the paper implemented the algorithm in VC and did a case study on the LED chip's location mark with the program,the result shows the algorithm has good location accuracy.Software of the image matching of the LED chip had been developed. At last, an example of the LED chip's location had been gived, which showed the process of the PPM.