Research On The Detection Algorithm Of Conductive Particles In ACF

With the rapid development of the electronics industry,the demand for LCD screens has also increased.Most liquid crystal panels use a chip on glass bonding technique,and the material that connects the chip to the glass is an anisotropic conductive film(ACF).The number of conductive particles in ACF is one of the important basis for judging the quality of LCD screen.The method of manual sampling requires the use of high-precision instruments to assist workers in testing,and only some products can be taken out of each batch for testing.Therefore,the manual sampling speed is slow,the detection is incomplete,and the accuracy is not high enough to meet the client demand.The use of automated optical inspection technology is the research direction to solve the above problems.The conductive particle detection system based on automatic optical detection needs to automatically segment the region of the conductive particles in the image,record the coordinates of the segmented conductive particle region in the standard file,and the subsequent conductive particle detection algorithm directly reads the conductive particle region in the standard file.Finally,the number of conductive particles is detected by a conductive particle detection algorithm in the conductive particle region.The ACF conductive particle detection system studied in this paper has been applied in the company and has been well received by the company.The main research work of this paper can be summarized the following three parts.1.Research on image pre-positioning algorithm.This paper proposes a method based on image similarity to judge whether the image is similar.Compared with the perceptual hash algorithm,the algorithm has better intuition,better judgment effect and higher running efficiency,and can be better applied to the image pre-positioning.2.Improved conductive particle segmentation algorithm.Aiming at the problem that the existing conductive particle region segmentation algorithm has complex distribution of conductive particles,poor generality of segmentation algorithm and low accuracy of segmentation results,this paper improves a pre-processed conductive particle segmentation algorithm.The algorithm first preprocesses the image,and obtains the pin area withconductive particles through binarization and region difference.The conductive particles are then enhanced to redraw the gray value of the image.Finally,the coordinates of the conductive particle region are accurately segmented by the horizontal and vertical grayscale projection of the image.The experimental results show that the proposed conductive particle segmentation algorithm can segment the coordinates of the conductive particle region more accurately and has better versatility.3.Improved conductive particle detection algorithm.Aiming at the influence of the existing conductive particle algorithm on the illumination direction and the inaccurate counting of the adhered conductive particles,an improved algorithm based on local mean difference is proposed.The algorithm increases the left and right local regions,and obtains the local mean difference in the horizontal direction,supplementing with the local mean difference in the original vertical direction to solve the influence of the illumination direction.In addition,for the large area,the adhered conductive particles are separated by the method of watershed combined with morphological operation,and the adhered conductive particles are accurately counted.The experimental results show that the proposed conductive particle detection algorithm improves the accuracy of detection and reduces the influence of illumination direction.