Research On Detection Algorithm Of Chip Dispensing Based On Machine Vision

In the manufacturing process of semiconductor and electronic products,dispensing operation is a key technology in the process of chip production and packaging,which has been widely used in the automatic production process of detection and assembly.The quality of dispensing directly affects the production of the chip.The consequences of too little glue may cause insufficient interconnection between chip and packaging materials,thus leading to difference or failure on device performance.While too much glue would waste glue,as well as the longer time of curing deformation caused by glue would also influence chip quality fluctuation.Therefore,taking the chips as the research object,this paper proposes a machine vision-based chip dispensing detection system.The prime tasks were summarized as follows:Firstly,the hardware of experimental platform for chip dispensing detection has been built,and the imaging system of chip has been designed.According to the requirements of imaging,the industrial camera,lens,and light source has been selected to complete the image acquisition of the dispensing chip.Secondly,according to the structural features of the chip,the specific process of the algorithm for extracting the region of the chip which is to be detected has been introduced.This paper preprocess the image by using the median filter and an image enhancement technology based on gray values and the double threshold segmentation method.For the image which has been processed,extract the sub-pixel contour of the region which is to be detected by using a sub-pixel edge detection method based on arc tangent function fitting.Then,initially extract the region which is to be detected through region filling of the contour.Finally,the region which is to be detected would be extracted completely by using morphological processing on the extracted region.Thirdly,a fast pyramid-based NCC(Normalized Cross Correlation)matching method has been proposed to locate the dispensing chip,and the region which is to be detected of the dispensing chip would be extracted by an affine transformation.According to the features of the glue in the detected region,the shape-area featurewas selected as the feature for the detection and classification of the chip dispensing situation.The glue region in the detection region would be extracted by threshold segmentation.Then the union and connected components of the region would be calculated,and get the complete glue region by using the morphological processing methods such as closing operation,hole filling,and opening operation.Classifying the type of chip dispensing according to the area of region which has been calculated.Finally,the software platform of the entire detection system has been built,which can realize the functions of image acquisition and display,detection and classification of chip dispensing and the display of experimental results.The detection accuracy of the entire detection system was 98.2%,and the average detection time was 427.6ms.It proves the feasibility of the detection algorithm in this paper.