Three-dimensional Appearance Inspection Of QFP Chips Based On Machine Vision

IC (Integrated Circuit) chips have been widely used in national defence, aerospace,energy resource and medical treatment fields, and their production technology is consideredas one of the high-tech which affects the development of21st century. The three-dimensionalappearance inspection is an critical step in the manufacturing process of chips. Specially, it’svery significant to detect the three-dimensional appearance of QFP (Plastic Quad Flat Package)chips, because they are one of important IC chips. This thesis investigates thethree-dimensional appearance inspection based on machine vision according to the features ofthe QFP chips, such as small dimensions, multi-pin, and large output. The major researchwork is as follows:The thesis finishes the demand analysis for the three-dimensional appearance inspectionsystem of QFP chips, and determines the major Technical indicators including the ability todetect the chip’s size range, test items, test accuracy and speed, and discusses the currentmachine vision inspection system applications. Finally, it designs the system which includesvisual image acquisition, testing software development based on the mil package,multi-threaded design, database management, serial communications, etc.The thesis designs the hardware for image acquisition and detecting software systemswhich meets the requirements of the inspection of QFP chips, and carries out optical design,chooses visual hardware according to optical requirements, and designs the mechanicalstructure. Among them, the optical path uses a single camera with four planar speculas on theway for the three-dimensional detection problem, which not only reduces the difficulties ofcalibration and the software for calculating the chip three-dimensional targets, but alsoreduces the cost, and uses parallel optical path design to solve the imaging difficulties dued tochips’ large range in dimensions. Besides, in order to improve the quality of chips image, ituses backlight to increase the image contrast; and uses a large depth of field lens to solve theproblem of imaging blurring caused by the different length of optical path. Additionally, thesoftware system uses object-oriented design methods, multi-threading technology and VisualC++6.0as a development platform, and image processing algorithm is based on Mil softwarepackage for secondary development, which reduces the development cycle and the difficultyof software development. Finally, the thesis completes the assembly and debugging work of the three-dimensionalappearance visual inspection system of QFP chips successfully, and makes some experimentsincluding detection speed, stability, repeatability and accuracy in the laboratory environmentconsidering the impact of random error. Besides, it processes the inspection results consist ofstatistics, analysis and printing reports. The experimental results show that the inspectionaccuracy is15microns and the inspection speed is187milliseconds, and they have reachedthe industrial detecting requirements.