Research On Surface Mount Chip Detection And Location Base On Machine Vision

In recent years,development of surface mount technology(SMT)has been continuously enhanced and remarkable results have been achieved.But there is still room for improvement at level of vision inspection technology.Limited by field of view of camera,the large size of chip can be captured by a camera with large field of view.But complete reliance on increasing field of view of camera has limitations,and the recognition speed of algorithm is also limited.In addition,for high-density integrated chip,the recognition through rate is low and influenced by external factors such as light and chip quality.And classical template matching algorithm used for positioning is time consuming,leading to a reduction in the speed of production placement.For this reason,this paper designs a parameter extraction algorithm and testing algorithm for FLIP CHIP,which is important to solve low recognition rate and improve production efficiency.For problem of breaking limitation of camera field of view,an algorithm of moving stitching photo(MFOV)is proposed.The head is controlled to move over fixed camera by setting movement step and movement type in the interactive interface and transmitting it to host computer.Then,the image is sent back to image processing system for cropping and stitching to obtain an image that exceeds field of view of camera and contains all information of chip.Cropping threshold is obtained by moving the step conversion to ensure that different thresholds are used for different sizes of chips.For FLIP CHIP parameter extraction problem,we propose a region screening algorithm,including area screening and roundness screening.In order to reduce interference and extract image information of the most valuable part.The grayscale center of gravity set of weld balls is obtained by using the connected-domain filtering method,and the circle fitting function in OpenCV is used to calculate weld ball radius.Pre-processed images are fitted with a straight line or a rectangular fit to calculate the chip body size,the rotation angle and the chip center point coordinates.An intercept method is proposed to calculate number of rows and columns of welded balls.The principle is that intercept of same row of weld balls is similar,intercept of different rows of weld balls differs greatly,and column intercept is based by the same calculation method.A remapping search algorithm is designed to calculate number of empty bits in chip.For FLIP CHIP test positioning problem,two methods are proposed: one is shape-based template matching combined with Fourier transform;the other is to use results of extracted parameters to directly convert rotation angle and center point coordinates of chip in image coordinate system into physical coordinate system.Then,the offset is output after making a difference with size of acquired ROI image.It is compared with traditional template matching algorithm,and it is superior in time with guaranteed detection accuracy.