Defect Detection Of Beer Bottles Based On DSP And Image Processing

Beer industry's rapid development in our country need for the improvements of the original production line urgently to enhance the efficiency of production,developing towards high-speed, high-precision and intelligent direction. At present the domestic study of beer bottle defect recognition system is still in the initial stage, and most manufacturers use artificial detection method,which is very susceptible to subjective factors such as the human eye distinguish ability and easy-fatigue and the quality and quantity of production tasks can not be completed ; Machine Vision system has the advantages such as high speed, accuracy, and non-contact, and applying it to beer bottle detection is an effective way to overcome the shortage of artificial inspection and increase the degree of automation and production efficiency of the beer production. Visual inspection system must improve the image processing speed to complete on-line measurement of beer bottles.The main research topics is beer bottle default defect technology based on image processing and DSP, in order to satisfy the beer bottle inspector's high demand of the speed and establish foundation of following development. The main work and innovation of this paper are as follows:(1) The beer bottle image processing method. Using Visual C + + development platform, common image processing methods like image filtering, histogram equalization,gray stretch,window transform,threshold segmentation,edge extraction, morphological operations are achieved and applied to the image process of beer bottles. The object of above work is to choose the image processing method meeting with the actual needs and prepare for the following defects detection and code transplantation on the DSP platform.(2) The image localization. The key of bottle finish and bottle base flaw inspection lies in the localization of the circle center. The external circle edge of the bottle finish ring is clear, for the localization of the bottle finish circle center the edge detection with line scan is used and then we use improved gravity method to search the external circle center and the radius. The internal edge is few, so we determine its radius by the proportion to external circle radius. Because the ring edge of the bottle base is not obvious, for the localization of the bottle base circle center we use a improved gradient Hough transform after the edge detection.(3) The defective bottle judgement method. For the bottle finish image, according to the actual image characteristics, select two eigenvalues, first is the external cicle edge points' saturation degree (that is, detected edge points' proportion to the points under ideal conditions), and the second is the number of edge points in sector and the points number adjacent of neighbor sectors after deviding the finish ring into 16 sectors using ring scanning algorithm. The bottle base is devided into and internal circle regiona,then anti-skid texture ring region is devided into 16 sectors ,using neighbor region pixel meanvalue comparison to judge defects,and the internal circle region is divided into inner square zone (divided into 16) and four arch regions outside the square, using a new algorithm based on symmetric region image block matching degree to judge defects. To the calibration region of bottle body, a mesh deviding method is used and statistic edge points' number in the grid to identify defects.(4) Familiar with DSP-based image processing system, implement the image processing algorithms on DSP, and in the experiment Ruitai company's ITECKDM642 development board is used to program and debug codes.In a word ,the research results show that the beer bottle defect inspecting method used in the article has not only theoretical meaning but also realistic value. Further research on the faster process method and hardware is worthy and required.