The Study On The Detection And Recognition Algorithm Of Safety Glass Fragmention Particles

China practice a compulsory detection regulations toward the safety glasson the automobiles, trains and the buildings since May 1,2003. Fragmentationstate test is one of the important contents in safety glass detection, whosepurpose is to evaluate the possibility of hurt and whose detection content is toevaluate the fragmentation particles' state by the information in the brokensafety glass particles. Up to now, the recognition of the results in this test isachieved by manpower. This paper will apply the digital image processingtechnology to the digital images to detect and recognize the fragmentationparticles to improve its detection efficiency and detection accuracy.If we study on the shape, the size and the count number of the fragmentationparticles, the first task is that we should draw its crack lines and depict thefragmentation particles. Based on the specialty of gray-level distribution in theinitial glass fragmentation gray images, according to the following steps separatethe crack lines from the inner regions of glass fragmentation particles.The first step is to detect the crack line. It requires the following process:detect the edge lines after which the rough contours of the crack lines couldappear;next, make the ununiform crack lines grow. After the above steps, anintial binary crack line image is obtained which is consistent with the initial grayimage. Propose an algorithm to make the crack lines grow adaptively bycontrasting to the initial gray image.The second step is to recognize fragmentation particles. Some imageprocessing before image segmentation is necessary including permiting someconcavity about the detected crack lines. It could be completed by computing thedistance function from the pixels in the fragmentation particles to those on thecrack lines, then using the grayscale reconstruction theory to reconstruct thedistance function image. Next, reverse the reconstructed distance function. Thusit prepares for the most critical step—image segmentation. At last, apply thewatershed transformation algorithm based on chaincode to the reversed image.The fragmentation image could be divided into several regions. We get the finalcrack lines whose width is one pixel and the number of the regions which is justthe number of the fragmentation particles.Propose an image segmentation algorithm based on edge detection, integratethe edge dection, region growth and image segmentation algorithms organically.In the total algorithm, traditional edge detection could draw the rough contour ofthe crack lines;the adaptive crack lines growth method could remove theredundant background pixels, connect the uncontinuous crack lines to form theclosed contour, then remove the false object pixels by getting rid of smaller noise;and watershed transformation achieved the image segmentation completely.Experiments show that the image segmentation algorithm based on edgedetection could draw the edge line of glass fragmentation and divide the imageinto regions correctly and count the number of fragmentation particles.