Research On The Technology Of The Print Blobs Detection

When prints are processed to package-case through the handling process such as cutting and chromatograph ,many defects such as blobs ,inks missing and press deviation are produced, because of the machines'performers and the systems'stability. Usually, prints are inspected by humans with low efficiency and high costs, and the products quality can not be ensured because of the workers'subjective. Besides these, The human inspection method usually does in off-line way,so materials may be wasted when there are many problem products are produced before inspection.This paper studies the detection method which use the machine vision to inspect the prints quality in on-line way, specially, we aim to detect the blobs in the cigarette case. For the detecting cigarette case image, when there is a pixel block which gray level is notable different from the standard one, this block is called a blob. We introduce a system algorithm to detect the print blobs. we first register the detecting image with the standard image. Then we compare the correspond pixels in the two images and the questionable points are found .Lastly we use the morphology methods to analysis the questionable points image and decide if there are any blobs in the detecting image. here, pixels in the detecting image are defined as questionable points when they are notable different from the correspond pixels in the standard image. This algorithm can detect the blob defects in the cigarette case rapidly and accurately with small mistake detection rate.To remove the mistake registration effect , the standard image edges are detected and different methods are used to analysis the detecting image's edge pixels and non-edge pixels ,which decreased the mistake detection rate largely. At the same time, we detect the questionable pixels edge types through edge type analyses, and the questionable points are processed based on their edge types.To meet with the system's high-speed and high-accuracy demands, we register the detecting image with the standard one making use of the corner points detection algorithm and the SSDA algorithm.