Research And Application On The Sorting Impurity From Waste Paper Based On Machine Vision

The secondary recycling of waste paper can effectively reduce the problems thatthe papermaking industry facing such as the shortage of funds, raw material supply andthe low production capacity, etc. Previous sorting methods basically rely on manualselection, and the efficiency and quality are difficult to maintain. The safety and healthof the workers are also difficult to protect. In automatic sorting, relevant equipment isstill in experimental stage, or the accuracy is still far from reaching the point of practicalapplication, which severely restricts the sustainable development of waste paperrecycling industry. With the increasing of energy conservation and emission reduction,it is extremely urgent to develop the highly integrated waste material recycling system.In this paper, the real-time acquisition and preprocessing technology of highdynamic range image based on the trace of impurity is researched and developed. Firstof all, the mathematical model based on PC, the target object, CCD sensor andmanipulator is established. The position and structure of CCD camera are designed inorder to acquire the best quality image. Then the high dynamic range image real-time isacquired based on the establishment of the image intensity and shutter database, as wellas the calibration technology of gray card CCD sensor. Then feature transform and edgedetection technology are used to realize image filtering and feature point extraction. Atlast, the multi-level and multiresolution image information is fused. The application ofwaste paper impurity motions tracking technology is promoted based on the precisepaper impurities image information.The main content of this paper include:1) A new method for image recognition based on the grass-roots level imageinformation is proposed. Image edge profile information as one of the most basic imageinformation can finish the reliable characterization with high promptness.2) The irregular distributions, uncertain shapes, mutations, and randomdistributions are considered, and a new recognition algorithm based on invariantmoments and edge histogram similarities (IM-EHCVS) is proposed. This algorithm cancombine the global and local information together, and finish comprehensivedescription of the image without hurt the characteristic information.3) An effective image processing algorithm (IMFD) based on parallel mechanismis proposed, in order to reduce redundant information during image information processing. The algorithm based on invariant moments and fractal dimension, whichcan be integrated the global information and local information into an integrationdescriptor. The new descriptor can express the reaction of the image informationcompletely. The algorithm can perform image processing and information extraction inparallel way, improve the system response speed and meet the real-time requirements inpractical application.4) The high-level information characteristics of image information as well as theclassification of the visual system is investigated. The recognition system is processedon the specific image matching effect for image. SIFT is used to extract the sensitivefeatures. The matching is processed in the global mode based on the high-levelinformation and concise expression. This method can classify the impurities intodifferent categories effectively.It is proved in practice that the intelligent sorting system can completeidentification and sorting the impurities from waste paper accurately and quickly withgood robustness and anti-interference.