| The application of machine vision technology to the online industrial detection has become one of the research focuses in recent years. Machine vision technology is widely used in the field of industrial detection and control because it can effectively reduce the detection cost, ensure products quality and raise the productivity effect. However, the application of machine vision in high speed on-line detection is still a difficult problem because of the huge quantity of visual information data and the complicated operation. Furthermore human and animal visual mechanism has not figured out completely so far, as a result, the development of the computer vision theory and the machine vision detection technology is restricted to a certain extent. The machine vision detection technology is not mature or perfect at all, which still needs massive research work. As a result of the limitation of the machine vision technology, we need to design different system structures and develop the corresponding algorithms to meet concrete demands in the actual industrial detection, according to various detection system.In the chemical fiber industry, the viscose filament is popular for its superior performance such as fine water absorbability, pliability, skin-proximity, etc. in addition the market demand of the high quality V. filament unceasingly rises, especially after China joined WTO. However, the filament made in China could not achieve the international standard, its exportation is restricted. One of the most important reasons is the insufficiency of monofilaments of V. filament, which causes the cotton material dyeing uneven. In the high speed production of V. filament, some pin holes of spurts board are sealed, so the number of monofilaments which should be twisted into one filament is insufficient. Besides, the problem is so difficult to discover that it always lasts for a long time, the product line may make lots of waste products, it is a waste of raw material and cost. At present, domestic factory mainly uses off-line detection for filament quality artificially, because of the acid liquor environment, on-line detection artificially is impossible. However, off-line detection is low efficiency and unable to prevent waste from being produced. Some oversea factories use microscope to carry on the manual observation, which still could not realize automatically interprets and computer data processing, with a low accurate rate. Our goal is to improve the product quality with low consumption, but manual detection cannot make it, so the application of machine vision technology to the online industrial detection has become one of the most urgent methods.In view of the above situation, founded on "Research on 50,000 spindles V. filament in bath detection system of Jilin chemical fiber Co. Ltd." of Jilin Province Science and Technology Planning Projects, we study the key technology of V. filament quality detection. We focus on how to inspect fast and accurately and judge whether the spurts board is sealed and the number of monofilaments is enough or not. The acid liquor's transparency is about 400mm to 500mm. in this foundation we design and the development the detection system, thus improves quality of the filament, and avoids dyeing cloth unevenly. This work could solve problems mentioned above.In this thesis, the system design schema and key algorithms are proposed. Firstly, we introduce background knowledge and the system structure, including the user requirement analysis, system function and architecture. The second, we particularly discuss the research of relative image processing algorithm following the requirement of online high-speed inspecting. The algorithms involve image segmentation, image recognition and object measurement and so on.The main jobs are: a improved algorithm about image segmentation by thresholding based on edge information is presented to take full advantage of the ample edges in V. filament's monofilaments. The inspiration is from the algorithm in fingerprint recognition system which can separate fingerprint from background;For uneven illumination and some noise exist in background, the image has obvious characters such as high contrast, the grey-level histogram is complex, so we design improved optimized segmentation by threshold based on grey-level histogram distribution and image segmentation by line-by-line adaptive thresholding;According to specific demand of the project, we design monofilament scanning and tracking algorithm and connected domain analysis algorithm in order to measure the monofilament information, as there isn't mature algorithm available;At the same time, we describe the hardware architecture briefly, according to the actual situation and the environment, we use special photosource to improve the uniformity of luminosity in the acid liquor bath and enhance the quality of the V. filament picture. We choose close field lens in order to guarantee the depth of field, make full use of area array CCD, enlarge the slender monofilament as much as possible. We also use anticorrosion system structure and material to ensure the system avoid from being corroded in acid liquor bath.We make use of DirectShow technology to realize shooting pictures of V. filament real time, design and develop MIS for results of detection. Our simulate experimentation shows that this system can detect the quality defects of V. filament on high-speed product line, and it performs well in aspects such as speed, accuracy and stability.At present, V. filament in bath detection system that can detect the quality defects intelligently and automatically is not exist in domestic chemical fiber industry. It will bring us enormous convenience to quality detection of textile industry if we introduce machine vision technology into this research area. Furthermore, machine vision technology will play a much more important role in industrial detection in the future!... |
PDF Full Text Request