| The viscose filament is a kind of viscose fiber, it is a material that can be well used in clothing industry, there are more requirements for viscose filaments with high quality in the market, especially after China join in the WTO. However, the export of viscose filaments in our country has been restricted due to the quality of domestic viscose filaments which is hard to come up to the international standard. One of the reasons which cause the quality problem is the uneven of dyeing of burden distributions caused by short of single viscose filament. The reason is because some spinneret holes are blocked during spinneret。Nowadays, in China, we use handwork to count the number of viscose filaments. It is very time-consuming, vigor- consuming, low- efficiency and low- accuracy, can not detect huge amounts of viscose filaments.Recently, with the development of computer vision, image processing, pattern recognition technique, the detection technology of machine vision has been improved rapidly, there are lots of successful cases which are good for the research of new detection system of machine vision. However, due to different targets, field, and detection environment, we need to do some pertinence work to different applications of machine vision system.This paper is about the study the field of machine vision; it introduces the structure and theory of viscose filament collection end surface of counting system, it also introduces a kind of image processing and counting algorithm which suits the project combined with image processing technique and practical imaging effect. We describe the debugging and application of the practical algorithm during the development process.Among which, the algorithm of the image processing of viscose filament collection end surface is the working emphasis. The machine vision system has a very ripe architecture, however, in different situations we need to design different organization forms and different measures to obtain image information. There are some points as follows:⑴According to the fiber properties and forms of viscose filament, we put forward a measure to count the numbers of collection end surface, through this measure, we can improve the efficiency in the factory.⑵We design hemispherical multiple angle LED as the light source. By practical verified, the effect of the light source is very well.⑶According the experiment condition and the cost, we use a special slicer which is made by myself to meet the requirement of field of view.⑷According to the image characteristic of collection end surface of viscose filaments, and the algorithm of gradient edge extraction, we puts forward a new measure suit for threshold segmentation to obtain the image information of the end surface accurately. ⑸We study the image of the end surface after threshold segmentation, and according to the image characteristic, we use the distance transform method to count the number of connected domains.We study the system from both sides of hardware and software.1. HardwareThe automation system in industry must be simple and applied, this system has four parts: light source, slicer photoelectric imaging, PC data processing and output.⑴Light source: in order to get a better effect, we use a hemispherical multiple angle LED, at the limited experiment condition, we get a even brightness image which meets the requirement.⑵Slicer: in order to meet the requirement of the field of view and the magnification, we use a modified slicer, the side has been charged from 3×0.8mm to 0.6×0.8mm.⑶Photoelectric imaging: we use a biomicroscope as the light path system, and a CMOS industrial camera to obtain the image information.⑷PC data processing: We use a common PC to do the job to reduce the cost.The working principle of the system: First, we use the light source to light the collection end surface of the viscose filament, according to the fiber properties of the viscose filament, the light will pass through the viscose filament, and after the process of the light path system, all the lights are focusing at the surface of COMS which will turn the light single to digital single. Then, the single will be sent to the PC to do image processing. At last, we get a result. All the parts of the system are hardware except the PC image processing software.2 .SoftwareThe viscose filament collection end surface counting system is a kind of image processing software based on the VC++ programming environment. The software do a comprehensive utilization of all the existing image processing algorithm, and the main algorithm is indigenous designed which can do a past process of the image. According the software processing flow, this paper introduces the algorithm of every step start from the loading of the image, and gives a compare of all the algorithms, includes the image preprocessing, image threshold segmentation, connected domains counting and threshold. The simple introduce as follows:⑴Due to the quality of the image, we need to do lots of enhance and preprocessing work. With which there are gray level transformation, image smoothing and so on. The preprocessing image is meet the requirement of the threshold segmentation.⑵Threshold segmentation use Gradient-based adaptive threshold segmentation which was independent design,which can remove the background part effectively when keep the effective information. After the morphological operation, median filtering and binary processing of the image, the quality is very well, which lighting the burdens for the next processing. ⑶Image end surface counting method use line segment encoding,then discriminant simply connected region according to regional characteristic parameters,then use simple counting method and regional core seed search method which according to distance transform,for one ting, it reduce the cost of time, for another, it do not influent the accuracy. We useThis system uses a biomicroscope as the light path system, which can meet the requirement of the experiment. After all the module combined, we use the SDK development kit which is serviced by DLC-300L to achieve the digitalization, and obtain the target image with high quality. We study the related algorithm by analyzing the image characteristic of the viscose filament. Threshold segmentation use Gradient-based adaptive threshold segmentation.Connected domain use line segment encoding and regional core search method.It can achieve the target proved by experiment. |
PDF Full Text Request