Cotton Yarn Inspection
Inspect Machined Parts for Defects Quickly and Efficiently using Pixuate's Technology
IntroductionDefect detection is highly important to fabric quality control. Traditionally, defects are detected by human eyes. The efficiency of this manual method is low and the missed rate is high because of eye fatigue. In the best case, a quality control person cannot detect more than 60–70 % of the present defects. Hence, an automatic inspection system is necessary for textile industry. Pixuate has developed an automatic inspection system using Deep Neural Networks to avoid manual errors and increase the efficiency.
Key Features Of Cotton Yarn Inspection
- A centralized repository for tracking defects.
- Automated notifications of resource assignments.
- Ability to define defect resolution status in order to map back to your defect management process.
- Provides management reporting, like the number of open defects grouped by various criteria such as open defects by project, severity, and priority.
- Ability to capture other items in addition to defects, such as customer suggestions and project-related issues. Items such as customer complaints or enhancement suggestions are often lost if not logged in a centralized system. If other tools are not already available, a defect management tool can be used to track these types of items as long as they can easily be filtered out or logically separated from defects.
- Ability to support internal and external teams. This feature provides the opportunity to involve external teams and in some situations, customers if appropriate.
How it works ?Pixuate® uses Deep Neural Networks for development of efficient inspection systems. In the conventional method, the analysis is done exclusively by the X-ray inspector. The progress in computer science and the artificial intelligence techniques have allowed the defect classification to be carried out by using pattern recognition tools, especially the methods that use artificial neural networks, which make the process automatic and more reliable, as it is not a subjective analysis. The techniques for pattern recognition implemented on a digital image involve, mainly, the location and isolation of the objects (defects) on the image, and later, the defect identification (classification). The pattern recognition assumes that the image may have one or more defects, and that each defect belongs to a preset type, category or pattern class of defects. If a digitized image has several defects, the process for pattern recognition consists of three main steps.
The first step is the image quality improvement or the defect isolation, in which each defect is located and isolated from the rest of the image. Several types of degradations and distortions, inherent in the processes of image acquisition, transmission and display contribute to limit the capacity to extract information. The main purpose of this image quality improvement step is to suppress or reduce the consequences of these barriers, making it easy to extract the relevant information. This improvement is usually done with the application of image digital filters.
The second step is the extraction of defect features, that is, where the defects are measured. One measurement is the value of any sizeable property of defect. A feature is a function of one or more measurements, which are registered in the computer to dimension any significant characteristic of the defect. This drastic reduction in the amount of information (if compared to the original image) represents all the knowledge on which the subsequent process of classification shall be based. Therefore, the fewer features existing to represent a certain class, the fewer information will be processed. The use of features of defects is one of the mostly used techniques to classify the defects after their detection. In this case, the correct selection of the most relevant features in the identification of each class has a great importance in the recognition of such classes done by the intelligent system. This selection is similar to the interpretation given by an inspector who, in most cases, recognizes first one type of welding defect in the radiography by visual characteristics, such as: location, shape, length, density (gray level), aspect ratio, etc., besides the welding conditions. So, an important study of the defect morphology at the image level is required to optimize the system performance.
The third step of the pattern recognition process is the classification. This output is merely a decision considering the class that each defect belongs to. Each defect is considered to be an individual class, and the recognition is implemented as a process of classification. Each defect is directed to one of the several preset classes, which represent all possible types of defects expected to exist in the image.
Advantages of Pixuate® Inspection System
- Reduces Fix and Remediation Cost
- Reduces Overall Spend
- Increased Manufacturing and QA Staff Productivity
- Reduces Business Risk Due to Outages
- Improves Security and Overall Manufacturing Quality
- Detect Defects Earlier
- Fulfill Architecture Standards
- Improve Manufacturing Quality
- Minimize Complexity
- Verifying Manufacturing Practices
- Lower Technical Debt
- Pixuate’s Inspection systems eliminates all the manual errors and provides a very high accuracy of defect identification.
- Use of cognitive methods such as deep learning, social learning, and tensor factorization which are based on machine learning and neural networks.
- Increases overall manufacturing throughput of the manufacturing process and reduces the risks due to defects.