about the author
Zhang Weiwei, male, School of Electronic Information, Xi'an Polytechnic University, 2019 graduate student, Zhang Hongwei Artificial Intelligence Research Group.
Research direction: machine vision and artificial intelligence.
Email: [email protected]
Personal CSDN homepage: Welcome to pay attention and exchange and learn from each other .
Project master plan
1. Demand analysis
my country is a big textile country, and the textile industry is one of the pillar industries of our national economy. At present, the automatic textile defect detection systems on the market at home and abroad are expensive to develop and maintain, and the investment recovery period is long, which cannot meet the requirements of universal application. In order to further improve industrial automation and reduce enterprise costs, an intelligent detection system for fabric defect detection is urgently needed. It is worth noting that for defect detection in small batch production fabric scenarios, due to the challenges of complex patterns and few defect samples, intelligent The algorithm design of the fabric defect detection system has become a difficult point, and the market urgently needs the development of an automatic fabric defect detection system with frequent pattern iterations. In addition, because deep learning relies on a large number of samples, the problem of feature extraction for small samples is also a difficulty in deep learning theory.
Therefore, the automatic fabric defect detection algorithm is of great significance to reduce enterprise costs and improve product quality.
2. Overall project design
The overall functions that need to be completed are : in the high-end fabric defect detection scene suitable for small batch production with complex and changeable patterns, instead of the worker inspection process between the cutting and sewing processes, the automatic fabric defect detection algorithm can be realized, and the detection speed is required. It is lower than the white grey cloth. At present, the enterprise requirements are more efficient than manual inspection, and the algorithm design and simple implementation are completed, and finally commercial iterative mass production.
The goals to be achieved:
(1) Universality: Since the types of textiles produced by each enterprise are different, the needs of each enterprise are also different, which requires our intelligent terminal equipment to be able to respond to any type of textiles according to the needs of the enterprise. (2) Economical practicability
: Our products are mainly targeted at domestic small and medium-sized textile enterprises. Due to the high cost and selling price of existing fabric defect equipment at home and abroad, many Small and medium-sized enterprises are discouraged. In contrast, our products are much cheaper in terms of cost and selling price, and the detection effect has not been greatly reduced;
(3) Easy industrialization: all components required by our products are Whether it is terminal equipment, monitors and mouse buttons, or cloud computing services, there is already a mature industry chain, and our employees only need to assemble and debug the purchased components, so our products are easy to realize industrialization. .
2.1 Software scheme design
In the figure, the software design is mainly divided into three parts, namely data acquisition and processing, algorithm design and analysis, and interface construction. The functions and corresponding work that the specific software needs to complete are shown in detail in the figure.
3.1 Hardware Solution Design
The following is a conceptual diagram of the system:
The intelligent detection system for small yarn-dyed shirt cuttings designed in this work is mainly composed of an image acquisition chassis, an NVIDIA Jetson TX2 embedded development board, and a display. The overall structure is shown in the figure below.
In the picture: 1. Semi-enclosed piece image acquisition chassis with ring light source; 2. Color industrial camera and lens for image acquisition of yarn-dyed pieces; 3. Ring light source; 4. Jetson TX2 embedded development board; 5. Display.