For enquiry (Roger Ge):
For support (Eason Tsang):
In the production process of the cloth, the work with high repeatability and intelligence like cloth quality detection can only be done by manual detection, and a lot of detection workers can be seen in the rear of the modernization pipeline to carry out the process, While adding huge labor cost and management cost to the enterprise, it is still not possible to guarantee 100% of the inspection qualification rate (i.e., the “zero defect”). the detection of the cloth quality is repetitive labor, and is prone to error and has low efficiency. the production line of the cloth production line is changed into a fast, real-time, accurate and high-efficiency pipeline. In the pipeline, the color and quantity of all cloth shall be automatically confirmed (hereinafter referred to as" “cloth detection” ").
The automatic identification technique of machine vision is used to complete the work that has been done manually. in a large-scale cloth detection, that product quality is low and the precision is not high by manual inspection, and the machine vision detection method can greatly improve the production efficiency and the automation degree of production.
Feature extraction identification
General cloth detection (automatic identification) first uses the high definition, high-speed camera lens to take the standard image, set a certain standard on this basis, then shoot the detected image, and then compare the two.
However, in the quality test of the cloth, it is more complicated:
1. The content of the image is not a single image, and the number, size, color and position of the impurities present in each area under test shall not be consistent.
2. The shape of the impurities is difficult to determine in advance.
3. A large amount of noise may be present in the image due to the reflection of the light by the rapid movement of the cloth.
4. On the pipeline, the cloth is tested and the real-time performance is required. For the above reasons, the image recognition processing should take the corresponding algorithm to extract the features of the impurity, and carry out the pattern recognition to realize the intelligent analysis.
In general, the images acquired from the color CCD camera are RGB images. That is, each pixel is composed of three components of red (R) green (G) blue (B) to represent a point in the RGB color space. the problem is that these color differences are different from the perception of the human eye. even small noise changes the position in the color space. So no matter how close we feel, it's different in the color space. For the above reason, we need to convert the RGB pixels into another color space CIELAB. The goal is to make the human eye feel as close as possible to the color difference in the color space.
According to the processing image obtained above, the impurity color spot is detected in a solid color background according to the requirement, and the area of the color spot is to be calculated to determine whether or not it is within the detection range. Therefore, the image processing software has a function of separating a target, detecting a target, and calculating an area thereof. Blob Analysis is the analysis of the communication domain of the same pixels in the image, referred to as Blob. The color spot in the image processed by Binary Thresholding can be considered to be a blob. The Blob analysis tool can separate the target from the background, and can calculate the number, position, shape, direction, and size of the target, and can also provide the topology between the related spots. instead of using a single pixel to analyze one by one in the process, the rows of the pattern are operated. Each row of the image is encoded with a run length (RLE) to represent an adjacent target range. This algorithm greatly improves the processing speed compared with the pixel-based algorithm.
Results processing and control
The application stores the returned result in a database or a user-specified location and controls the mechanical part to do the corresponding movement according to the result. and the information management is carried out in the database according to the result of the identification. the information can be retrieved at any time, and the manager can know the busy of the pipeline during a certain period of time, and arrange the work of the next step; and the quality condition of the inner cloth and the like can be known.