Automatic detection of defective products is generally based on machine vision technology, also known as vision detection system equipment
In the assembly line of automobile parts, visual inspection system has been an indispensable part.
Automatic assembly line has the characteristics of mass production, strong dependence between each process, high precision and stability requirements. Therefore, the production environment, the characteristics of the object under test, and the accidental factors that can and cannot be predicted greatly affect the design scheme of the visual system, and the rationality and accuracy of the measurement algorithm need to be carefully analyzed and verified.
In the actual design process, on the one hand, the production department will put forward the allowable error range of standard products, on the other hand, the visual system has its own accuracy level that can be achieved. Visual inspection is effective only when the level of accuracy achieved by the visual system is higher than that of the standard product. This paper will take the size detection of spring assembly as an example, study and analyze the accuracy problem in visual detection, and put forward corresponding solutions.
1. System description
The vision system is used to check the assembly quality of the spring assembly. The whole detection process uses two cameras, set at 90° Angle on both sides. When the spring assembly moves to the front of the lens, the lens collects images from two directions at the same time, and the measurement data is calculated by special vision software. The quality status of the product is displayed as OK or NG on the LCD screen, and the measured values of each parameter are given. If the quality is OK, continue to perform the following process; If it is NG, the workstation will alarm and suspend work for processing
2. System functions
The visual system monitors the processing size and assembly quality of the main and auxiliary spring components of each product type. The main indicators include: the outer diameter of the spring, the outer diameter of the piston and the measurement of concentricity. Strictly control each index within the error range of quality requirements, comprehensive evaluation of product quality, to achieve the function of quality control.
Visual inspection system design
The following steps are usually used to build a visual detection system
● Build an optical system to obtain good quality images. The image should be able to highlight the characteristics of the tested object and facilitate object extraction, which is the prerequisite to determine the success of the system design.
● Image preprocessing. Filter image noise and extract object features by binarization, edge sharpening and other image processing algorithms.
● Position adjustment and calibration. Locate the area of interest and convert pixels to millimeters.
● Detection algorithm design. For the specific application design appearance size measurement, character reading, two-dimensional code reading and other detection algorithms, programming and design software expansion module for special application occasions.
● System test. The stability and precision of the system need to be tested in the trial production for the large volume detection task. Analyze misjudgment, missed judgment and other abnormal conditions, improve the above steps, count the success rate of visual detection, until the production requirements are met.
2 light source, lens selection and imaging effect
Since the measurement is the external dimension of the spring assembly, using an LED backlight can effectively highlight the edges of the object being measured. In addition, considering that the installation position of the optical system consisting of lens, object and light source is limited by the reserved space on the machine, the standard lens with 12mm focal length is adopted. The maximum length of the measured object is about 100mm, that is, the field of view is required to be greater than 100mm, and the distance between objects is approximately 200mm. Turning down the aperture to suppress reflections on metal surfaces and fine-tuning the relative distance between the lens, object and light source can result in a sharp image with an object's edge.
3 Image preprocessing
For exterior dimension detection, edge clarity directly affects the accuracy of detection. In this system, 3×3 edge sharpening template is used to process the original image.
Position adjustment and calibration
Since the position of each product in front of the camera will change slightly, the measured object may deviate from the preset ROI after the lens magnification, so the position of ROI needs to be adjusted accordingly. Firstly, find the obvious and stable part in the image, such as the three threads in the middle of the spring, and then adjust the position of other ROI according to the position of this part. The image below shows the effect after adjustment.