The composition of the machine vision system
Machine vision system refers to the realization of human visual function by computer, that is, using computer to realize the recognition of objective three-dimensional world. As is now understood, the sensory part of the human visual system is the retina, which is a three-dimensional sampling system. The visible portion of the three-dimensional object is projected onto the omentum, and the person performs a three-dimensional understanding of the object according to the two-dimensional image projected onto the retina. The so-called three-dimensional understanding refers to an understanding of the shape, size, distance from the observation point, texture and motion characteristics (direction and speed) of the observed object.
The input device of the machine vision system can be a camera, a drum, etc., all of which use a three-dimensional image as an input source, that is, a two-dimensional projection of a three-dimensional view of the world into the computer. If the three-dimensional objective world to the two-dimensional projection image is regarded as a kind of positive transformation, then the machine vision system has to do the inverse transformation from the two-dimensional projection image to the three-dimensional objective world, that is, according to the two-dimensional projection. The image is to reconstruct the three-dimensional objective world.
The machine vision system consists of three main parts: image acquisition, image processing and analysis, output or display.
Nearly 80% of industrial vision systems are used primarily for inspection, including for improving production efficiency, controlling product quality during production, and collecting product data. Product classification and selection are also integrated into the inspection function. The following describes the composition and function of the system through a single camera vision system for the production line.
The vision system detects the products on the production line, determines whether the products meet the quality requirements, and according to the results, generates corresponding signals to the upper computer. The image acquisition device includes a light source, a camera, etc.; the image processing device includes a corresponding software and hardware system; the output device is a related system connected to the manufacturing process, including a process controller and an alarm device. The data is transmitted to the computer for analysis and product control. If a non-conforming product is found, the alarm is alarmed and excluded from the production line. The result of machine vision is the source of quality information for the CAQ system and can also be integrated with other CIMS systems.
Image acquisition
The acquisition of images is actually to convert the visual image and intrinsic features of the measured object into a series of data that can be processed by the computer. It is mainly composed of three parts:
* Lighting
* Image focusing formation
* Image determination and formation of camera output signals
1, lighting
Illumination and important factors affecting the input of the machine vision system, as it directly affects the quality of the input data and at least 30% of the application effect. Since there is no universal machine vision lighting device, the corresponding lighting device should be selected for each specific application example to achieve the best results.
In the past, many industrial machine vision systems used visible light as a light source, mainly because visible light was readily available, inexpensive, and easy to operate. Several commonly used sources of visible light are white, fluorescent, mercury, and sodium. However, one of the biggest drawbacks of these sources is that light energy cannot be kept stable. Taking fluorescent lamps as an example, in the first 100 hours of use, the light energy will drop by 15%, and as the use time increases, the light energy will continue to drop. Therefore, how to make the light energy stable to a certain extent is an urgent problem to be solved in the practical process.
On the other hand, ambient light will change the total light energy that these light sources illuminate onto the object, causing noise in the output image data. Generally, a protective screen is added to reduce the influence of ambient light.
Due to the above problems, in today's industrial applications, for some highly demanding inspection tasks, invisible light such as X-rays or ultrasonic waves is often used as a light source. However, the invisible light is not conducive to the operation of the detection system, and the price is high. Therefore, in practical applications, visible light is still mostly used as a light source.
The illumination system can be divided into: backlight, forward illumination, structured light and stroboscopic illumination according to its illumination method. Among them, the back illumination is placed between the light source and the camera, and has the advantage of obtaining a high contrast image. The forward illumination is that the light source and the camera are located on the same side of the object to be tested, which is convenient for installation. The structured light illumination projects a grating or a line source or the like onto the object to be measured, and demodulates the three-dimensional information of the object to be measured according to the distortion generated by them. The stroboscopic illumination is to illuminate a high-frequency light pulse onto an object, and the camera is required to synchronize with the light source.
2, image focus formation
The image of the object to be measured is focused on the sensitive component by a lens, just like a camera. The difference is that the camera uses film, while the machine vision system uses sensors to capture images, and the sensor converts the visible image into an electrical signal for computer processing.
Selecting a camera in a machine vision system should be based on the requirements of the actual application, where the lens parameters of the camera are an important indicator. The lens parameters are divided into four sections: Magnification, Focal Length, Depth of Field, and Lens Mount.
3. Image determination and formation of camera output signals
The machine vision system is actually a photoelectric conversion device, which converts the lens received by the sensor into an electrical signal that can be processed by the computer. The camera can be an electronic tube or a solid state sensing unit.
Tube camera development was earlier. It was applied to commercial TV in the 1930s. It uses a vacuum tube containing light-sensitive components for image sensing, and converts the received image into an analog voltage signal output. A camera with an RS-170 output system can be connected directly to a commercial TV display.
The solid state camera was developed in the late 1960s when the Bell Telephone Lab invented the Charge Coupler (CCD). It is composed of a linear array or a rectangular array of photodiodes distributed on the respective pixels, and the image optical signals are converted into electrical signals by outputting voltage pulses of each diode in a certain order. The output voltage pulse sequence can be directly input into the standard TV display in the RS-170 format, or input into the computer's memory for numerical processing. CCD is the most commonly used machine vision sensor.
Image processing technology
In machine vision systems, visual information processing technology mainly relies on image processing methods, including image enhancement, data encoding and transmission, smoothing, edge sharpening, segmentation, feature extraction, image recognition and understanding. After these processes, the quality of the output image is improved to a considerable extent, which not only improves the visual effect of the image, but also facilitates the analysis, processing and recognition of the image by the computer.
1, image enhancement
Image enhancements are used to adjust the contrast of the image, highlight important details in the image, and improve visual quality. Image enhancement is typically performed using grayscale histogram modification techniques.
The gray histogram of an image is a statistical chart showing the gray distribution of an image, which is closely related to the contrast.
Generally, a two-dimensional digital image represented in a computer can be represented as a matrix, and the elements in the matrix are image gray values ​​located at corresponding coordinate positions, which are discretized integers, generally taking 0, 1, ..., 255. This is mainly because the value range represented by one byte in the computer is 0~255. In addition, the human eye can only distinguish about 32 gray levels. Therefore, it is sufficient to represent the gray scale in one byte.
However, the histogram can only count the probability of occurrence of a certain level of grayscale pixels, and can not reflect the two-dimensional coordinates of the pixel in the image. Therefore, different images may have the same histogram. Through the shape of the gray histogram, the sharpness and black-and-white contrast of the image can be judged.
If the histogram effect of obtaining an image is not ideal, it can be appropriately modified by the histogram equalization processing technique, that is, a pixel mapping gray scale in a known gray probability distribution image is transformed into some kind, so that it becomes A new image with a uniform grayscale probability distribution for the purpose of making the image clear.
2, the smoothing of the image
The smoothing processing technique of the image, that is, the denoising processing of the image, is mainly to remove the image distortion caused by the imaging device and the environment during the actual imaging process, and extract useful information. It is well known that in the process of forming, transmitting, receiving and processing, the actual obtained image inevitably has external interference and internal interference, such as the sensitivity of the sensitivity of the sensitive component of the photoelectric conversion process, the quantization noise of the digitization process, and the transmission process. The error, as well as human factors, can degrade the image. Therefore, removing noise and restoring the original image is an important content in image processing.
3, image data encoding and transmission
The amount of data in a digital image is quite large, and the amount of data of a 512*512 pixel digital image is 256 Kbytes. If it is assumed that 25 frames per second are transmitted, the channel rate of transmission is 52.4 Mbits/sec. High channel rates mean high investment, which also means an increase in the difficulty of popularization. Therefore, it is very important to compress the image data during transmission. The compression of the data is mainly done by encoding and transforming the image data.
Image data coding generally uses predictive coding, that is, the spatial variation law and sequence change law of image data are represented by a prediction formula. If the previous adjacent pixel values ​​of a certain pixel are known, the pixel value can be predicted by a formula. With predictive coding, it is generally only necessary to transmit the start value and prediction error of the image data, so that 8 bits/pixel can be compressed to 2 bits/pixel.
The transform compression method divides the entire image into small pieces (8*8 or 16*16), and then classifies, transforms, and quantizes the data blocks to form an adaptive transform compression system. The method can compress the data of one image into a few dozens of special transmissions, and then transform back at the receiving end.
4, edge sharpening
Image edge sharpening is mainly to enhance the contour edges and details in the image to form a complete object boundary, to achieve the purpose of separating the object from the image or detecting the area representing the same object surface. It is a basic problem in early visual theory and algorithms, and one of the important factors in the success of the medium and late visual.
5, the image segmentation
Image segmentation is the division of an image into parts, each part corresponding to the surface of an object. When segmentation is performed, the grayscale or texture of each part conforms to a certain uniform measure. One essence is to classify pixels. The classification is based on the gray value, color, spectral characteristics, spatial characteristics or texture characteristics of the pixel. Image segmentation is one of the basic methods of image processing technology, such as chromosome classification, scene understanding system, machine vision and so on.
There are two main methods for image segmentation: one is the gray threshold segmentation method in view of the metric space. It is based on the image gray histogram to determine the image spatial domain pixel clustering. However, it only uses the grayscale features of the image, and does not use other useful information in the image, so that the segmentation result is very sensitive to noise; the second is the spatial domain region growing segmentation method. It is a segmentation region for pixel connected sets with similar properties in a certain sense (such as gray level, organization, gradient, etc.). This method has a good segmentation effect, but the disadvantage is that the operation is complicated and the processing speed is slow. Other methods, such as edge tracking, focus on maintaining the edge properties, tracking the edges and forming a closed contour, and segmenting the target; the cone image data structure method and the marker relaxation iteration method also use the spatial distribution of pixels to align the edges. Pixels are properly merged. The knowledge-based segmentation method utilizes the prior information and statistical characteristics of the scene. Firstly, the image is initially segmented, the region features are extracted, and then the domain knowledge is used to derive the interpretation of the region. Finally, the region is merged according to the interpretation.
6, image recognition
The image recognition process can actually be regarded as a marking process, that is, the recognition algorithm is used to identify the divided objects in the scene, and the specific markings are given to these objects, which is a task that the machine vision system must complete.
According to image recognition, from easy to difficult, it can be divided into three types of problems. In the first type of recognition problem, a pixel in an image expresses a certain piece of information of an object. For example, a certain pixel in a remote sensing image represents a reflection characteristic of a certain spectral band of a ground object at a certain position on the ground, and the type of the ground object can be discriminated by the same. In the second type of problem, the object to be identified is a tangible whole, and the two-dimensional image information is sufficient to identify the object, such as character recognition, some three-dimensional body recognition with a stable visible surface, and the like. However, such problems are not easily expressed as feature vectors like the first type of problems. In the process of identification, the object to be identified should be correctly segmented from the background of the image, and then the attribute map of the object in the established image should be managed. Assume that the model library's attribute maps match. The third type of problem is the three-dimensional representation of the measured object from the input two-dimensional map, feature map, 2.5-dimensional graph, and so on. There is a question of how to extract the implicit three-dimensional information, which is a hot topic in this research.
The current methods for image recognition are mainly divided into decision theory and structural methods. The basis of the decision theory method is the decision function, which is used to classify the pattern vector based on the timing description (such as statistical texture); the core of the structure method is to decompose the object into a pattern or pattern primitive, and different Object structures have different primitive strings (or strings). By using a given pattern primitive for an unknown object to find the encoding boundary, a string is obtained, and its genus is determined according to the string. This is a method that relies on symbols to describe the relationship between the objects being measured.
So, what are the difficulties in the design of machine vision systems? This article mainly summarizes the following five points.
First: the stability of lighting
Industrial vision applications are generally divided into four categories: positioning, measurement, detection, and identification. Measurements require the highest stability of illumination, because light changes as long as 10-20%, and the measurement results may deviate by 1-2 pixels. This is not a software problem. This is a change in illumination, which causes the edge position of the image to change. Even if the software is not solved, the problem must be solved. From the perspective of system design, the interference of ambient light must be eliminated, and the active illumination source must be guaranteed. Luminescence stability. Of course, the improvement of resolution by hardware camera is also a way to improve accuracy and resist environmental interference. For example, the previous camera's corresponding object space size is 1 pixel 10um, and by increasing the resolution to become 1 pixel 5um, the accuracy is approximately doubled, and the interference to the environment is naturally enhanced.
Second: the inconsistency of the workpiece position
Generally, the measurement project, whether it is offline detection or online detection, as long as it is a fully automated testing equipment, the first step is to find the target to be tested. Every time the object to be tested appears in the field of view, it is necessary to know exactly where the object to be tested is. Even if you use some mechanical fixtures, you cannot ensure that the object to be tested appears in the same position every time. This requires the positioning function. If the positioning is not accurate, the position where the measuring tool appears may be inaccurate, and the measurement result may sometimes have a large deviation.
Third: calibration
Generally, the following calibrations are required for high-precision measurement, an optical distortion calibration (if you are not using a software lens, generally must be calibrated), the calibration of the second projection distortion, that is, the image distortion correction represented by the installation position error The calibration of the three object image space, that is, the size of the corresponding object space of each pixel is specifically calculated.
However, the current calibration algorithms are based on plane calibration. If the physics to be measured is not planar, the calibration will need to be processed by some special algorithms. The usual calibration algorithm can't be solved.
In addition, some calibrations, because the calibration plate is not used, must also design a special calibration method, so the calibration may not be completely solved by the calibration algorithm already in the software.
Fourth: the speed of movement of the object
If the object being measured is not stationary, but is in motion, then the motion blur must be considered for image accuracy (fuzzy pixels = object motion speed * camera exposure time), which is not software solver.
Fifth: the measurement accuracy of the software
In the measurement application, the accuracy of the software can only be considered in 1/2-1/4 pixels, preferably in 1/2, but not as 1/10-1/30 pixel accuracy as the positioning application, because in measurement applications. The software is able to extract very few feature points from the image.
The lamp is used in the production area of the factory and mine. Applicable to shipyards, mines, workshops, factories, warehouses, highway toll stations, gas stations, large supermarkets, exhibition halls, gymnasium and other places need industrial and mining lighting. Mining lamp is the use of ultra-bright three-color phosphor light source, the shell is made of aluminum alloy, the cover with a high efficiency pc or tempered glass, heat using copper heat conduction aluminum alloy heat dissipation natural air circulation makes the heat more perfect, combined with different customer Demand shape will continue to change, the lamp is only 20% of the traditional lamp, the equivalent of 100W LED mining lamp can replace the traditional 250W lamp, especially in line with the country's "green low-carbon.
UFO LED High Bay Light,LED High Bay Light Fixture,LED Linear High Bay Light,LED High Bay Light 150W
Shenzhen Ri Yue Guang Hua Technology Co., Ltd. , https://www.ledlightinside.com