Machine Vision is an important branch in the field of artificial intelligence. It is currently in a stage of continuous breakthrough and maturity. It is generally believed that machine vision "is an image that automatically accepts and processes an image of a real scene through optical devices and non-contact sensors, and obtains the desired information or a device for controlling the motion of the machine by analyzing the image". It can be seen that the intelligent image processing technology is in the machine. Vision plays a pivotal position.
Intelligent image processing refers to a kind of computer-based image processing and analysis technology adaptive to various applications. It is an independent theory and technology field, but it is also a very important technical support in machine vision.
Machine vision with intelligent image processing function is equivalent to people giving the machine intelligence while pressing the eyes of the machine, so that the machine can be "visible" and "see the right", which can replace or even better than the human eye to make measurements and Judging, the machine vision system can achieve high resolution and high speed control. Moreover, the machine vision system has no contact with the object to be inspected, and is safe and reliable.
1 machine vision technology
The origins of machine vision can be traced back to the image processing research of the American scholar LR Roberts on the multifaceted volume wood world in the 1960s, and the "machine vision" course of the MIT artificial intelligence laboratory in the 1970s. In the 1980s, a global wave of machine vision research began to emerge, and some machine vision-based applications appeared. After the 1990s, with the rapid development of computer and semiconductor technology, the theory and application of machine vision have been further developed.
After entering the 21st century, machine vision technology has developed at a faster rate and has been applied to many fields on a large scale, such as intelligent manufacturing, intelligent transportation, medical and health, and security monitoring. At present, with the rise of the artificial intelligence wave, machine vision technology is in a new stage of continuous breakthrough and maturity.
In China, the research and application of machine vision began in the 1990s. Starting from tracking foreign brand products, after more than 20 years of hard work, domestic machine vision has grown from scratch, from weak to strong, not only has theoretical research progressed rapidly, but also some competitive companies and products have emerged. It is estimated that with the continuous deepening of research, development and promotion of machine vision in China, it is not unattainable to catch up with and surpass the world level.
Common machine vision systems can be divided into two categories, one is based on computers, such as industrial computers or PCs, and the other is more compact embedded devices. Typical industrial machine-based machine vision systems mainly include: optical systems, cameras and industrial computers (including image acquisition, image processing and analysis, control / communication) and other units. The machine vision system requires accurate, fast and stable algorithms for the core image processing. At the same time, it also requires low system implementation cost and easy upgrade.
2 intelligent image processing technology
The machine vision image processing system calculates and analyzes the digital image signals of the scene according to specific application requirements, and controls the actions of the field devices according to the obtained processing results. The common functions are as follows:
(1) Image acquisition
Image acquisition is the process of acquiring scene images from the work site. It is the first step of machine vision. The acquisition tools are mostly CCD or CMOS cameras or cameras. The camera captures a single image and the camera captures continuous live images. In the case of an image, it is actually a projection of a three-dimensional scene on a two-dimensional image plane, and the color (brightness and chromaticity) of a point in the image is a reflection of the corresponding point color in the scene. This is the fundamental basis we can use to capture images instead of real scenes.
If the camera is an analog signal output, the analog image signal needs to be digitized and sent to a computer (including an embedded system) for processing. Most cameras now output digital image signals directly, eliminating the need for analog-to-digital conversion. Not only that, but now the camera's digital output interface is also standardized, such as USB, VGA, 1394, HDMI, WiFi, Blue Tooth interface, etc., can be directly sent to the computer for processing, in order to avoid adding an image between the image output and the computer The trouble of collecting cards. Subsequent image processing is often done in software by a computer or embedded system.
(2) Image preprocessing
For the collected digital scene images, due to the influence of equipment and environmental factors, they are often subject to different degrees of interference, such as noise, geometric deformation, color imbalance, etc., which will hinder the subsequent processing. To do this, the acquired image must be preprocessed. Common pre-processing includes noise cancellation, geometric correction, histogram equalization, and the like.
Time domain or frequency domain filtering is usually used to remove noise in the image; geometric transformation is used to correct the geometric distortion of the image; and histogram equalization and homomorphic filtering are used to reduce the color deviation of the image. In summary, through this series of image pre-processing techniques, the acquired images are "processed" to provide "better" and "more useful" images for body machine vision applications.
(3) Image segmentation
Image segmentation is to divide the image into regions with different characteristics according to the application requirements, and extract the target of interest from it. Common features in images are grayscale, color, texture, edges, corners, and the like. For example, the image of the assembly line of the automobile is divided into a background area and a workpiece area, and is supplied to the subsequent processing unit for processing the workpiece mounting portion.
Image segmentation has been a problem in image processing for many years. So far, there are many kinds of segmentation algorithms, but the effects are often not ideal. Recently, people use the deep learning method based on neural network to perform image segmentation, and its performance is better than traditional algorithms.
(4) Target identification and classification
In the manufacturing or security industries, machine vision is inseparable from the identification and classification of the target of the input image, in order to complete the subsequent judgment and operation on this basis. There are many similarities in the identification and classification techniques, and often the target category is clear after the target recognition is completed. Recently, image recognition technology is crossing traditional methods to form intelligent image recognition methods based on neural networks, such as convolutional neural networks (CNN) and recurrent neural networks (RNN).
(5) Target positioning and measurement
In smart manufacturing, the most common job is to install the target workpiece, but it is often necessary to locate the target before installation, and the target needs to be measured after installation. Both installation and measurement need to maintain high precision and speed, such as millimeter accuracy (or even smaller), millisecond speed. This high-precision, high-speed positioning and measurement is difficult to achieve by relying on conventional mechanical or manual methods. In machine vision, the image processing method is used to process the installation site image, and the complex mapping relationship between the target and the image is processed to complete the positioning and measurement tasks quickly and accurately.
(6) Target detection and tracking
The moving target detection and tracking in image processing is to detect whether there is a moving target in the scene image captured by the camera in real time, and predict its next moving direction and trend, that is, tracking. These motion data are submitted to subsequent analysis and control processing in time to form corresponding control actions. Image acquisition generally uses a single camera, and if necessary, two cameras can be used to mimic the binocular vision of a person to obtain stereoscopic information of the scene, which is more advantageous for target detection and tracking processing.
3 machine vision applications
Machine vision is used in a wide range of applications, such as security, manufacturing, education, publishing, medical, transportation, and military. Intelligent image processing is indispensable in the application of these machine amounts. Here are just a few of the applications.
(1) Intelligent manufacturing
In order to achieve the grand goal of China's smart manufacturing 2025, machine vision is inseparable. Xuntong Technology is a system for automatic positioning, detection and identification of door stoppers developed by a well-known automobile manufacturer. The system automatically detects whether the model is correct or not, and replaces the manual operation by intelligent image recognition. The detection accuracy reaches 100%. Previously, each station required 4 workers to check and locate 16 types of positioners with their eyes. The employees were not only very tired, but also often made mistakes.
(2) Educational examination
Examination papers often find that students are affected by typography or printing errors. Using intelligent image processing technology, the machine automatically compares the printed test papers with the original test papers. When the inconsistencies are found, they will automatically prompt and alarm. Manually verify the test paper.
(3) Publishing and printing
Similar to the education test, professional publishing and printing companies often make mistakes in typesetting and printing due to the variety of printed books, newspapers and magazines, and the variety of packaging and promotional materials from companies. To this end, many professionals need to be arranged for proofreading, which consumes a lot of money and time. Automated proofreading by using intelligent image processing technology not only improves proofreading accuracy, but also shortens proofreading time, reduces printing costs, and shortens the publication delivery cycle.
(4) Security monitoring
This is an area that is currently being watched by machine vision. Machine vision breaks the limitations of traditional video surveillance systems, increases the intelligence of the system, and enables intelligent video analysis to be implemented step by step. Taking video surveillance in public places as an example, by using machine vision technology, automatic detection, face recognition, and real-time tracking of suspicious people can be realized. If necessary, multi-camera tracking can be realized, and alarms can be issued to store on-site information.
(5) Intelligent transportation
Machine vision has a wide range of applications in the transportation sector. For example, on the expressway and at the bayonet, the vehicles and licenses of the vehicles are recognized, and even the violations of the vehicles are identified. The driver's face image is analyzed on the car to determine whether the driver is in a fatigue driving state. For another example, the driverless car uses the machine vision technology to automatically plan and control the safe driving of the vehicle by using the camera, laser/millimeter wave/ultrasonic radar, GPS, etc. to sense the road environment information.
According to statistics, the global market for machine vision systems in 2016 was about $4.6 billion, about $5 billion in 2017, and is expected to reach $5.5 billion in 2018, with an annual growth rate of around 10%. The growth of China's machine vision market began in 2010. In 2017, the market size was about 6.8 billion yuan. It is expected to reach 78 billion yuan by 2020, and the market growth rate will exceed 100%.
4Technical bottlenecks and future development
In the development of intelligent image processing technology for machine vision, there are still many technical bottlenecks, such as:
1) Stability: Some processing methods tend to perform well in research and development, but in complex and variable application environments, problems arise from time to time. For example, the face recognition system can recognize the recognition rate up to 95% when the target is matched, but in the actual monitoring environment, the recognition rate will be greatly reduced.
2) Real-time: If the image acquisition speed and processing speed are slow, coupled with the newly introduced deep learning algorithm, the difficulty of real-time processing of the system is increased, and the rhythm of machine operation and control cannot be kept up.
3) Accuracy: The machine vision system requires image recognition and measurement accuracy close to 100%, and any slight error can have unpredictable consequences. For example, the error in target positioning will make the assembled equipment not meet the requirements.
4) System Capability: At present, the embedded image processing system has problems such as insufficient computing power of the chip and limited storage space, and often cannot satisfy image processing operations with large computational complexity, such as neural network iterative operations, large-scale matrix operations, etc. .
The development of intelligent image processing in machine vision in the future is mainly reflected in the following aspects:
1) Algorithm: Traditional algorithms continue to make breakthroughs. A new wave of artificial intelligence brings a lot of new and excellent image processing algorithms, such as deep learning (DL), convolutional neural network (CNN), and generation of confrontation networks ( GAN), and so on.
2) Real-time: There are more hardware platform support with novel structure, sufficient resources and fast computing, such as computers based on multi-CPU, multi-GPU parallel processing structure, mass storage units, etc.
3) Embedded: A new high-speed signal processor array, a very large-scale FPGA chip.
4) Fusion processing: From single image sensor development to multi-sensor (multi-viewpoint) fusion processing, field information can be more fully acquired. It can also integrate multiple types of sensors, such as image sensors, sound sensors, temperature sensors, etc. to complete the on-site target location, identification and measurement.
In short, whether it is "Made in China 2025" or "Industry 4.0" is inseparable from artificial intelligence, computer vision is inseparable, and intelligent image processing is the core technology of machine vision. With the continuous improvement of image processing level, there will be some power. Promote the rapid development of machine vision.
GALOCE Force Washer Transducers are miniature load cells designed specifically for measuring fastener clamping forces. The design provides high stiffness in a small package, making these load cells ideal for static and dynamic measurements on fasteners, or structural test applications where space limitations exist. Force washer transducers come in a variety of English and Metric sizes, by the way, GALOCE OEM&ODM group can customized the Load Cell according the drawing.
Force Washer Load Cell,Force Washer Transducers,Washer Typer Load Cell,Through-hole force sensor
GALOCE (XI'AN) M&C TECHNOLOGY CO., LTD. , https://www.galoce-meas.com