Machine vision plays an important role in the intelligent factory, which can effectively increase the production capacity and improve the product qualification rate. When choosing a small machine vision system, the traditional industrial intelligent camera has the advantages of small size, high integration and easy development and use; the advantage of embedded machine vision system is that the configuration is quite flexible, it can be equipped with high-level CPU processor, support multi-channel camera, and has high scalability. Is there a new type of small machine vision system that can combine the advantages of both, reduce the cost and meet the increasingly demanding and rapid visual application test?
Machine vision plays an important role in the intelligent factory, which can effectively increase the production capacity and improve the product qualification rate. When choosing a small machine vision system, the traditional industrial intelligent camera has the advantages of small size, high integration and easy development and use; the advantage of embedded machine vision system is that the configuration is quite flexible, it can be equipped with high-level CPU processor, support multi-channel camera, and has high scalability. Is there a new type of small machine vision system that can combine the advantages of both, reduce the cost and meet the increasingly demanding and rapid visual application test?
The demand of intelligent factory for machine vision
Performance and processing capacity. How to improve production capacity, system performance and throughput play a key role. As far as general machine vision system is concerned, high resolution and high frame rate (display frames per second) are like fish and bear's paw. In general practical applications, the copyright of Control Engineering China is reserved. It is usually an application combination with high resolution but low frame rate or low resolution but high frame rate. If you want to have both, the only way out is to use high-end CPU processors to make up for the resolution and frame rate multiplication results. How to obtain the best processing performance with reasonable cost is the concern of system developers.
Production line environment. The environment of the factory is usually bad. For example, in the production line of beverage production and packaging, the system may come into direct contact with liquid. But in the environment of tool machining, it is a bad environment full of cutting workpiece. If the machine vision system must be configured in the harsh production line environment nearby, the product with waterproof and dustproof ability can meet the demand.
Multiple production workstations. In the factory environment, all rights are reserved. For a finished product to be put on the market, it is necessary to go through different work stations for the manufacturing of components, the taking and placing of semi-finished products, the quality inspection and the packaging for shipment. For example, CNC machine is responsible for the turning of components. The workpiece is picked up by industrial robot and positioned by industrial camera. After completion, it enters the detection platform for defect detection. The finished product is scanned in the packaging area. How to integrate and communicate easily between multiple production workstations is a key to the intelligent factory.
Software development environment. The difficulty and integration degree of software solution development is a big worry in the minds of all the engineers who import the intelligent system, and it is often the most important factor to determine the success or failure of the project. How to shorten the development time and reduce the system development cost is an important key.
Choose the winning point of small machine vision system
Processor computing performance. Because of its small size and limited space, the traditional smart camera can only carry single core Atom processor or arm processor. Although its power consumption is low, due to its limited performance, it can only complete image analysis and processing of single task, such as counting, scanning barcode, etc. With the release of the Intel atomtm e3840 processor family, the processor performance is improved by two times compared with the previous generation processor family. Control Engineering China Copyright All rights reserved, and it also has the advantage of low power consumption. This means that small size can also bring high performance, and multi task image processing can be realized. The new generation of small machine vision system can simultaneously carry out measurement, counting, positioning, two-dimensional code reading and other multi task processing capabilities. From the perspective of cost of ownership, it has the ability of more than one machine.
Image sensor and image quality. Image sensor is the soul of machine vision system. The size of sensor directly represents the quality of image. In the past, the application of smart camera was defined in the primary image detection. The advantages and disadvantages of sensor size and image quality were not easy to be highlighted. However, if we want to apply machine vision to high-end and high-speed detection applications, then the size of the sensor becomes the key point to be considered when selecting the system.
Comparison of rolling shutter and global shutter. The difference between rolling shutter and global shutter lies in the time difference of exposure. The shutter of rolling shutter tells the photosensitive component to expose in sequence until the whole picture is exposed. The global shutter exposes the whole picture at the same time. With the improvement of system processing performance, the system performance will no longer be the bottleneck. If there is a demand for the detection of high-speed moving objects, the global shutter sensor can collect the correct image without residual image.
Coprocessor. In the process of machine vision image acquisition and analysis, image quality plays an important role. If we can optimize the quality of the collected images before the image analysis, we can ensure the correctness of the image analysis. In the past applications, after the image data was collected to the system, it was necessary to calculate and optimize the image quality through the system processor, because limited by the CPU computing resources, the amount of image data that can be processed will also be limited. However, if we can filter and optimize the image matrix calculation before the CPU calculation, we can greatly speed up the image processing performance and reduce the CPU resources. On the one hand, we can leave the system resources to the core of the machine vision system - image algorithm, on the other hand, we can process large amount of images in real time, so that the high-speed and complex The preprocessing functions such as look-up table, ROI, region of interest, shading correction and other image quality optimization functions are realized.
GPU graphics and multimedia image processing performance. Compared with the previous generation of Intel atomtm d2550 processor family, the new generation of Intel atomtm e3840 processor has a performance improvement of about six times. It can simultaneously process multi-channel image compression transmission through Intel HD graphics 4000 technology. Through the improvement of CPU and GPU performance, image detection results can be recorded and saved, or provide raw data for further comparison and analysis, so that the information system of the factory has more intelligent functions.
System display performance. In the factory environment, all rights are reserved. The traditional intelligent camera can only transmit data through Ethernet for the monitoring of the central control terminal. If the machine vision system can support VGA output interface, the machine vision system can output images through VGA and Ethernet ports at the same time, and connect to the screen at the HMI or production line end to check the results in real time and find problems, which will effectively improve the performance of the production line.
64 bit architecture. Because image analysis software needs to process a large amount of data, most of the mainstream applications in the market have already supported 64 bit. Therefore, the choice of machine vision system, of course, must also choose the system supporting 64 bits, in order to give full play to the application of maximum benefits.
System storage capacity. The storage capacity of small machine vision system represents that users can store more image recognition contrast samples, also can store detection data, or make backup. It is very beneficial to the stability of the whole system.
Total cost of ownership. The total cost of ownership of system purchase is not only the cost of machine vision system itself. Users are smart, how to help customers reduce costs from the perspective of total cost of ownership is the king way.
EN
