1. Introduction Modern industrial automation involves various applications of inspection, production monitoring, and part identification, such as dimensional inspection in batch processing of automotive parts and integrity checks in automated assembly, automatic component positioning in electronic assembly lines, and character recognition on ICs. These highly repetitive and intelligent tasks are typically performed by the human eye. However, in certain special cases, such as precise and rapid measurement of minute dimensions, shape matching, and color recognition, the human eye simply cannot perform these tasks continuously and stably, and other physical quantity sensors are also inadequate. Researchers began to consider using CCD cameras to capture images and then feeding them into computers or dedicated image processing modules. Through digital processing, the size, shape, and color are determined based on pixel distribution, brightness, color, and other information. This method combines the speed and repeatability of computer processing with the high intelligence and abstraction capabilities of human vision, thus giving rise to the concept of machine vision testing technology. Visual testing technology is an emerging testing technology built upon research in computer vision. Unlike computer vision research, which focuses on visual pattern recognition and visual understanding, visual testing technology emphasizes the measurement of an object's geometric dimensions and position. Examples include measuring the three-dimensional dimensions of a car body, rapidly measuring the three-dimensional surface shapes of molds, and measuring the coaxiality and coplanarity of large workpieces. It can be widely applied to active, real-time measurement processes such as online measurement and reverse engineering. Visual testing technology has developed rapidly abroad. As early as the 1980s, the U.S. National Bureau of Standards predicted that 90% of future inspection tasks would be completed by visual testing systems. Therefore, in the 1980s alone, more than 100 companies in the U.S. entered the visual testing system market, demonstrating its promising future. Recent Beijing International Machine Tool Exhibitions have showcased advanced instruments developed using visual inspection technology, such as mobile optical coordinate measuring machines (CMMs), high-speed, high-precision digital scanning systems, and non-contact optical CMMs. 2. Composition, Classification and Working Principle of Machine Vision Testing System 2.1 System Composition and Working Principle (1) System Composition A typical vision system generally includes a light source, lens, CCD camera, image processing unit (or image acquisition card), image processing software, monitor, communication/input/output unit, etc. (2) Working Principle The output of the vision system is not an image or video signal, but the detection result (such as size data) after calculation and processing. Usually, machine vision testing is to use machines to replace the human eye to make measurements and judgments. First, the CCD camera is used to convert the captured target into an image signal, which is transmitted to a dedicated image processing system. According to the pixel distribution and information such as brightness and color, it is converted into a digital signal. The image system performs various calculations on these signals to extract the features of the target, such as: area, length, quantity, position, etc. Finally, the results are output according to the preset tolerance and other conditions, such as: size, angle, offset, number, qualified/unqualified, present/absent, etc. After the host computer (such as PC and PLC) obtains the detection results in real time, it commands the motion system or I/O system to execute the corresponding control actions (such as positioning and classification). 2.2 System Classification Based on the operating environment, vision systems can be classified into PC-based systems and PLC-based systems. PC-based systems leverage their openness, high programming flexibility, and user-friendly Windows interface, while also having a lower overall cost. PC-based systems include high-performance image acquisition cards, typically supporting multiple lenses and providing library function support. Currently, the world's leading PC-based vision system manufacturer, Data Translation Corporation (USA), has established its MACH series (such as the DT3155) and MV series PCI industrial vision cards as industry standards. Regarding supporting software, its 32-bit SDK for Windows 95/98/NT provides DLLs for C/C++ programming, the DT Active Open Layer visual control provides a graphical programming environment for VB and VC++, and DT Vision Foundry is an object-oriented machine vision configuration software for Windows, allowing users to quickly develop complex and advanced applications. Similarly, National Instruments (NI) in the US combines machine vision and motion control functions with its widely used LabVIEW virtual instrument software, achieving significant results. Compared to the US companies' strong focus on PC architecture, Japanese and German companies are at the forefront of PLC-based systems. In PLC systems, vision systems function more like intelligent sensors. The image processing unit operates independently of the system, exchanging data with the PLC via a serial bus and I/O. Panasonic's Image Checker M100/M200 system is a prime example. This system utilizes a high-speed dedicated ASIC for 256 levels of grayscale detection, incorporating logic conditions and mathematical operations. The system software is embedded in the image processor, and configuration of the menu displayed on the monitor is achieved through a simple device similar to a game keyboard. This results in a short development cycle and high system reliability. Their next-generation products, the A110/A210, embody integration, miniaturization, high speed, and low cost. Companies like Omron and Keyence also offer similar systems, but their technical performance is relatively simpler, making them more suitable for presence/absence detection or shape matching. The Siemens SIMATIC VS 710 intelligent PROFIBUS industrial vision system from Germany provides an integrated, distributed, high-end image processing solution. It integrates the processor, CCD, and I/O into a single chassis, offering PROFIBUS networking or integrated I/O and RS232 interfaces. More importantly, it can be configured via Pro Vision software under PC/Windows. The VS 710 is the first system to combine the flexibility of a PC, the reliability of a PLC, distributed network technology, and integrated design, allowing Siemens to achieve a perfect balance between PC and PLC architectures. 3. Typical Application Areas and Market Status of Machine Vision Testing Systems The development of modern vision theory and technology not only simulates the functions that the human eye can perform, but more importantly, it enables the completion of tasks that the human eye cannot. Based on the continuous maturation and improvement of technologies such as electronics, optics, and computers, vision technology, as an emerging technology category, has also developed rapidly. Machine vision is characterized by automation, objectivity, non-contact operation, and high precision. Compared with general image processing systems, machine vision systems emphasize accuracy, speed, and reliability in industrial environments. Machine vision is particularly suitable for quality inspection in mass production processes, such as: part assembly integrity, assembly dimensional accuracy, part machining accuracy, position/angle measurement, part identification, and characteristic/character recognition. It is mainly used in fields including automotive, pharmaceutical, electronics and electrical, manufacturing, packaging, food, beverage, and medical. Applications include inspecting the machining accuracy of automotive dashboards, rapid positioning of electronic components on high-speed placement machines, checking the number of pins, identifying characters printed on IC surfaces, inspecting capsule wall thickness and appearance defects in capsule production, checking the number and breakage of ball bearings in bearing production, identifying production dates on food packaging, checking label placement, and determining cell count and properties in the medical field. Because machine vision systems can quickly acquire large amounts of information, are easy to automate, and are easily integrated with design and processing control information, they are widely used in modern automated production processes for condition monitoring, finished product inspection, and quality control. Machine vision systems are characterized by improved production flexibility and automation. They can replace human vision in hazardous work environments unsuitable for manual labor or in situations where human vision is insufficient. Furthermore, in large-scale industrial production, manual inspection of product quality is not only inefficient but also lacks precision, while machine vision inspection methods can significantly improve production efficiency and automation. In addition, machine vision facilitates information integration, making it a fundamental technology for computer-integrated manufacturing. Internationally, the application of vision systems is booming, with the market size reaching $4.6 billion in 1998 alone. In China, however, industrial vision systems are still in the conceptual introductory stage. Leading companies in various industries have only begun to focus on visual measurement automation after solving the problems of production automation. 4. Application of Machine Vision Testing System in Inspection Machine vision systems are widely used in various fields of industrial online inspection. (1) Measurement of parallelism and perpendicularity of large workpieces: The parallelism and perpendicularity measuring instrument for large workpieces using laser scanning and CCD detection system uses a stable collimated laser beam as the measurement baseline, equipped with a rotating axis system, and a rotating pentagram prism to scan out mutually parallel or perpendicular reference planes, and compares them with the surfaces of the large workpiece being measured. When processing or installing large workpieces, this error detector can be used to measure the parallelism and perpendicularity between surfaces. (2) Online dynamic detection system for geometric parameters of hot-rolled threaded steel: This system uses strobe light as the illumination source and uses area array and line array CCD as detection devices for the outer contour dimensions of threaded steel to realize the dynamic detection of online measurement of geometric parameters of hot-rolled threaded steel. (3) Real-time monitoring of bearing status: The bearing load and temperature changes are monitored in real time using vision technology to eliminate the danger of overload and overheating. This technology transforms the traditional passive measurement of ensuring processing quality and safe operation by measuring the surface of the ball into active monitoring. (4) Intelligent Integrated Testing System for Instrument Panel Assembly Based on Machine Vision: The instrument panel assembly of an automobile is equipped with speedometer, water temperature gauge, fuel gauge, ammeter, signal alarm lights, etc. Its production volume is large, and a final quality inspection is required before leaving the factory. The inspection items include the indication error of the five instrument pointers such as the speedometer, whether the 24 signal alarm lights and several lighting lights are damaged or missing, etc. Usually, manual visual inspection is used, but the error is large and the reliability is poor, which cannot meet the needs of automated production. The intelligent integrated testing system based on machine vision testing technology changes this situation, realizes intelligent, fully automatic, high-precision and fast quality inspection of the instrument panel assembly, overcomes the various errors caused by manual inspection, and greatly improves the efficiency and reliability of inspection. (5) Automatic Flaw Detection System for Metal Plate Surface: When inspecting special large metal plates with high surface quality requirements, the original inspection method is to use manual visual inspection or use a dial indicator with a probe. This method is not only easily affected by subjective factors, but may also bring new scratches to the surface being tested. The automatic flaw detection system for metal plates utilizes machine vision testing technology to automatically inspect metal surface defects. It enables high-speed and accurate detection during production, and its non-contact measurement method avoids the possibility of creating new scratches. The system uses a laser as the light source, filtering out stray light around the laser beam with a pinhole filter. A beam expander and collimator are used to parallelize the laser beam, uniformly illuminating the surface of the metal plate at a 45-degree incident angle. The metal plate is placed on an inspection table that can move in the x, y, and z directions. The camera uses a TCD142D 2048-line CCD with a standard camera lens. The CCD interface circuit uses a microcontroller system. The PC host primarily performs image preprocessing and defect classification or scratch depth calculation, and displays the detected defect or scratch images on the monitor. The CCD interface circuit and the PC communicate bidirectionally via an RS.232 port, forming an interactive human-machine data acquisition and processing system. This system mainly utilizes the self-scanning characteristics of linear CCD combined with the movement of the inspected steel plate in the x-direction to extract three-dimensional image information of the metal plate surface. (6) Car body contour dimension accuracy detection system The 100% online detection of the contour dimension accuracy of the 800 series cars of the British ROVER car company is a typical example of machine vision system used in industrial inspection. The system consists of 62 measurement units, each of which includes a laser and a CCD camera to detect 288 measurement points on the car body shell; the car body is placed under the measurement frame and the precise position of the car body is calibrated by software. Each laser and camera unit is calibrated offline, and there is also a calibration device calibrated offline by a coordinate measuring machine to calibrate the camera online; the detection system detects one car body every 40 seconds and can detect three types of car bodies; the system compares the detection results with the qualified dimensions extracted from the CAD model, and the measurement accuracy is ±0.1mm. ROVER's quality inspection personnel use this system to determine the dimensional consistency of key parts, such as the overall shape of the car body, doors, and glass windows. Testing practice has proven that the system can successfully perform online testing of the dimensional accuracy of the body contour of the 800 series automobiles, and will be used to test the dimensional accuracy of the body contour of other ROVER series. (7) Audi Body-in-White Surface Quality Inspection System Audi has recently developed a system capable of fully automatic detection of surface defects in body-in-white, named "Intelligent Control Body-in-White Surface Quality Inspection System". This inspection system integrates direct phase acquisition of projection grating, high-speed digital image processing, automatic identification of surface defect image patterns, intelligent quality judgment, adaptive system learning technology, high-speed digital information network, loose self-adjusting hardware and software structure, and robot system control technology. It can perform 100% online inspection of the welded body-in-white on a production line with a transmission speed of 5m/min. The vehicle inspection time is 1 minute and 20 seconds. Through automatic testing and analysis, surface defects that were previously indistinguishable by the naked eye are directly marked on the body, so that the defects can be polished before the body-in-white enters the painting process, saving the polishing process in the surface painting process, saving a lot of manufacturing costs, and improving the surface quality of the body. Furthermore, machine vision systems can effectively address situations where other methods are difficult to use. The application of machine vision is increasingly replacing humans in many tasks, which undoubtedly greatly improves the level of production automation and the intelligence level of inspection systems. 5. Integration of Machine Vision Systems with CMM With the intensification of international market competition, manufacturing enterprises in various countries are increasingly aware that product quality is the key to the success or failure of their production and operations. With the diversification of the market environment, enterprises have placed higher demands on the collection, processing, and transmission of massive amounts of quality-related data. More flexible and automated CAQ systems are showing the following development trends: ① When necessary, CAQ systems can inspect 100% of products, unlike the currently prevalent sampling inspection; ② Integrating inspection planning into the processing process to form a closed-loop feedback control system, determining the deviation of products from standard dimensions during inspection and correcting it online, thus obtaining nearly 100% high-quality products; ③ Machine vision, advanced image processing technology, and reverse engineering technology have been widely applied to automated inspection, thus achieving intelligent, flexible, fast, and low-cost inspection goals. ④ Inspection technologies applicable to different product structures can directly transfer new product technical requirements from the CAD/CAM database to the inspection system, eliminating the need for operators to write special programs. The development of technologies such as machine vision and reverse engineering, and their integration with CMM, can further improve the measurement efficiency of CMM. For measurement objects with original CAD models, machine vision systems can be used to quickly identify the object's shape and its position and status on the measurement platform, completing the transformation between the machine coordinate system, workpiece coordinate system, and camera coordinate system, helping CMM to automatically form inspection paths and judge measurement results. The machine vision system transmits the collected information to the computer, which simultaneously controls the operation of the vision system. On the other hand, the computer transmits the generated inspection plan to the CMM controller, which controls the CMM measurement and then feeds the measurement results back to the main control computer, forming a closed-loop feedback inspection system. To generate inspection rules, information from the CAD/CAM database is used to match the image data obtained from machine vision with CAD data, automatically determining the workpiece position and selecting inspection items, inspection points, and inspection paths. The method for determining measurement points is as follows: to minimize measurement errors, measurement points are pre-assigned at equal intervals for each object being measured. Finally, the measurement instructions generated by the CMM are transmitted to the CMM controller to begin measurement. For objects without an original CAD model, reverse engineering can be used, i.e., by processing the 3D coordinates of the measurement points acquired by the machine vision system, the CAD model of the object can be reconstructed. 6. Conclusion Machine vision testing systems can significantly reduce inspection costs, improve product quality, accelerate production speed, and increase production efficiency. As a high-precision, non-contact measurement solution, vision systems involve optical and image processing algorithms and are highly specialized products. In the entire measurement and control system, they often need to cooperate with motion control systems to complete position and feed control. Furthermore, when performing synchronous and continuous inspection of multiple processes on a production line, the vision system must have distributed networking capabilities. The integration of machine vision with advanced technologies such as motion control and network communication is changing the face of industrial automation production. With the maturation and development of machine vision technology, it is expected to be increasingly widely used in modern and future manufacturing enterprises. Editor: He Shiping