1. Trang chủ
  2. » Kỹ Thuật - Công Nghệ

Understanding And Applying Machine Vision Part 2 ppt

25 258 0

Đang tải... (xem toàn văn)

Tài liệu hạn chế xem trước, để xem đầy đủ mời bạn chọn Tải xuống

THÔNG TIN TÀI LIỆU

Thông tin cơ bản

Định dạng
Số trang 25
Dung lượng 641,33 KB

Nội dung

The vast majority of machine vision vendors are players in niche applications in specific manufacturing industries. While generic machine vision platforms have been applied in many industries, no single company has emerged within the machine vision industry as a dominant player with a product(s) that has been applied across a significant number of manufacturing industries. Several companies offer general-purpose vision platforms that have sufficient functionality permitting them to be configured for a variety of applications. Some of these same companies are suppliers of products that address a specific set of applications such as optical character recognition (OCR) and optical character verification (OCV). Some companies are suppliers of image processing board sets that also offer the functionality of a vision platform and can be utilized to address many applications like the general-purpose vision platforms. Page 29 Figure 3.5 Turnkey system from Perceptron performing 3D dimensional analysis on "body-in-white" in automotive assembly plant for critical dimensions, gap andflushness. The vast majority of the suppliers that make up the machine vision industry are suppliers of industry-specific niche application products. There is often as much value associated with peripheral equipment necessary to provide a turnkey solution, as there is value of the machine vision content in the system. It is becoming increasingly more difficult to classify companies in the machine vision market. Suppliers of general- purpose systems are extending their lines to include products that might have earlier been classified as application- specific machine vision systems. Similarly, suppliers of image processing boards are offering boards with software that makes their products appear to be a general-purpose machine vision system. There are a couple of board suppliers that today actually offer turnkey, application-specific machine vision systems. There are several suppliers of application- specific machine vision systems with turnkey systems that address specific applications in different markets (e.g., unpatterned and patterned/print web inspection (Figure 3.6), or 3D systems for semiconductor and electronic applications). Page 30 Figure 3.6 Pharma Vision system from Focus Automation inspecting a roll of pharmaceutical labels on a rewinder for print defects. 3.5— Machine Vision Industry-Related Definitions The following are definitions associated with the different segments of the machine vision industry: Merchant machine vision vendor - A company that either offers a general-purpose, configurable machine vision system or a turnkey application-specific machine vision system. In either case, without the proprietary machine vision functionality, there would be no purchase by a customer. The proprietary machine Page 31 vision hardware could be based either on commercially available image board level products or proprietary vision computer products. Image processing board set suppliers (IPBS) - A company offering one or more products, such as a frame grabber, that often incorporates an A/D, frame storage, and output look-up tables to display memorized or processed images. These boards can operate with either digital or analog cameras. In some cases, they merely condition the image data out of a camera making it compatible with processing by a personal computer. Often these boards will be more "intelligent," incorporating firmware that performs certain image-processing algorithmic primitives at real time rates, and off-loading the computer requirements to the firmware from the computer itself. The interface supplied generally requires a familiarity with image processing and analysis, since one will generally start at the algorithm level to develop an application. IPBS can be sold to GPMV builders, ASMV, builders, merchant system integrators, OEMs, or end-users. General-purpose machine vision system vendor (GPMV) - A company offering products that can be configured or adapted to many different applications. The vision hardware design can be either based on commercially available image board level products or proprietary vision computer products. The graphic user interface is such that little or no reference is made to image processing and analysis. Rather, the interface refers to generic machine vision applications (flaw inspection, gaging, assembly verification, find/locate, OCR, OCV, etc.) and walks the user through an application set-up via menus or icons. These systems may or may not have the ability to get into refining specific algorithms for the more sophisticated user. GPMV systems are sold to application-specific machine vision system builders, merchant system integrators, OEMs, or end-users. A GPMV supplier can use some combination of: Proprietary software Proprietary flame grabber + proprietary software Commercial frame grabber + proprietary software Proprietary IPBS + proprietary software Commercial IPBS + proprietary software Proprietary hardware + proprietary software. Application-specific machine vision vendor (ASMV) - A company supplying a turnkey system that addresses a single specific application that one can find widely throughout industry or within an industry. Interface refers specifically to the application itself, not to generic machine vision applications or imaging functions. In other words, machine vision technology is virtually transparent to the user. Page 32 The vision hardware can be either based on commercially available image board level products, general-purpose machine vision systems, or proprietary vision computer products. ASMV systems are typically sold directly to end- users. An ASMV supplier can use some combination of: Proprietary frame grabber + proprietary software Commercial frame grabber + proprietary software Proprietary IPBS + proprietary software Commercial IPBS + proprietary software Proprietary hardware + proprietary software Commercial GPMV + proprietary software. Machine vision software supplier (MVSW) - A company supplying software that adapts image processing and analysis hardware to generic machine vision applications (flaw inspection, gauging, locate/find, OCR, OCV, etc.). Usually the software is designed to adapt a commercially available image processing board for use in machine vision applications. Alternatively, it may adapt a personal computer to a machine vision application. MVSW can be sold to GPMV builders, ASMV, builders, merchant system integrators, OEMs, or end-users. Web scanner supplier - A company providing a turnkey system to inspect unpatterned products produced in webs (paper, steel, plastic, textile, etc.). These systems can capture image data using area cameras, linear array cameras, or laser scanners. The vision hardware used in the system can be based on commercially available image board level products, general-purpose machine vision systems or proprietary vision computers. Web scanners are typically sold to end-users. 3D-machine vision or laser triangulation supplier - A company providing a system that offers 3D measurements based on the calculation of range using triangulation measurement techniques. The system can use any number of detection schemes (lateral effect photodiode, quadrant photodetector, matrix array camera, linear array camera) to achieve the measurement. The lighting could be a point source, line source, or specific pattern. The simpler versions collect data one point at a time. Some use a flying spot scanner approach to reduce the amount of motion required to make measurements over a large area. Others use camera arrangements to collect both 2D and 3D data simultaneously. Laser triangulation-based machine vision systems can be sold to GPMV builders, ASMV, builders, merchant system integrators, OEMs, or end users. Merchant system integrator - A company providing a machine vision system with integration services and adapting the vision system to a specific customer's requirements. A system integrator is project-oriented. Merchant system integrators typically sell to an end user. A merchant system integrator provides: Page 33 1. Turnkey system based on: Commercial frame grabber + proprietary software or commercial software Commercial IPBS + proprietary software of commercial software Commercial GPMV + proprietary software or commercial software 2. Plus value added: application engineering, GUI, material handling, etc. Captive system integrator - A company purchasing a machine vision product for its own use and employing its own people to provide the integration services. The machine vision product will typically be either a general-purpose machine vision system or an image board set. Original equipment manufacturer (OEM) - A company offering a product with a machine vision value adder as an option. An OEM includes machine vision in its product, but without machine vision, the system would still have functionality for a customer. Absent from this list of supplier types "value adder remarketer (VAR)." This term is so general that it loses its meaning. Virtually every other type of company associated with applying machine vision is essentially a value adder. In other words, a company that manufactures application-specific machine vision systems based on a commercial general-purpose machine vision product or image processing board set is a value adder to those products. An OEM is a company adding a whole lot of value - generally the functionality required by the user of its piece of equipment. A merchant system integrator adds value to either a general-purpose machine vision system or image processing boards — the value being project-specific software and hardware application engineering. The distinctions between an ASMV, OEM, and merchant system integrator are: ASMV - turnkey system provider; functionality purchased includes entire system; any single element of system has no value to customer alone; sells many of the same system OEM - machine vision is an optional value adder to existing functionality Merchant system integrator - project-based business. 3.6— Summary This discussion is meant to clarify the vendor community for the prospective buyer of a machine vision system. It is important to understand that there are different players with different business goals as well as expertise. Successful deployment depends on matching the supplier's product and skill mix to the application. Page 35 4— The ''What" and "Why" of Machine Vision Machine vision, or the application of computer-based image analysis and interpretation, is a technology that has demonstrated it can contribute significantly to improving the productivity and quality of manufacturing operations in virtually every industry. In some industries (semiconductors, electronics, automotives), many products can not be produced without machine vision as an integral technology on production lines. Successful techniques in manufacturing tend to be very specific and often capitalize on clever "tricks" associated with manipulating the manufacturing environment. Nevertheless, many useful applications are possible with existing technology. These include finding flaws (Figure 4. 1), identifying parts (Figure 4.2), gauging (Figure 4.3), determining X, Y, and Z coordinates to locate parts in three-dimensional space for robot guidance (Figure 4.4), and collecting statistical data for process control and record keeping (Figure 4.5) and high speed sorting of rejects (Figure 4.6). Machine vision is a term associated with the merger of one or more sensing techniques and computer technologies. Fundamentally, a sensor (typically a television-type camera) acquires electromagnetic energy (typically in the visible spectrum; i.e., light) from a scene and converts the energy to an image the computer can use. The computer extracts data from the image (often first enhancing or otherwise processing the data), compares the data with previously developed standards, and outputs the results usually in the form of a response. Page 36 It is important to realize in what stage of the innovation cycle machine vision finds itself today. Researchers who study such cycles generally classify the stages as (1) research, (2) early commercialization, (3) niche-specific products, and (4) widespread proliferation. In the research stage, people that are experts in the field add new knowledge to the field. In the early commercialization phase, entrepreneurial researchers develop products that are more like "solutions looking for problems." It requires a good deal of expertise to use these products. The individuals applying stage 2 technology are generally techies who thrive on pioneering. Stage 3 sees the emergence of niche-specific products. Some suggest this is the stage machine vision finds itself in today. Machine vision systems embedded in a piece of production equipment are generally totally transparent to the equipment operator. Application-specific machine vision systems generally have a graphic user interface that an operator can easily identify with as it speaks only in terms with which he is familiar. Nevertheless, while the fact that a machine vision system is being used may be disguised, it still requires an understanding of the application to use it successfully. Figure 4.1 Early version of a paint inspection system that looks for cosmetic defects on auto body immediately after paint spray booth. Page 37 Figure 4.2 Cognex Vision system verifying and sorting foreign tires based on tread pattern identification. Stage 4 is characterized by technology transparency - the user does not know anything about it, other than that it is useful. Most car drivers understand little about how a car operates, other than what it does when you turn the key. Interestingly, when the car was a "stage 2" technology, a driver also had to be able to service it because of frequent breakdowns experienced. Since then an infrastructure of service stations and highways has emerged to support the technology. In stage 2 there were over 1100 car manufacturers in the United States alone! The industry consolidated as it moved from stage 2 to stage 4. Clearly, while some consolidation has taken place in the machine vision industry, there are still hundreds of players. This is an indicator of more of a Stage 3 technology. This means that one should have some level of understanding of the technology to apply it successfully. Machine vision is far from a commodity item. The first step is to become informed - the very purpose of this book. Page 38 Figure 4.3 Early system installed on a steel line by Opcon designed to measure cylindrical property of billet. It is not clear that machine vision as we have defined it will ever become transparently pervasive in our lives or truly a stage 4 technology. The reality is that the underlying technology will definitely become stage 4 technology. The area of biometrics that often uses the same computer vision technology is expected to become a major tool in accessing automated teller machines, cashing checks, accessing computers, etc. There is no doubt there will be other markets in which the underlying technology will become pervasive. For example, if the automobile is to ever achieve autonomous vehicle status, computer vision in some form will make it possible. Page 39 Figure 4.4 Adept vision-guided robot shown placing components on printed circuit board. 4.1— Human Vision versus Machine Vision Significantly, machine vision performance today is not equal to the performance one might expect from an artificially intelligent eye. One "tongue-in-cheek" analysis by Richard Morley and William Taylor of Gould's Industrial Automation Section quoted in several newspaper articles in the mid-1980's suggests that the optic nerve in each eye dissects each picture into about one million spatial data points (picture elements). Retinas act like 1000 layers of image processors. Each Page 40 Figure 4.5 Early RVSI (Automatic) system at end of stamping line examining hole presence and dimensions to monitor punch wear (a) and example of data (b). Page 41 Figure 4.6 Zapata system inspecting bottle caps to verify presence and integrity of liners at rates of 2600 per minute. layer does something to the image (a process step) and passes it on. Since the eye can process about 10 images per second, it processes 10,000 million spatial data points per second per eye. While today there are machine vision systems that operate at several billion operations per second, these still do not have anywhere near the generic vision capacity of humans. Significantly, the specification of MIPS, MOPS, and so on, generally has little relevance to actual system performance. Both hardware and software architectures affect a system's performance, and collectively these dictate the time needed to perform a complete imaging task. Based on our eye-brain capacity, current machine vision systems are primitive. The range of objects that can be handled, the speed of interpretation, and the susceptibility to lighting problems and minor variations in texture and reflectance of objects are examples of limitations with current technology. On the other hand, machine vision has clear advantages when it comes to capacity to keep up with high line speeds (Figure 4.6). Similarly, machine vision systems can conduct multiple tasks or inspection functions in a virtually simultaneous manner on the same object or on different objects (Figure 4.7). With multiple sensor inputs it can even handle these tasks on different lines. Some comparisons that can be made between human and machine vision are as follows: Human vision is a parallel processing activity. We take in all the content of a scene simultaneously. Machine vision is a serial processor. Because of sensor Page 42 Figure 4.7 (a) Early RVSI (Automatix) system with multiple cameras inspects tie rod to verify presence of thread, assembly, completeness and swage angle; (b) with multiple cameras inspects tie rod to verify presence of thread, assembly, completeness, and swage angle; (c) with multiple cameras to inspect tie rods to verify presence of thread, assembly, completeness, and swage angle; and (d) with multiple cameras to inspect tie rods to verify presence of thread, assembly, completeness, and swage angle. [...]... RVSI/Itran) The challenge in machine vision is the computational power required to handle the amount of image data generated: 25 6 × 25 6 × 30 ~ 2 MHz 5 12 × 5 12 × 30 ~ 8 MHz These are 8-bit bytes if processing 25 6 shades of gray images Data arrives at the rate of one pixel in every 100 or so nanoseconds This is why in the many machine vision systems, resolution is only nominally 5 12 × 5 12, and each picture element... Human vision is based on the interaction of light reflected from an image In machine vision any number of illumination methods are possible, and the specific one used is a function of the application Figure 4.8 Light spectrum Page 45 Figure 4.9 Rendering of eye (courtesy of RVSI/Itran) Tables 4.1 and 4 .2 summarize the comparison between machine vision and human vision A key difference is that machine vision. .. meteorological and earth resources data are examples Machine vision has been defined by the Machine Vision Association of the Society of Manufacturing Engineers and the Automated Imaging Association as the use of devices for optical, noncontact sensing to automatically receive and interpret an image of a real scene in order to obtain information and/ or control machines or process Significantly, machine vision. .. control, machine control, and robot control Page 47 Table 4 .2 Machine Vision versus Human Vision: Evaluation of Performance PERFORMANCE CRITERIA MACHINE VISION HUMAN VISION Resolution Limited by pixel array size High resolution capability Processing speed Fraction of second per image Real-time processing Discrimination Limited to high-contrast images Very sensitive discrimination Accuracy Accurate for part. .. unstructured scene 4 .2 Machine Vision Definition What do we mean by machine vision? Distinctions are made between image analysis, image processing, and machine vision Image analysis generally refers to equipment that makes quantitative assessments on patterns associated with biological and metallurgical phenomena Image processing refers generally to equipment designed to process and enhance images for... 56 Table 4.5 Machine Vision Applications: Inspection Highly quantitative mensuration, critical dimensions: critical exterior and interior dimensions of key features of workpieces A Qualitative-semiquantitative mensuration 1 Label reading and registration 2 Sorting 3 Integrity and completeness a All parts and features present; right parts b Burrs; cracks; warping; defects, approximate size and location... nerve and are passed onto the occipital nerve, where cognitive processing of the image starts Generally speaking, early on we establish models of our surroundings and interpret what we observe based on a priori known relationships stemming from learned models Machine vision has a long way to go Page 46 Table 4.1 Machine Vision versus Human Vision: Evaluation of Capabilities CAPABILITIES MACHINE VISION. .. addition, this property can be influenced by the medium between object and illumination and object and sensor, by filters between object and illumination and object and sensor, by optical properties such as vignetting and dirt on the optics, and by sensor pixel sensitivity variations Figure 4.13 reflects the digital representation of a scene, and Figure 4.14 depicts the digitally encoded values of the gray... machine vision system can infer dimensional data and the distance of an object An alternate approach to obtain three-dimensional detail has been to employ two cameras and use stereo correspondence analysis based on triangulation principles 4.3— Machine Vision Applications Table 4.4 depicts the type of information that can be extracted and analyzed from an image of an object: spectral, spatial, and temporal... on and the type of analysis that a machine vision system must perform is a function of application, which includes task, object, and related application issues (material handling, staging, environment, etc.) The task refers to: Inspection Gauging Cosmetic (flaw detection) Verification Recognition Identification Location analysis Position Guidance Tables 4.5 and 4.6 depict taxonomies of generic machine . RVSI/Itran). Tables 4.1 and 4 .2 summarize the comparison between machine vision and human vision. A key difference is that machine vision can be quantitative while human vision is qualitative and subjective. The. quality control, machine control, and robot control. Page 47 Table 4 .2 Machine Vision versus Human Vision: Evaluation of Performance PERFORMANCE CRITERIA MACHINE VISION HUMAN VISION Resolution. challenge in machine vision is the computational power required to handle the amount of image data generated: 25 6 × 25 6 × 30 ~ 2 MHz 5 12 × 5 12 × 30 ~ 8 MHz These are 8-bit bytes if processing 25 6 shades

Ngày đăng: 10/08/2014, 02:21

TỪ KHÓA LIÊN QUAN