In modern industrial automation, various inspection, production monitoring, and part identification applications are involved, such as dimensional inspection of batch processing of parts, integrity inspection of automated assembly, automatic component positioning in electronic assembly lines, and character recognition on ICs. The human eye typically cannot continuously and stably perform these highly repetitive and intelligent tasks, and other physical quantity sensors also have limited applications.
This led to the consideration of using photoelectric imaging systems to acquire images of controlled targets, which were then digitized by computers or dedicated image processing modules. Based on information such as pixel distribution, brightness, and color, the size, shape, and color of the images were determined. In this way, the speed and repeatability of computers were combined with the high intelligence and abstraction capabilities of human vision, thus giving rise to the concept of machine vision.
A successful machine vision system is a system meticulously engineered to meet a set of clearly defined requirements. Once these requirements are fully determined, the system is designed and built to satisfy these precise requirements.
The advantages of machine vision include the following:
■High precision
As a precise measuring instrument, a well-designed vision system can perform spatial measurements on one of a thousand or more parts. Because this measurement is non-contact, it eliminates wear and tear and hazard to fragile components.
■ Continuity
Vision systems can save people from fatigue. Because there is no human operator, there are no human-induced changes in operation. Multiple systems can be set to operate independently.
■ High cost efficiency
With the dramatic drop in computer processor prices, machine vision systems have become increasingly cost-efficient. A $10,000 vision system can easily replace three human operators, each requiring a $20,000 annual salary. Furthermore, the operating and maintenance costs of vision systems are very low.
■ Flexibility
Vision systems can perform a wide variety of measurements. When applications change, only the software needs to be modified or upgraded to adapt to the new requirements.
Many companies using Satisfactory Process Control (SPC) are considering using machine vision systems to deliver continuous, coordinated, and precise measurement SPC commands. In SPC, manufacturing parameters are continuously monitored. Controlling the entire process involves ensuring these parameters remain within certain ranges. This allows manufacturers to adjust process parameters if production becomes uncontrollable or defective parts occur.
Machine vision systems offer greater adaptability than optical or machine sensors. They enable automated machines to be diverse, flexible, and reconfigurable. When changes to the production process are needed, "tool change" for machine vision involves only a software modification, not the replacement of expensive hardware. Furthermore, vision systems can often be reused after production line reconfiguration.
Composition of machine vision system
Machine vision technology uses computers to analyze an image and draw conclusions based on the analysis. Currently, machine vision has two main applications. Machine vision systems can detect parts, where optics allow processors to observe targets more precisely and make effective decisions about which parts can be used and which need to be discarded; machine vision can also be used to create parts, that is, to directly guide the manufacturing process using a combination of complex optics and software.
Although machine vision applications vary, they all include the following processes;
■Image Acquisition
The optical system acquires images, converts the images into analog format, and transfers them to the computer's memory.
■Image Processing
The processor uses different algorithms to improve image elements that have a significant impact on the conclusion.
■Feature Extraction
The processor identifies and quantifies key features of the image, such as the location of holes on a printed circuit board or the number of pins on a connector. This data is then transmitted to the control program.
■Judgments and Control
The processor's control program draws conclusions based on the received data. This data may include whether holes on a printed circuit board are within required specifications or how an automated machine must move to pick up a component.
Machine Vision System Analysis
A typical vision system generally includes: a light source, an optical system, a camera, an image processing unit (or image acquisition card), image analysis and processing software, a monitor, and a communication/input/output unit.
Image Acquisition
Image acquisition is essentially the process of converting the visual image and intrinsic features of the object under test into data that can be processed by a computer, which directly affects the stability and reliability of the system. Generally, images of the object under test are acquired using light sources, optical systems, cameras, and image processing units (or image capture cards).
■Light Source
Light source is a crucial factor affecting the input of machine vision systems, as it directly impacts the quality of the input data and accounts for at least 30% of the application's effectiveness. Since there is no universal machine vision lighting equipment, appropriate lighting must be selected for each specific application to achieve optimal results. Many industrial machine vision systems use visible light as their light source, primarily because it is readily available, inexpensive, and easy to operate. Commonly used visible light sources include incandescent lamps, fluorescent lamps, mercury lamps, and sodium lamps.
However, a major drawback of these light sources is their instability. For example, fluorescent lamps experience a 15% drop in light energy within the first 100 hours of use, and this drop continues over time. Therefore, maintaining stable light energy to a certain extent is a pressing issue that needs to be addressed in practical applications. Furthermore, ambient light alters the total light energy reaching the object, introducing noise into the output image data. This is typically addressed by using a protective shield to reduce the impact of ambient light. Due to these problems, in modern industrial applications, for certain demanding inspection tasks, invisible light sources such as X-rays and ultrasound are often used as the light source.
Illumination systems composed of light sources can be classified according to their illumination methods, including: backlighting, front lighting, structured light, and stroboscopic lighting. Backlighting places the object under test between the light source and the camera, offering the advantage of obtaining high-contrast images. Front lighting places the light source and camera on the same side of the object under test, facilitating installation. Structured light illumination projects gratings or line light sources onto the object under test, demodulating the object's three-dimensional information based on the resulting distortions. Stroboscopic lighting illuminates the object with high-frequency light pulses, requiring the camera's scanning speed to be synchronized with the light source's stroboscopic speed.
■Optical System
For machine vision systems, images are the sole source of information, and image quality is determined by the proper selection of the optical system. Errors caused by poor image quality often cannot be corrected by software. Machine vision technology combines optical components and imaging electronics, using a computer-controlled system to distinguish, measure, classify, and detect parts passing through automated processing systems. Machine vision systems can typically detect up to 100% of the products being processed without slowing down the production line. This capability is crucial as more and more manufacturers require "6-sigma" (less than three parts per million) results to be more competitive in today's quality-conscious market. Furthermore, these systems integrate ideally with Satisfactory Process Control (SPC).
The main parameters of an optical system are related to the format of the photosensitive surface of the image sensor, and generally include: aperture, field of view, focal length, F-number, etc.
■Camera
A camera is essentially a photoelectric conversion device, which converts the optical image received by the image sensor into electrical signals that a computer can process. Photoelectric conversion devices are the core components of a camera. Currently, typical photoelectric conversion devices include vacuum tubes, CCDs, and CMOS image sensors.
A vacuum television camera tube consists of two parts: a camera target and an electron gun, both sealed within a glass tube. The camera target converts the illuminance distribution of the input optical image into a two-dimensional spatial distribution of the corresponding pixel charge on the target surface, primarily performing photoelectric conversion and charge storage. The electron gun, on the other hand, performs the scanning and acquisition of the image signal. Television camera tube imaging systems offer advantages such as high definition, high sensitivity, wide spectral density, and high frame rate imaging. However, because television camera tubes are vacuum tube devices, they are relatively heavy, bulky, and consume significant power.
CCDs are currently the most commonly used image sensors in machine vision. They integrate photoelectric conversion, charge storage, charge transfer, and signal readout, making them typical solid-state imaging devices. A key characteristic of CCDs is that they use electric charge as a signal, unlike other devices that use current or voltage. These imaging devices form charge packets through photoelectric conversion, which are then transferred, amplified, and output as image signals under the action of driving pulses. A typical CCD camera consists of an optical lens, a timing and synchronization signal generator, a vertical driver, and analog/digital signal processing circuitry.
As a functional device, CCD has advantages over vacuum tubes, such as no burn-in, no hysteresis, low voltage operation, and low power consumption.
The development of CMOS (Complementary Metal Oxide Semiconductor) image sensors first appeared in the early 1970s. In the early 1990s, with the development of Very Large Scale Integration (VLSI) manufacturing technology, CMOS image sensors experienced rapid development. A CMOS image sensor integrates a photosensitive element array, image signal amplifier, signal readout circuit, analog-to-digital converter, image signal processor, and controller onto a single chip, and also has the advantage of programmable random access to local pixels. Currently, CMOS image sensors are widely used due to their excellent integration, low power consumption, wide dynamic range, and virtually ghosting-free output images.
Image processing and analysis
In machine vision systems, the primary function of a camera is to convert the received light signal into a voltage amplitude signal for output. To obtain a digital signal that can be processed and recognized by a computer, the video information needs to be quantized. Image acquisition cards are crucial tools for this quantization process.
■Image Acquisition/Processing Card
Image acquisition cards primarily handle the digitization of analog video signals. The video signal is first filtered by a low-pass filter, converting it into a continuous analog signal. According to the image resolution requirements of the application system, a sample-and-hold circuit is used to sample the video signal at intervals in time, converting it into a discrete analog signal. Then, an A/D converter converts it into a digital signal for output. In addition to analog-to-digital conversion, image acquisition/processing cards also provide video image analysis and processing capabilities, and can effectively control the camera.
■Image processing software
In machine vision systems, visual information processing techniques primarily rely on image processing methods, including image enhancement, data encoding and transmission, smoothing, edge sharpening, segmentation, feature extraction, and image recognition and understanding. After these processes, the quality of the output image is significantly improved, enhancing both its visual appeal and facilitating computer analysis, processing, and recognition.
Applications of machine vision systems
Machine vision systems are an effective way to achieve precise control, intelligence, and automation of instruments and equipment, and can be considered the "machine eyes" of modern industrial production. Their greatest advantage is:
(1) Achieves non-contact measurement. No damage is caused to either the observer or the observed, thus improving the reliability of the system;
(2) It has a wide spectral response range. Machine vision can use dedicated photosensitive elements to observe the world that humans cannot see, thus expanding the range of human vision.
(3) Long-term operation. Humans find it difficult to observe the same object for extended periods. Machine vision systems, on the other hand, can perform observation, analysis, and recognition tasks for long periods and can be applied in harsh working environments.
For more information, please follow the Machine Vision channel.
