Machine vision-Suzhou Xispek Inspection Technology Co., Ltd.

Introduction to machine vision system
Machine vision, as its name implies, is to make the machine have the same visual function as human, so as to realize various functions such as inspection, judgment, recognition, measurement and positioning. Machine vision can significantly improve the production efficiency, automation and intelligence. A typical machine vision system includes: light source, lens, camera, image processing hardware, image processing software, execution unit, etc.
Machine vision integrates the technology of optics, machinery, electronics and computer software and hardware, involving image processing, pattern recognition, artificial intelligence, optical and mechanical integration and other fields. In recent years, the rapid development of image processing and artificial intelligence technology greatly promotes the development of machine vision.

Working principle of machine vision system
Machine vision system converts optical signal into image signal through image acquisition hardware (camera, lens, light source, etc.) and transmits it to image processing software. According to the pixel brightness, color distribution and other information, the image processing software extracts the features of the target, and makes the corresponding judgment. According to the results, it controls the field equipment and realizes the inspection function.
The visual processing system consists of hardware and software. According to the different hardware, machine vision system is divided into intelligent camera and PC based vision system. The core part of a PC based machine vision system consists of four parts: light source, lens, camera and visual processing system.
The main function of the lens is to image the measured target onto the sensitive chip of the camera.
The main function of the camera is to collect images, convert optical signals into electrical signals, and then output images to the computer.
The core technology of the software is image processing and analysis algorithm, which includes image enhancement, image segmentation, feature extraction, image recognition and analysis. Through image processing and analysis, the product quality is judged and the size is measured, and the result signal is transmitted to the corresponding hardware for display or execution.
Comparison between machine vision and artificial vision
In the past large-scale industrial production process, the measurement, identification and analysis of products mainly rely on artificial vision. Due to the low efficiency, poor stability and low precision of artificial vision processing, machine vision can greatly improve the processing efficiency and automation degree; at the same time, in some dangerous working environment which is not suitable for manual operation or the human eye is difficult to meet the requirements, machine vision is also commonly used to replace human eyes, such as nuclear power plant monitoring, wafer defect inspection; Moreover, machine vision is easy to realize information integration, which is one of the basic technologies of computer integrated manufacturing. Because machine vision system can obtain a lot of information quickly, and it is easy to process and integrate information automatically, machine vision system is widely used in assembly positioning, product quality inspection, product identification, size measurement and so on. The main difference between machine vision and artificial vision is shown in the right table.

	Machine vision	Artificial vision
Efficiency	Efficient	Low efficiency
Wpeed	Fast	Slow
Reliability	The inspection effect is stable	Easy to fatigue, emotional, not easy to maintain the inspection effect
Working hours	Can work 24 hours a day	Easy to fatigue, limited working time
Information integration	Information integration can be realized	It is not easy to realize information integration
Cost	Cost (one time input)	High cost
Environmental Science	Suitable for hazardous inspection environment	Not suitable for hazardous inspection environment

The development of machine vision
The development of machine vision at home and abroad
Foreign machine vision technology started earlier, the concept of machine vision appeared in the 1950s, began to develop in the 1970s, and entered the high-speed development period of the industry in the late 1990s. There are two great leaps in the development of machine vision. One is the appearance of CCD image sensor in 1970s, which makes optical image into digital signal that can be processed by computer, which is an important turning point in the development of machine vision; the other is the rapid progress of image processing technology such as CPU and DSP in 1980s, which provides favorable conditions for the rapid development of machine vision. From the current situation of global machine vision industry, China, Germany, the United States, Japan and other industrial powers occupy the vast majority of the market of machine vision technology and application. In foreign countries, machine vision is widely used in semiconductor, electronic information, automobile, food, medical and other industries. In the 21st century, although the growth of foreign machine vision market is slowing down, it is still in a leading position in technology.
Domestic machine vision started in the 1980s, and entered the early stage of development at the end of the 20th century and the beginning of this century, and has been developing at a high speed since 2010. With the continuous improvement of industrial automation and more stringent requirements for quality, it is inevitable for machine vision to replace manual inspection. In addition, China's early industrial equipment automation is generally low, so a large number of updates are needed, which constitute a large market demand for machine vision. With the gradual progress of machine vision technology, domestic technology companies continue to develop and launch corresponding products. From cameras, lenses, light sources to image processing software, a number of mature research and development companies have emerged in China. China has become one of the most potential and active regions of machine vision development in the world due to the combined effect of multiple factors, such as the decline of demographic dividend in manufacturing industry, the stimulation of favorable policies for intelligent manufacturing, and the urgent need to improve factory automation.
Development trend of machine vision
With the complexity and diversity of machine vision application scenarios, it is more and more closely integrated with deep learning algorithm, 3D application technology, interconnection standards and other technologies.
Deep learning algorithm: deep learning algorithm simulates the hierarchical structure similar to human brain, and establishes the mapping from low-level signal to high-level semantics through deep neural network to realize hierarchical feature expression of data. Deep learning algorithm is introduced into the machine vision image processing system for appearance inspection, which makes the recognition process more intelligent and the visual information processing ability more powerful.
3D application technology: with the deepening of 3D application technology, more and more 3D reconstruction technologies are introduced into machine vision, such as structured light, DFF, TOF, stereo vision, photometric stereo method, etc. The algorithm of 3D image processing and analysis has been studied more and more widely, which will become a mainstream development direction of machine vision.
Interconnection standard: it is necessary to carry out interconnection within the machine vision system, between the machine vision system and the intelligent manufacturing equipment, and between the enterprise management system, so as to make the equipment and manufacturing management develop towards a more intelligent direction. At present, genlcam standards have been jointly developed within the visual industry, including EMVA, AIA, cmva, JIIA, etc. AIA has developed GigE vision, usb3 vision and other camera communication protocols. The vision industry also cooperates with other industry associations to continuously expand the extension of interconnection, aiming to promote the interconnection between vision system and other industries.
Typical applications of machine vision
In the process of modern automatic production, machine vision system is widely used in electronics, SMT, semiconductor, medicine / medical, tobacco, printing, food / beverage, automobile, lithium battery, photovoltaic and other industries. Application projects include assembly positioning, product quality inspection, product identification, product size measurement and so on.
Visual application in electronic industry
The rapid development of electronic industry has brought great opportunities and challenges to machine vision. Machine vision technology has continuously penetrated into all aspects of the electronic industry chain, from the design and manufacturing of electronic products to product quality inspection, re inspection, packaging, etc., it has injected new strength into the development of the electronic industry.
The inspection contents include:
Capacitance and inductance appearance inspection; LCD AA area positioning, mobile phone card slot positioning, mobile phone appearance defect inspection, mobile phone case logo inspection, etc.

Visual application in SMT industry
SMT (surface mounted technology) is the most commonly used technology and process in the electronic assembly industry. Machine vision is mainly used for positioning and quality inspection of SMT production line, including alignment of steel mesh and PCB in printing machine, 3D scanning of solder paste, component positioning of Mounter, AOI after printing, AOI after mounting, AOI after furnace, etc.
The inspection contents include:
It includes dispensing inspection, component positive and negative pole judgment, component assembly positioning, PCB solder recheck (false solder, more tin, less tin, etc.), OCR (optical character recognition), surface two-dimensional code recognition, etc.

Visual applications in semiconductor industry
Machine vision has been widely used in semiconductor industry, which involves the inspection of semiconductor appearance defects, size, quantity, flatness, spacing, positioning, solder joint quality and bending.
The inspection contents include:
Metal bar, wafer size measurement, wafer dicing positioning, wafer cutting positioning, appearance, foreign body defect inspection, IC pin flatness inspection, SMD packaging inspection, character recognition, etc.

Visual applications in medicine / medical industry
The wide applicability of machine vision technology in the pharmaceutical industry has won it a broader market space.
The inspection contents include: the filling positioning of liquid preparation, the inspection of unqualified capsules, the inspection of impurities and covers in bottles, the inspection of capsule contamination, the barcode inspection of pharmaceutical products' outer packaging, the appearance inspection of outer packaging, and the full box inspection of outer packaging carton.

Visual application in tobacco industry
Cigarette production speed is very fast, in the production process, there are many unqualified products. It is very necessary to find out the unqualified products and eliminate them by machine vision.
The inspection contents include:
The inspection of tobacco raw materials and impurities, the size measurement of filter cigarette holder, the defect inspection of roll paper packaging, the inspection of two-dimensional code on the surface of cigarette box, the deformation inspection of cigarette package, and the counting of cigarette box.

Visual application in printing industry
Printing industry is one of the common application industries of machine vision. Machine vision system can quickly and accurately detect all kinds of defects in printed matter, improve product quality and production efficiency, and reduce enterprise costs. There are various forms of printed matter to be tested, which can be divided into paper, plastic and metal plate according to the type of printing materials; and it can be divided into curl material and single sheet product from the printing form.
The inspection contents include:
Material defect inspection (such as holes, foreign matters, etc.), printing defect inspection (such as flying ink, knife thread, rubbing plate, inaccurate overprint, etc.), color defect inspection (such as light printing, color deviation, exposure, etc.).

Visual application in food / beverage industry
In recent years, in the high-speed production line of food / beverage, manual inspection can not meet the requirements of enterprises for food / beverage quality inspection. The rapid development of machine vision has greatly improved the inspection technology level of food / beverage industry. As a series of application problems have been solved, the degree of automation of food / beverage production based on machine vision application has been significantly improved.
The inspection contents include:
Quality inspection of glass bottles, such as bottle mouth damage, bottle body, bottle bottom foreign body inspection. Bottle counting, beverage filling positioning, filling liquid level inspection, foreign matter inspection after filling, label position and spray code identification.

Visual application in automobile industry
Due to the increase of labor cost, the shortage of human resources and the requirements of automobile industry for high efficiency, high precision, high quality and high intelligence, machine vision technology has been widely used in automobile industry, which can not only greatly improve work efficiency, but also improve product quality.
The inspection contents include:
Automobile hardware size measurement, appearance defect inspection, part bar code reading, sheet metal solder joint inspection, automobile parts assembly positioning, automobile lampshade, buckle and other character recognition, appearance inspection, panel identification inspection, etc.

Visual application of lithium battery industry
In recent years, China has vigorously promoted the development of new energy, and the lithium battery industry has developed rapidly. In the production process of lithium battery, the defects of negative electrode such as exposed foil, dark spot, bright spot and material dropping will be caused by coating machine and editing press. The use of machine vision inspection instead of manual selection of defective products can realize the automatic inspection of electrode pieces and ensure the safety of products.
The inspection contents include:
Coating appearance inspection, electrode lug welding positioning, dimension measurement, crease inspection, welding burst point inspection, battery surface character recognition, finished product appearance inspection, etc.

Visual application in photovoltaic industry
With the increase of labor cost, in order to ensure the product quality and avoid the quality problems caused by many human factors, in the process of solar panel manufacturing, from the front end to the back end, any link error will affect the power generation efficiency of the finished solar panel, such as secondary pollution and fragments. It is the development trend of photovoltaic manufacturing technology to use machine vision inspection technology instead of manual to detect and sort silicon and solar cells.
The inspection contents include:
Tin coating and dimension measurement of solar cells, welding positioning of solar panels, dirt inspection of welding surfaces, defect inspection of solar cells (cracks, impurities, voids, corner fracture, grid line fracture, slurry stains, color deviation), etc.