Intelligent image processing refers to a class of computer-based image processing and analysis technologies that are adaptive to various applications. It is an independent theoretical and technical field, but at the same time, it is a very important technical support for machine vision.
Machine vision with intelligent image processing capabilities is like giving machines eyes while endowing them with intelligence, enabling them to "see" and "see accurately." It can replace or even surpass the human eye in measurement and judgment, allowing machine vision systems to achieve high-resolution and high-speed control. Furthermore, machine vision systems operate without contact with the object being inspected, making them safe and reliable.
Machine vision technology
The origins of machine vision can be traced back to the image processing research of American scholar L.R. Roberts on the world of multifaceted blocks in the 1960s, and the introduction of the "Machine Vision" course at the MIT Artificial Intelligence Laboratory in the 1970s. By the 1980s, a global surge in machine vision research began, and several application systems based on machine vision emerged. Since the 1990s, with the rapid development of computer and semiconductor technologies, the theory and application of machine vision have been further developed.
Since the beginning of the 21st century, machine vision technology has developed at an even faster pace and has been widely applied in various fields, such as intelligent manufacturing, intelligent transportation, healthcare, and security monitoring. Currently, with the rise of artificial intelligence, machine vision technology is in a new stage of continuous breakthroughs and maturation.
In China, research and application of machine vision began in the 1990s. Starting by tracking foreign brands, and after more than two decades of effort, domestic machine vision has grown from nothing to something, from weak to strong. Not only has theoretical research progressed rapidly, but some highly competitive companies and products have also emerged. It is estimated that with the continued deepening of domestic research, development, and promotion of machine vision, catching up with and surpassing world-class levels is no longer a distant dream.
Intelligent image processing technology
Machine vision image processing systems perform calculations and analyses on digital image signals from the field according to specific application requirements, and control the actions of field equipment based on the processing results. Common functions include:
(1) Image acquisition
Image acquisition is the process of obtaining images of a scene from the work site. It is the first step in machine vision, and the acquisition tools are mostly CCD or CMOS cameras or video cameras. Cameras acquire single images, while video cameras can acquire continuous images of the scene. In terms of a single image, it is actually a projection of the three-dimensional scene onto a two-dimensional image plane. The color (brightness and chromaticity) of a point in the image reflects the color of the corresponding point in the scene. This is the fundamental basis for using acquired images to represent the real scene.
If a camera outputs analog signals, the analog image signal needs to be digitized before being sent to a computer (including embedded systems) for processing. Most cameras now directly output digital image signals, eliminating the need for analog-to-digital conversion. Furthermore, modern cameras have standardized digital output interfaces, such as USB, VGA, 1394, HDMI, WiFi, and Bluetooth, allowing direct input to a computer for processing and eliminating the need for a separate image acquisition card between the image output and the computer. Subsequent image processing is typically performed in software by the computer or embedded system.
(2) Image preprocessing
Acquired digitized images are often subject to varying degrees of interference due to equipment and environmental factors, such as noise, geometric distortion, and color misalignment, all of which can hinder subsequent processing. Therefore, image preprocessing is essential. Common preprocessing techniques include noise reduction, geometric correction, and histogram equalization.
Typically, time-domain or frequency-domain filtering methods are used to remove noise from images; geometric transformations are employed to correct geometric distortions; and histogram equalization and homomorphic filtering are used to mitigate color deviations. In short, this series of image preprocessing techniques "processes" the acquired images, providing "better" and "more useful" images for machine vision applications.
(3) Image segmentation
Image segmentation involves dividing an image into regions with distinct characteristics according to application requirements, and then extracting the target of interest from these regions. Common features in images include grayscale, color, texture, edges, and corners. For example, segmenting an image of an automobile assembly line into a background region and a workpiece region provides the data to the subsequent processing unit for processing the workpiece assembly section.
Image segmentation has been a challenging problem in image processing for many years, and while numerous segmentation algorithms exist, their effectiveness is often unsatisfactory. Recently, deep learning methods based on neural networks have been used for image segmentation, achieving performance superior to traditional algorithms.
(4) Target identification and classification
In industries such as manufacturing and security, machine vision relies heavily on the identification and classification of targets in input images to enable subsequent judgments and operations. Recognition and classification technologies share many similarities; often, once target recognition is complete, the target's category is also clear. Recent image recognition technologies are transcending traditional methods, forming intelligent image recognition methods dominated by neural networks, such as Convolutional Neural Networks (CNNs) and Regressive Neural Networks (RNNs), which offer superior performance.
(5) Target location and measurement
In intelligent manufacturing, the most common task is the installation of target workpieces. However, before installation, the target often needs to be positioned, and after installation, it needs to be measured. Both installation and measurement require high precision and speed, such as millimeter-level accuracy (or even smaller) and millisecond-level speed. Such high-precision, high-speed positioning and measurement are difficult to achieve using conventional mechanical or manual methods. In machine vision, image processing methods are used to process the images of the installation site, processing them according to the complex mapping relationship between the target and the image, thereby quickly and accurately completing the positioning and measurement tasks.
(6) Target detection and tracking
Moving target detection and tracking in image processing involves real-time detection of moving targets within scene images captured by a camera, predicting their next direction and trend of motion—that is, tracking. This motion data is then promptly submitted to subsequent analysis and control processing to generate appropriate control actions. Image acquisition typically uses a single camera, but two cameras can be used if needed to mimic human binocular vision and obtain stereoscopic information of the scene, which is more conducive to target detection and tracking.
Shenzhen Haotianchen Technology Co., Ltd. is a high-tech enterprise integrating R&D, customized solutions, and sales, and a provider of production line equipment upgrade and transformation solutions. Since its establishment, the company has focused on the field of machine vision inspection, independently developing and producing machine vision inspection equipment, automated vision inspection equipment, machine vision appearance inspection equipment, optical automated inspection equipment, CCD vision inspection equipment, optical sorting machines, and machine vision inspection systems. It also provides customized machine vision inspection solutions, offering non-standard automated inspection equipment to various enterprises and manufacturers. These solutions automate the measurement of dimensions, appearance defects, and character recognition for products in fields such as new energy batteries, PCB circuit boards, precision components, and electronic components, helping customers improve production efficiency, product quality, reduce labor costs, and enhance market competitiveness.
