Machine vision uses machines to replace human eyes for measurement and judgment. A machine vision system refers to a system where machine vision products (i.e., image acquisition devices, either CMOS or CCD) convert the captured target into an image signal, which is then transmitted to dedicated image processing software. Based on pixel distribution and information such as brightness and color, this signal is converted into a digital signal. The image processing software performs various calculations on these signals to extract the target's features, and then controls the on-site equipment based on the judgment results. With the booming development of China's manufacturing industry, China is becoming one of the most dynamic regions for machine vision development internationally. Large-scale applications have expanded from the initial electronics and pharmaceutical industries to packaging, automobiles, and other sectors, and are also widely used in transportation and printing. The application scope virtually covers all sectors of the national economy. Choosing the right machine vision software has become a crucial part of achieving successful automation. Most vision companies, due to their limited experience, cannot systematically explain this to newcomers. Vision Image, as a seasoned company with ten years of experience, is willing to share its experience and progress together with everyone through years of technological accumulation. This article focuses on the key points of selecting machine vision software and introduces some commonly used vision software.
I. Key Considerations for Machine Vision Software Selection
1. Accuracy of the positioner
Accurate target or feature localization is a crucial function of any detection system or vision-guided motion system. Traditional object localization relies on grayscale values for object identification. While this technique is widely used, it lacks stability when image quality deteriorates. Image quality degradation can be caused by factors such as clutter, varying brightness, and obscuring elements. In contrast, geometric target localization is a more recent approach that uses the target's outline to identify its features. Vision Image's self-developed image processing software largely employs geometric target localization, offering extremely high efficiency and effectively assisting users in solving localization problems.
2. Tool library or application software?
Machine vision software is primarily sold in two typical ways: one is as a tool library containing various processing algorithms, and the other is as application software specifically designed for a particular type of task. Each has its advantages and disadvantages, and it's crucial to confirm this characteristic before making a choice. Vision Image provides image processing software that encompasses both, offering solutions for professional applications while also meeting the diverse needs of its clients.
3. Easy to program and operate
A concise and intuitive graphical interface is key to ease of use and setup. The main difference between today's machine vision products lies in their graphical interfaces. An interface can be evaluated from two aspects: "setup" and "operation." It may be quite complex for an engineer, but very simple for an operator.
4. Subpixel precision
The resolution of a visual system is the smallest feature that the system can distinguish. For example, a visual field size (FOV) of 1'' using a 640x480 pixel computer image will yield a resolution of 1/640, or 0.00156''. In fact, machine vision algorithms have sub-pixel capabilities. That is, these algorithms can measure or derive units smaller than a pixel.
5. Easy to upgrade
Machine vision systems can be used in a variety of applications, ranging from high-tech cameras to surveillance systems. When selecting a system, future upgrades should be considered. General-purpose vision software systems are easily upgradable. End users typically consider future system needs based on additional cameras, changes in lighting, and changes in vision tools.
6. Image preprocessing algorithm
Most image processing software provides preprocessing algorithms, so it is necessary to compare the richness of the algorithms they offer, their efficiency, etc.
7. System Integration
If you are not very proficient in machine vision skills, then you will need a system integrator for your project. Ideal vision products can be widely accepted by system integrators.
II. Introduction to Commonly Used Machine Vision Software
1. OpenCV, a free and open-source image processing library.
OpenCV is a free and open-source computer vision library released in recent years. Its functions make digital image and video processing very convenient. Furthermore, using the object-oriented VC++ 6.0 programming tool, OpenCV algorithms are written in C++, greatly facilitating digital image processing in VC++ programming. It is poised to become a powerful tool in the field of image and video processing.
Disadvantages: As it is open-source software, there are many versions and complex function libraries, resulting in reduced execution efficiency. It is more suitable for scientific research and learning than for industrial applications.
2. Halcon, a powerful image processing library
HALCON, developed by the German company MVtec, is a complete and standard machine vision algorithm package with a widely used integrated development environment (IDE). It reduces product costs and shortens software development cycles—HALCON's flexible architecture facilitates the rapid development of machine vision, medical imaging, and image analysis applications. It is widely recognized in European and Japanese industries as the highest-performing MachineVision software.
Disadvantages: The price is relatively high, and a new license needs to be purchased each time it is distributed.
3. NIVision, a rapidly validated image processing library (including Vision Assistant and VBAI)
NI Vision software's graphical programming and vision algorithm engineering enable greater development efficiency. Choose from configurable environments and comprehensive programming libraries to better meet your needs and get started quickly. They work well with all hardware combinations, allowing for easy use of existing code and management and maintenance of multiple hardware systems.
Disadvantage: Good things are bound to be expensive.
4. Matlab: the best choice for verifying underlying algorithms.
MATLAB is easy to use, especially since it integrates many image processing functions, covering almost all image processing methods. Common operations are made into corresponding built-in functions, freeing users from worrying about the minutiae of fixed and mature algorithms such as how to read images and how to convert color spaces, allowing them to focus their main efforts on algorithm research.
Disadvantages: MATLAB does not have sufficient control over the details of the processing method.
