Hello everyone, it's hard to believe I've been working in the machine vision automation industry for six years now. I'm sharing this article in the hopes of providing beginners with some learning experience and helping them avoid some pitfalls. Please excuse my amateurish attempts.
I studied Mechanical Design, Manufacturing and Automation in Dongguan. At the time, I felt that while mechanical design graduates could easily find jobs in the Pearl River Delta, earning a high salary required many years of experience. Since many people were already doing this, it lacked innovation. So, I became an assistant in the mechanical engineering lab, learning PLCs and touchscreens. My teacher told me that mastering these skills would lead to a high salary after graduation, so I gradually drifted away from mechanical design and focused on learning Mitsubishi PLCs and KingSCADA touchscreen programming software. It wasn't until my first semester of senior year, when I was looking for a job, that I discovered there were many people working with PLCs, including those with low educational backgrounds. One particularly disheartening experience was at a job fair organized by Zhilian in Dongguan. A guy asked me how much I wanted, and I said 3,000 RMB. He told me, "You're a graduating student with no work experience; you can only offer me 1,500 RMB at most." I was stunned.
Later, I took a course at school called "Machine Vision," which I found quite interesting. I felt that this might be a future direction for automation, so I diligently searched for a lot of information online to learn. For example, I bought a cracked version of "Halcon 10" software on Taobao to study on my own. I also found a job at Rui Company in Nancheng, Dongguan, which specializes in machine vision solutions. After seeing the projects done by experienced colleagues, I thought about learning C++ and bought Sun Xin's C++ video tutorial on Taobao. That's how I truly embarked on this industry.
Later, I worked at Foxconn for two or three years, working on projects such as automatic scanning bars, hole inspection machines, steel sheet attaching machines, and adhesive attaching machines. I used Dalsa and Basler cameras, while other colleagues used GrayPoint and Baumer cameras. For machine vision software development kits, I used Halcon, VisionPro, and OpenCV. Some colleagues also used Dalsa's own image processing software package, but I forgot the name. For light sources, I used Opto and Christie (now I think it's called Ruishi). It seemed like there were many light source manufacturers in Chang'an, Dongguan. We also used motion control cards from the Industrial Technology Research Institute (ITRI) in Taiwan. Later, I joined my current company, working on die bonders and AOI inspection machines, which I've been doing ever since. I've used GrayPoint and ImageSource cameras, MIL image processing software, and Leadshine and Googol motion control cards. There's also the "Production Management Manager" app, a mobile app for monitoring equipment capacity. It's quite good; customers can check equipment capacity anytime, anywhere using their phones. Its API for automatically uploading data is simple and easy to use. You can search for the WeChat ID "lindrss" and ask for "Production Management Manager."
After rambling on for so long, let's summarize:
1. It's best to acquire images using the camera's built-in API interface, rather than using image acquisition software based on standard image acquisition protocols. Camera manufacturers understand their own methods and know how to achieve more stable image acquisition. Also, always use software triggering. Software triggering means sending a software command to the camera to acquire an image only when needed. Using both API interface and software triggering ensures more stable image acquisition, preventing software crashes at the customer's site after a period of operation, greatly saving time and effort.
2. Image processing software packages: Halcon is the best, VisionPro is simple and easy to use, OpenCV is the most economical, and MIL is cheap. At Foxconn, a computer-based system running Halcon and VisionPro costs around 15,000 RMB. If you want to learn, you can just get a cracked version from Taobao. If you want to use a genuine and cheap version, I recommend MIL; MIL's positioning module is only around 2,000 RMB. As for OpenCV, its positioning templates aren't very good, so it's fine for simple applications. Its biggest advantage is that it's free and open-source.
3. Cameras: Basler and Baumer are the most expensive, Dasla line scanners are better, gray dot cameras offer the best value, and ImagesSource from Germany is also decent. Basler's SDK functions provide error capture functionality, which I think is excellent; other brands of cameras don't seem to have this feature. It allows you to control image capture when the camera malfunctions, making it easy to troubleshoot the cause without popping up a memory read error message that leaves you clueless about how to troubleshoot.
4. Motion control cards: It seems that automation engineers in Hong Kong and Taiwan prefer ITRI brands, while mainland China generally uses Googol and Leisai. ITRI is more expensive, but its stability is excellent. Googol is a few hundred yuan more expensive than Leisai, but Googol's cards and adapters are of better workmanship and quality. Their function libraries are similar. I use Leisai the most; my boss thinks it's cheap and sufficient, avoiding unnecessary complexity.
5. Light source: Lighting is crucial and directly affects the quality of the acquired image. A good image has clear outlines and perfectly captures the characteristics of the product. If your company has the budget, you can have a lighting company test your lighting setup. They offer a wide variety of lights, including ring lights, coaxial lights, high-power lights for use with line scan cameras, as well as white, red, and blue light, ultraviolet light, etc. If your company doesn't have the budget, you can buy a light source for a few hundred yuan on Taobao.
6. Capacity Manager: This app allows you to monitor equipment capacity. It's a current trend. I remember many standard machines, such as foreign wire bonders and die bonders, offer corresponding network management software, but not apps. This Capacity Manager provides both an app and a complete set of network management software. Vision software engineers can simply use its API; it's very easy. You can search for the WeChat ID "lindrss" and ask for "Capacity Manager".
7. This line of work is quite lucrative right now; I don't have to worry about finding a job. Headhunters often call to ask if I want to switch jobs. Since I've chosen this path, no matter how bumpy the road ahead, I'll finish it, even if I have to crawl on my knees.
When your talent can't yet support your ambition, obediently continue to work hard to improve yourself, learn more, communicate more, and learn from industry veterans to avoid detours. If you do these things, you will definitely grow quickly!
Disclaimer: This article is a reprint. If it involves copyright issues, please contact us promptly for deletion (QQ: 2737591964). We apologize for any inconvenience.
