Unfortunately, a fair portion of our business is fixing failed machine vision installations. Below I talk about the top three reasons for failure we typically see. But for mostly they all have one thing in common: they fail to recognize that the camera, lighting and software have to work together to expose the defect properly.
Machine vision OEM's would have you believe it's all about the hardware. The best camera in the world won't find your defect if all the other pieces don’t work together in harmony. Making machine vision work is all about getting the right pieces to work together to deliver an acceptable image.
At SolVIS we call this process “getting the front end correct.” The front end is the camera, optics, lighting, triggers, component speeds and algorithms working together as a system. If the optics distort the image too much, you’ve got problems. If the lights don't expose your defect adequately, good luck. Pick a camera with too high a resolution, and you may not be able to process the image fast enough. This give and take is what the machine vision engineer works to balance.
Top 3 Reasons Machine Vision Systems Fail
1. Using the wrong lighting
"Light makes photography. Embrace light. Admire it. Love it. But above all, know light. Know it for all you are worth, and you will know the key to photography." ~ George Eastman
Every good photographer knows lighting is key. The goal of proper lighting in machine vision applications is to make the area of interest stand out from the rest of the image. Think contrast. There are a half dozen or so lighting techniques and variations of each. Find the right one and your area of interest will pop out of the image. You can send SolVIS sample parts so we can use our extensive machine vision lab tools to find the ideal lighting solution for you. Or, you can take a deep dive on the subject by reading these two great white papers from Vision Online here and here.
2. Insufficient resolution and optics
Your smallest defect needs to show up at least 4 pixels wide. Generally speaking, more pixels means better resolution of your defect, but comes at a higher financial, speed and computing cost. You need to think of the problem as you would in regular photography; your eyes find it much easier to discern a crisp, clear image versus a blurry image. The same holds true of your machine vision system. If the area of interest is not crisp and clear, your system won't stand a chance. Read our white paper on selecting camera resolution here.
3. Expecting too much from the system
We tend to forget that our brains do most of the seeing, not our eyes. Don't believe me? How do you know the difference between a dog and a cat? They both have four feet, fur, similar noses and eyes. You have to combine everything your senses are telling you to make the determination. In the machine vision world, assuming a good picture, the solution here is software. In the past, we solved these problems with lots of code. But recently, deep learning and artificial intelligence products have come to market, making these problems solvable. For example, Cognex's VisionPro Vidi system uses new deep learning techniques and artificial intelligence in an easy-to-deploy, powerful solution. If you want to take a peek under the hood on machine vision AI and deep learning, I highly recommend watching this video here . And remember, some types of machine vision problems need some very powerful code and coding techniques to make it work.
It can take years to gain the experience and wisdom you need to avoid the traps machine vision in the factory can present. Fortunately, the engineers at SolVIS have spent the past 20 years in the trenches of machine vison and automation. We have seen what works and what does not. We learn from our mistakes and constantly try the latest machine vision techniques on customer products every week. We solve problems and put together complex machine vision solutions that just plain work. I encourage you to send us your hardest challenge today, and we will find a solution for you.
Let us help you get a clear vision of what's possible.
Adam Nelson