Vision systems are helpful in the sense that they allow you to do visual checks automatically without employee inspection, but they can be difficult to set up ideally.
As an example, we could see how a vision system might apply to selection of hams. When cutting the pig into its individual components, the ham is typically cut right above a joint where the aitchbone is located. The distance from the cut to the aitchbone relates directly to the diameter of the ham bone at the cut. This can be used to make a general comparison of ham sizes.
The five different muscle groups in a ham can be used for a variety of things. Certain sized hams get put into machines that cut them into spiral cut hams and some are sold as whole hams.
To categorize each ham for the proper use, weight and size are needed. This could be done with the combination of conveyor scale and vision system. While the ham crosses over the conveyor scale, a camera can look at the cross section to detect the size of the bone.
This is where the process can become difficult. Vision systems can run into problems with lighting and color that you might not encounter with human handling. For example, a dirty bulb which discolors the light could in turn discolor the meat for the camera. This makes it very tough for the camera to identify things correctly because vision systems compare the product to a number of acceptable images to justify what range the product falls into.
On the color side of things, the process can become difficult based on where the ham was cut. The vision system detects pixels that vary from the red hue of the meat in order to find the bone, but the color of the bone isn’t always pronounced. If the bone is cut in such a way that it is less visible, the vision system could miss the bone entirely as the marrow itself often contains colors similar to the ham.
While this operation has a lot of potential, it requires that the cutting of hams be precise for bone visibility and that lighting be optimized. On the other hand, a human could spot the bone fairly easily after a little examination. This shows that both human estimation and machine vision have a place. Vision systems have the power to greatly speed up certain processes, but the current technology for them does much better with well-defined trait differences. Feel free to ask questions about some of the processes that machine vision could be beneficial for.