Posted on

Picking Solutions: How Complex Must Your System Be?

Bin-picking, random picking, pick and place, pick and drop, palletization, depalletization—these are all part of the same project. You want a fully automated process that grabs the desired sample from one position and moves it somewhere else. Before you choose the right solution for your project, you should think about how the objects are arranged. There are three picking solutions: structured, semi-structured, and random.

As you can imagine, the basic differences between these solutions are in their complexity and their approach. The distribution and arrangement of the samples to be picked will set the requirements for a solution. Let’s have a look at the options:

Structured picking

From a technical point of view, this is the easiest type of picking application. Samples are well organized and very often in a single layer. Arranging the pieces in a highly organized way requires high-level preparation of the samples and more storage space to hold the pieces individually. Because the samples are in a single layer or are layered at a defined height, a traditional 2-dimensional camera is more than sufficient. There are even cases where the vision system isn’t necessary at all and can be replaced by a smart sensor or another type of sensor. Typical robot systems use SCARA or Delta models, which ensure maximum speed and a short cycle time.

Semi-structured picking

Greater flexibility in robotization is necessary since semi-structured bin picking requires some predictability in sample placement. A six-axis robot is used in most cases, and the demands on its grippers are more complex. However, it depends on the gripping requirements of the samples themselves. It is rarely sufficient to use a classic 2D area scan camera, and a 3D camera is required instead. Many picking applications also require a vision inspection step, which burdens the system and slows down the entire cycle time.

Random picking

Samples are randomly loaded in a carrier or pallet. On the one hand, this requires minimal preparation of samples for picking, but on the other hand, it significantly increases the demands on the process that will make a 3D vision system a requirement. You need to consider that there are very often collisions between selected samples. This is a factor not only when looking for the right gripper but also for the approach of the whole picking process.

Compared to structured picking, the cycle time is extended due to scanning evaluation, robot trajectory, and mounting accuracy. Some applications require the deployment of two picking stations to meet the required cycle time. It is often necessary to limit the gripping points used by the robot, which increases the demands on 3D image quality, grippers, and robot track guidance planning and can also require an intermediate step to place the same in the exact position needed for gripping.

In the end, the complexity of the picking solution is set primarily by the way the samples are arranged. The less structured their arrangement, the more complicated the system must be to meet the project’s demands. By considering how samples are organized before they are picked, as well as the picking process, you can design an overall process that meets your requirements the best.

Posted on

Add Depth to Your Processes With 3D Machine Vision

What comes to mind first when you think of 3D? Cheap red and blue glasses? Paying extra at a movie theater? Or maybe the awkward top screen on a Nintendo 3DS? Neither industrial machine vision nor robot guidance likely come to mind, but they should.

Advancements in 3D machine vision have taken the old method of 2D image processing and added literal depth. You become emerged into the application with true definition of the target—far from what you get looking at a flat image.

See For Yourself

Let’s do an exercise: Close one eye and try to pick up an object on your desk by pinching it. Did you miss it on the first try? Did things look foreign or off? This is because your depth perception is skewed with only one vision source. It takes both eyes to paint an accurate picture of your surroundings.

Now, imagine what you can do with two cameras side by side looking at an application. This is 3D machine vision; this is human.

How 3D Saves the Day

Robot guidance. The goal of robotics is to emulate human movements while allowing them to work more safely and reliably. So, why not give them the same vision we possess? When a robot is sent in to do a job it needs to know the x, y and z coordinates of its target to best control its approach and handle the item(s). 3D does this.

Part sorting. If you are anything like me, you have your favorite parts of Chex mix. Whether it’s the pretzels or the Chex pieces themselves, picking one out of the bowl takes coordination. Finding the right shape and the ideal place to grab it takes depth perception. You wouldn’t use a robot to sort your snacks, of course, but if you need to select specific parts in a bin of various shapes and sizes, 3D vision can give you the detail you need to select the right part every time.

Palletization and/or depalletization. Like in a game of Jenga, the careful and accurate stacking and removing of parts is paramount. Whether it’s for speed, quality or damage control, palletization/ depalletization of material needs 3D vision to position material accurately and efficiently.

I hope these 3D examples inspire you to seek more from your machine vision solution and look to the technology of the day to automate your processes. A picture is worth a thousand words, just imagine what a 3D image can tell you.