Vision Systems in the Warehouse and the Middle Child Syndrome

Warehousing and distribution centers continue to be very active areas for incorporating the latest innovations in automation technology. Today, some systems are conveying and sorting up to 300 cartons per minute. This is almost mind numbing considering the variety of package sizes and weights that exist in most warehouses.

To handle these increasing demands, warehouses and distribution centers continue to undergo a quiet revolution in the adoption of advanced technologies to store, retrieve and convey goods. Palletizing robotics, autonomous fork lifts, tracking systems, and scanning tunnels are now the foundation of a new generation of highly automated warehouses. Underpinning many of these systems is machine vision technology. Machine vision is defined as the combination of a digital camera, lighting and sophisticated algorithms used for location detection and identifying 2D and 3D object features. At one point in time, machine vision was relegated to the semiconductor and electronics markets where investments of $50 to $100 thousand were the typical price tag for an application deployment in a production cell or machine. However, over the last 10 to 12 years, the steady decrease in pricing of machine vision systems has reduced complex machine vision applications to under $10 thousand and vision sensors can be applied for about $1.5 thousand. In conjunction with lower system prices an increase in processing capabilities and the ease of deployment have enabled vision to be applied economically to challenges in the warehousing.

Machine vision integrated with robotic systems and high speed automation is making inroads into warehouse on several fronts including: barcode reading, volumetric scanning, AGVs using natural feature navigation, smart conveyance, package inspection, 3D guided robotic palletizing.

In applications such as barcode scanning, machine vision has not displaced the laser scanner. Over the last decade there have been numerous attempts to replace the 1D barcode with RFID tags or 2D barcodes. However, the 1D barcode has been and remains the dominant method for tracking movement of packages throughout the warehouse. Neither the 2D barcode or RFID has been able to provide the cost effectiveness and speed of using a 1D barcode inside the warehouse. The laser scanner continues to remain the dominant method for scanning barcodes because it is inexpensive, fast and reliable. However, this may begin to shift as sustainability initiatives are bringing the reuse of shipping cartons into widespread use. Shipping cartons with numerous barcodes are challenging for the traditional laser scanner to isolate the correct barcode on the package. This is where machine vision sensors are now starting to demonstrate their prowess. Vision systems are beginning to displace the barcode scanners in warehousing operations that reuse containers as the vision algorithms are able to discern a valid barcode in a crowded field.

Machine vision systems are enabling conveyance systems to create order out of disorder. The scanning tunnel preceding the cross belt sorters are organizing packages by volumetric size. Using high speed vision arrays that employ multiple cameras and laser scanning technology, scanning tunnels are able to determine the volumetric size of packages moving at conveyance speeds. Shipping container and packaging material are minimized as packages are automatically being sorted by size. Some operations have employed these sortation systems to preselect shipping cartons for the operators down the end of the line. Others applications have used this as another form of quality inspection making sure packages have not been damaged prior to shipping. Vision systems can be used for this purpose to detect deformities and bulges in packages.

AGV and forklift truck technology has largely undergone a transformation where many of the well-entrenched material handling companies have integrated more robust navigation systems into these mobile platforms. For example, Kiva robots are equipped with vision systems that read stickers affixed to a concrete warehouse floor. The stickers form a grid covering the entire robot operating area. The next step for this market is natural feature recognition which is being deployed commercially in fork lift trucks. Low cost camera technology has been employed as a substitute to laser-based guidance. Laser-based systems have relied upon placing of reflective tape in key locations in the facility to enable the autonomous vehicle to determine its coordinates; camera technology will instead utilize object recognition in the facility for guidance purposes.

It is very expensive to ship the wrong item to a customer. To avoid this many warehouses drive goods at high speeds over weigh stations to insure the right product is going to the customer. If the weight matches what the system says the SKU should weigh, the item goes on its way. If not, it is kicked out into an exception process. In some warehouse operations, new SKUs are added and deleted every day. Continually measuring and weighing products and adding this master data into a WMS can slow down the operation. Vision systems can also be used to determine the volume and weight of an item without interrupting the flow of goods. When these goods are routed to a pack station, an appropriately size box can be available. By putting items in boxes that are not too big, major savings in parcel freight spend can be achieved.

Robotic palletizing is often integrated with 3D vision systems to depalletize incoming orders or prepare pallets for shipment. 3D vision systems along with robust robotic software applications are now being exploited to replace manual labor used to build pallets. Stereoscopic 3D vision systems have been the primary technology used in these applications, however “time of flight” technology is beginning to show some promise. “Time of Flight” is being employed in the consumer market for applications such as autonomous driving and gesturing in interactive media. As this technology proliferates, the cost is will rapidly decline due to scale thereby surely displacing expensive stereoscopic systems.

The combination of vision systems with smart conveyors, where a small section of a conveyor has its own logic and sensors to allow it to move items within its small domain more efficiently, will permit new warehouse capabilities. Wii technology provides a useful analogy of what is likely to come to the warehouse in the future. The Wii is an electronic game that senses the motions of players; it is inevitable that this form of technology will come to the warehouse. So, for example, it is possible to imagine a worker loading a truck supported by an extendable conveyor. One can imagine that worker making hand gestures to speed up or slow down the conveyor speed and other motions that signal the conveyor to extend further into the trailer or begin to pull back into the warehouse.

In conclusion, new technologies are impacting the warehouse on many fronts. But we should not lost sight of the increased price/performance of visions systems and how vision systems are working hand in hand with many of these new technologies. Vision systems in the warehouse are like a middle child – they just don’t get enough attention.

Sal Spada is the Research Director Discrete Automation at the ARC Advisory Group.  Sal’s research includes material handling, robotics, and of course Vision Systems.