Warehouse environments and processes are generally less structured when compared to manufacturing environments. If you look at any given warehouse, it follows a line from inbound to outbound goods, with various stages of processing in between, from pallet level to case level to item level, with plenty of variation depending on the items being processed including fashion, food and beverage or general merchandise.
In logistics, items are generally variable – individual products, parcels, packages, and pallets come in a range of shapes, sizes, and materials, and they’re usually presented to machine vision cameras in a less structured manner, with items on wide conveyor belts and loading areas in all sorts of positions and clusters. It means machine vision, 3D and industrial scanning technology needs to be able to deal with wider variations and unexpected scenes and formations.
Machine vision has been the preserve of the manufacturing industry for the past 25 years – it’s where the technology and use cases began and were hardened over decades. But over the past three to five years, the accessibility of neural processing, the development of 3D sensing, and AI algorithms have improved. Software suites and hardware platforms can apply these newer tools in manufacturing, such as more subtle defects or anomaly detection, as well as in much less structured environments like logistics.
Technical Advances – Here’s the Proof
Visual automation solutions generally slot into four main categories where machine vision, industrial scanning, AI, and 3D can be brought to bear on a host of applications between inbound and outbound for each of these four categories.
The first is vision-guided robotics to pick and sort items. For example, an industrial bakery in Europe is securing lower error rates and higher throughput of goods using machine vision software guiding a robot. The bakery inspects its full range of breads using this integrated solution.
It can carry out efficient, intelligently automated picking with the robotic grip handling between 25 and 30 packages per minute, without damaging bread or packaging. It’s estimated that the solution has secured a 75% cost savings compared to traditional camera and lighting inspection approaches and eliminated the need for frontline workers to carry out repetitive manual visual inspection and picking.
Second is inspection, for product integrity, damage detection and order completeness. For example, hyperspectral imaging can be applied to understand whether boxes have leakage, possibly signifying a damaged product, which may be seen with a colour camera but not necessarily with a normal greyscale camera.
The third is measurement, specifically dimensioning items, parcels and pallets to understand different shapes and sizes so they fit into various packages, loading areas, and vehicles. The logistics team at Dimar, a national grocery retailer has cut workflow times by up to 50% with a new intelligent automation solution. Its built on a fleet of mobile devices with integrated parcel dimensioning software using modern time-of-flight sensors combined with AI algorithms. Working together, it virtually reconstructs and measures standard cuboidal parcels and irregular non-cuboidal items.
The final category is identification, which uses various methodologies such as barcode, data matrix, colour, optical character recognition (OCR), pattern and shape. Third-party logistics company Noerpel-Group receives up to 25 truckloads of goods each week at one of its sites from a leading fashion and lifestyle retail customer. The associated data is logged in Noerpel’s enterprise resource planning (ERP) system. Each box is redistributed to its final destination and has achieved over 50% in time savings in inbound storage operations using fixed industrial scanners.
During the first five months after its installation, the solution has enabled Noerpel to intelligently automate the scanning and validation of around 700,000 packages. As the complete process time for each package has been dramatically reduced, follow-up processes have been further automated, and staff now only need to intervene by exception.
Another example is an industrial bakery called Zeelandia, which is saving at least €20,000 annually with fixed industrial scanners affixed to forklifts and powered by machine vision software. Zeelandia’s distribution teams are better connected to workflows, and benefit from less downtime and stress, more consistent output for customers, and more time to complete strategic tasks.
Today’s 3D Casts new Light
As with 2D machine vision, more structured manufacturing sectors have led the way with 3D scanning. I know of a vision specialist helping automotive OEMs secure 10-15% defect rate reduction for items as complex as car doors. The machine vision specialist creates a solution built on dual-camera, single-laser 3D sensor integrated with AI software. The 3D sensor scans items like car doors, capturing thousands of data points, and turns those into highly detailed point cloud and depth map representations for the AI software to interpret and inspect for defects. Thankfully, warehousing and logistics organisations are also catching up here.
In logistics, traditional structured light 3D scanners have been popular, as they provide submillimetre resolution and high accuracy for scanning static scenes. But if the sensor or the scene moves during the scanning process – which is always possible given the unstructured nature of logistics goods and conveyor belts – the 3D scan will be distorted. Traditional time-of-flight systems are also useful for certain use cases and offer very fast scanning speed and data acquisition, but some may compromise on resolution and noise levels.
However, newer generations of 3D sensing are delivering highly accurate scanning using unique parallel structured light technology. This allows for real-time, high-resolution 3D scanning of objects moving at speed, by constructing multiple virtual images within one exposure window. This opens new possibilities in warehousing and logistics robotic guidance, damage detection, advanced bin and conveyor belt picking, uniform and mixed palletisation and depalletization, and digital twinning.
You can learn more about intelligent automation solutions for visual inspection here.
