How we used lasers & 3D camera’s for faster package deliveries
Researching the challenge
When delivering hundreds of parcels a day, finding each box in your van a couple of seconds quicker has a significant impact on the delivery schedule. Bit teamed up with a multinational courier service to test the feasibility of using depth-sensing cameras in detecting the location of parcels during the delivery process.
Challenge: Saving time through faster parcel detection and delivery
How can we help the driver save time when delivering? Parcels get sorted per postal code in the sorting centre, but often still get lost in the van. Drivers are often looking for the right packages for minutes. One way of helping the driver save time when delivering is helping to find the box quicker. Our challenge was to test whether we could do this, using depth-sensing cameras. To figure this out, we constructed a simulation of a truck where parcels are loaded in and installed 3D depth cameras which would be able to capture the placement of boxes.
Hypothesis: To locate a parcel faster, we assumed that we could use 3D depth cameras to capture the placement of boxes. When a new parcel gets added to the shelf of a truck, the cameras extract the position and potentially dimensions of the package and save those to a 3D model.
Prototyping the solution
To do so, we mounted 18 depth-sensing cameras to shelves in the truck and a laser to the ceiling. These cameras scan the shelves on the opposite side before the driver is loading the parcels. The infrared cameras together create a depth map, which detects empty spots in the van. The driver scans a box with the barcode scanner, puts the parcel in the truck and when he is out, the cameras detect the position of that parcel, because now suddenly an empty space is filled. The cameras detect whether a person is still in the van, and will only start scanning once no person is in the truck. By using block detection, we decreased potential issues of visual noise.
The centroid of the detached parcel is saved together with the barcode ID. When the truck is loaded, and the driver would start his route, he can enter a barcode ID. This triggers the laser to turn on, which then can point at the desired parcel at any location in the van. If another package is in front of the desired parcel, the laser will keep blinking, indicating it is behind this second object. With this technique, the driver can immediately see which package he has to take out and deliver.
Conclusion
Through using 18 depth-sensing cameras, we were able to build a system that scans the van for empty spots and new packages through using block detection. By combining this technique with a barcode scanner and a 360 degrees laser at the ceiling, the driver can find the location of a parcel immediately when entering the barcode ID, saving them minutes, if not hours, throughout a day of delivery.
