Neural-network-based system classifies waste batteries

A Swedish company originally formed by Claes Strannegård, an assistant professor at Chalmers University of Technology, has developed a vision-based system that can automatically sort waste batteries.

Snap Battery 1207vsd

A Swedish company originally formed by Claes Strannegård, an assistant professor at Chalmers University of Technology, has developed a vision-based system that can automatically sort waste batteries.

Snap Battery 1207vsd

Optisort's battery sorter identifies different sorts of batteries, after which it sorts them into six different types: alkaline/zinc carbon, Ni-MH, Ni-Cd, lithium, Li-ion, and unknown.

Images of each battery are captured by a vertically mounted scout camera from Basler as they travel along a conveyor at 100 cm/sec.

Features from the images of each battery are then extracted and presented to a pretrained PC-based neural network system where they are matched into different classes. The system, which can be updated to identify new battery types, can classify batteries correctly even when large parts of the images are covered or distorted.

After the system has classified the batteries, they are separated by air jets into respective types with an accuracy of 98%. The machine can automatically sort all kinds of portable batteries from large D types to button cells (which are separated but not sorted) at a rate of between 4 and 8 metric tons per day.

Besides sorting, the system also collects statistical data such as the number of different brands, size distribution, and chemistry -- statistics that provide battery manufacturers with valuable market data.

More Vision Systems Issue Articles
Vision Systems Articles Archives
More in Cameras & Accessories