Almond shells and other foreign matters are removed from the product flow in order to provide a clean product for consumers.
In addition to color cameras, hyperspectral cameras are used to detect foreign matter types which are optically identical to actual almonds. They enable the differentiation of materials based on their chemical composition.
The data of the used color and hyperspectral cameras is synchronized precisely, thus ensuring the distinct identification of a single object which is represented by more than one image source.
Another challenge is the fact that many objects overlap with other objects in the product flow. To prevent the resulting decline of the sorting accuracy, these object clusters are separated by the software so that each object can be evaluated individually.
- seamless synchronization of sensors
- separate evaluation of overlapping objects
- reaction time < 1 ms
- classification accuracy > 99,99%
- real time evaluation
Original Image: almonds
Result Image: almonds
Original Image: foreign matter
Result Image: foreign matter
Original Image: overlapping almonds
Result Image: separated almond cluster