Vision Partners conducted a study that used a hyperspectral camera to take images of the filleted product in the NIR spectral region (900 - 1700 nm) with the goal of reliably and industrially identifying remaining meat, cartilage, fat and bone. The camera measures 224 different frequencies in the near-infrared region and can reveal the chemical composition of the product for analysis, or detect moisture or foreign objects in the target.
With software processing technology developed by Vision Partners, from the hyperspectral images the different relevant parts (such as cartilage, fat, meat or buck) are classified with k-NN classification. And this in real-time so that the efficiency of the production machine is immediately available.
In addition, a possible cost reduction was investigated by transitioning from a hyper-spectral camera to a multi-spectral camera where the captured is which frequency intervals are most relevant for the different product categories.
Thanks to this hyperspectral classification in the 900-1700 NIR spectral region, in-line chemical composition of animal product can be determined, thus developing a more sustainable food chain.
"What is important is that the machine vision system integrator understands our processes and thus can develop a customer-specific product that fits our needs seamlessly."
