Pin Bone Detection: X-ray Bone Detection in Salmon
Pink Salmon has some of the smallest and hardest known fish bone that could be as small as small as 0,1 mm in diameter. This is why pin bone detection is a very challenging quality assurance task. The samples of Salmon fillets have been tested at varying speeds and using InnospeXion’s HYMCIS Fish SL & DL and have been analyzed using our Fish Bone Detection Software to determine the capability and limitations of our low-energy X-ray systems.
Pin bone detection is very important, not least in the Salmon case, where the ability to remove the bones in the pre-rigor condition (i.e., shortly after slaughtering) may represent a large value since the fillet can be sold e.g. for sushi, etc.
The challenge is, that bone removal in the pre-rigor condition is very difficult due to attachment of the bones. Therefore, there is an important need of determining if some bones were not (completely) removed in the process.
Fig. 0 Low-energy X-ray image of salmon
In order for fish bone detection software to correctly analyze the digital radiographs (X-Ray images in .tiff format) a reasonable contrast is needed between the bones and the meat. This contrast is further enhanced using image analysis algorithms and exploited in detecting the fish bones. Ultimately the product with bones can either be rejected and sorted out manually or the bones identified and marked by X-Ray machines can be cut through.
There are number of parameters that could be adjusted to obtain the best contrast including the power settings (voltage and current) of X-Ray source and speed of conveyor, which determines the speed at which objects are scanned by the detector.
Fig.1 Fresh pink salmon fillet entering our HYMCIS Fish SL system for bone detection.
Salmon fish fillets have been scanned at speeds of 3 m/min, 9 m/min, 15 m/min and 26 m/min. At low speed the object is exposed to X-Rays for the longer duration of time and good contrast is obtained but as the speed of conveyer is increased, the duration of exposure to X-Rays is decreased, and image quality is compromised.
Fig.2 Graphic illustrating the pixel size difference between our X-ray technology and the standard one.
In the case that each pixel is sufficiently sensitive so the signal greatly exceed the noise level, small pixels enables a much better definition.
This is especially important for fish bones, since critical bones may be as small as 0.35 mm.
To the right is compared the size in the X-ray image of a 0.4 x 2.4 mm bone imaged using standard technology (6 pixels), compared to InnospeXions technology (96 pixels). Obviously, automatic detection by software is much easier when 96 pixels make up a bone compared to 6, and this makes our technology perfect for pin bone detection.
Fig.3 Graphic illustrating the attenuation differences between meat and bones when exposed to low-energy technology compared to standard X-ray technology.
The X-ray attenuation of materials depends e.g. on the atomic composition . The difference between e.g. meat and bone is exponential. At low X-ray energy, the differences are very pronounced (i.e. curve distance at the leftmost arrows, cp. To the small contrast at the rightmost arrows). Imaging at low energy therefore gives a much higher contrast and thus greatly improves the detection capability, therefore performing superior to standard technology during pin bone detection.
Fig.4 Pin bone detection in sample 1 using our low-energy inspection technology.
Fig.5 Pin bone detection in sample 2 using our low-energy inspection technology.
Fig.6 Pin bone detection in sample 3 using our low-energy inspection technology.