DETECTION OF FISHBONES IN FISH FLOSS PRODUCTS USING CURVELET TRANSFORM BASED SQUARE-RING BAND-HIGHPASS FILTERING TECHNIQUES

摘要

Both fish and fish-related products are common foodstuffs for many cultures around the world. Fish floss is a preserved fish product made from finely chopped or mashed fish, which are boiled in seasonings and then stir-fried until the mass is arid and pulverous. During consumption, fishbones constitute the most frequently ingested extraneous objects. In the making of commercial fish floss, fishbone inspection is currently conducted by experienced inspectors who use their sense of touch and sight to detect bones. This reliance on human inspection leaves considerable room for error, which poses a consumer health risk as fishbones may cause physical harm when ingested. This study seeks to replace the human inspectors in this process by exploiting a frequency filtering method based on computer vision to inspect fishbones in fish floss products. The proposed method applies the curvelet transform with square-ring band-highpass energy filtering to removing the random patterns of background and to deleting the angle direction of background texture. The approximated and partially detailed components regarding fishbones and uniform background are preserved in the low and medium frequency bands; the filtered image is then inversely converted to spatial domain. In the reconstructed image, the background random texture is attenuated and the fishbone areas are enhanced. Experimental outcomes show that the suggested inspection method can effectively identify and locate the visible fishbones, and outperforms the existing methods of fishbone inspection on the bumpy, irregularly textured surfaces of fish floss products.