Detection Of Fish Bones In Cod Fish Fillets Based On UV Fluorescence Technology

Fish bone detection is a very important aspect in the productive process of boneless fish products, especially for the higher value-added seafood. In most factories such a performance is implemented by manual operation based on touching and vision sense under daylight lamps, which with greater labor intensity and low efficiency. Moreover, the increasing labor cost also makes it more and more difficult to exploit such a technique even in many developing countries. Therefore, it is imperative to develop some innovative methods with better efficiency and higher intelligentialize degree for fish bones detection.In this paper, the main experimental material is cod fillets with fish bones. The UV fluorescence properties of fish bones, UV fluorescence imaging technology and computer image recognition technology were applied in the study of fish bones detection. The efficiency of the new methods proposed in this study were established, and the main results were listed as following:1. Result of fluorescence characteristics research of cod fish bones and muscle showed that the optimum excitation wavelength of fish bones and muscle were 320 nm and 358 nm, respectively corresponding optimum emission wavelength were 515 nm and 580 nm. For bones with different size, no significant difference was investigated in the fluorescence characteristics. Under the irradiation of UV-A light (320 nm-380 nm), cod fish bones could emit significant fluorescence with intensity of 1500-2500 equipment units that is much higher than fish muscle’s.2. Based on the results of the fluorescence characteristics, the UV fluorescence assisted-candling was for the first time developed for the detection of bones in fishery products. Using cod fillets as samples, the detection ratio of this technique was calculated as 90.86%, significantly higher than that of traditional candling under daylight (76.78%). Moreover, the working efficiency of the new technique was increased about 26% over the traditional method.3. On the basis of preliminary studies, the UV fluorescence image of cod fish fillet were collected by image acquisition device built and improved by ourselves. Making full use of the fluorescence characteristics of fish bone in the UV-A light, the fishbone detection method of UV fluorescence imaging based on principal component analysis (PCA) was built at the based platform of image processing software. The purpose of separation of fish bones and fish muscle background initially realized when the second main component image was segmented with fixed threshold value in particular Digital Number range.4. In order to overcome the fluorescence intensity fluctuation of fish bones between different batches of image samples. The fish bone image recognition algorithms based on linear discriminant analysis (LDA) was designed and established which has batter adaptability for complex image samples. To ensure the implementation of the algorithm, the true and false samples of training set were collected that each contained 300 image patches (size of 40 pixels×40 pixels). Here the Euclidean distance summation was taken as a simple computation mode for the decision making of the test image patch detection. Testing by 120 image patches from four batches of test set, the accuracy of LDA algorithm was confirmed as about 65%.