Detection Of Anisakid Larvae In Marine Fish Fillets By Imaging Technique

Anisakid larvae，parasitizing marine animal all over the world, causes anisakiasis.The disease mainly results from consuming uncooked fish or fish product infected bythe third-phase Anisakid larvae. Anisakiasis cases were reported in more than20countries, including Japan, Netherlands, England, France, Germany, and Spain. Withthe prevalence of uncooked marine food product consuming, the possibility ofAnisakiasis will increase. Therefore, the potential crisis arouses widely attention.At present, candling is still used in aquatic products processing plants to detectAnisakid larvae, which is inaccuracy and labor consumptive. Although some accuratemethods were reported, none of them fits Anisakid larvae detection on-site. In thisstudy, hyperspectral imaging and UV fluorescent imaging were developed to detectAnisakid larvae in fish. Main results were listed as following:1. The hyperspectral imaging technology was applied in Anisakid larvae detection.In this part, two methods were developed: band math and principal componentanalysis (PCA). Image of550nm and800nm were selected as characterized bands forbond math according to the difference between muscle and Anisakid larvae inspectrum. Image of550nm segmented by （0.15，0.17）minus image of800nm. Theresult image was further segmented by （-0.08，-0.06）and then processed by7×7median filtering. The position of Anisakid larvae was showed in the result imagewithout any interference. In order to increase the efficiency of detection, PCA wasapplied in processing600hyperspectral images, from400nm to1000nm. The resultshowed that hyperspectal images processed by PCA within region of interestdistinctly revealed the position of Anisakid larvae.2. UV fluorescent image was selected as method for further image process aftercompared with other imaging technology. UV imaging was established according tothe principle that Anisakid larvae emit visual fluorescence when excited by UV lightof360nm. On the plat form of ENVI4.7, PCA was used to process UV fluorescentimage of cod fillets. The second principle component was segmented by [-320,-280].The image was then fused with Red component of crude UV fluorescent image so thata reference of cod fillets was showed.3. Based on programming in Matlab, automatic UV fluorescent imaging was established which leads to results of quantitative analysis. Image process methodbased on gray value was applied in Anisakid larvae, and the detection rate was41.67%. The morphologic characteristics of Anisakid larva were studied to improvethe method. And the detection rate increased to61.11%. By combining the twomethods, an efficient method was established with detection rate86.11%. And thismethod was verified by samples of four different batches. The overall detection was83.89%and coefficient of variation is1.31%. There was no significant differencebetween all the experiments (P>0.05).4. A device for UV fluorescent image was made according to parameteroptimization. In single factor experiment of height of UV illuminant,5levels of25cm,45cm,65cm,85cm, and105cm were studied. The detection rates were38.05%,72.23%,71.65%,74.97%,29.52%, respectively. Five different conveyor belts weretested as background. Conveyor belt of white smooth, green smooth, black smooth,black diamond, black quadrate and black wood grained presented detection rate25.30%,67.40%,80.13%,89.29%,83.66%,84.31%,respectively. Digital Single LensReflex cannot be replaced by common computer camera, and screenshot of video isunqualified for detection. The method was applied in Pollack, Atlantic cod, butterflyfish, and salmon, leading to detection rate85.50%,83.23%,82.13%,85.78%,respectively.