Auto-detection of Anisakid larvae in Cod Fillets by UV fluorescent imaging with OS-ELM

In this paper, one auto-detection scheme of Anisakid larvae in cod fillets is developed on the basis of online sequential extreme learning machine (OS-ELM) in a single hidden layer feedforward neural networks (SLFN). One UV fluorescent imaging system is first set up to collect and extract the typical image patches with and without Anisakid larvae inside the fish muscles, the UV fluorescent image patches are then fed into SLFN sequentially to learn how to nondestructively identify the parasites in real-time, particularly for a growing size of the training set with new observations arrived again and again. It has been shown in the simulation experiments that the developed nondestructive approach could get online auto-detection performance in both good accuracy and efficiency during the test, even for those Anisakid larvae deeply embedded in the cod fillets.