Research On The Methods Of Automatic Shellfish Identification And Location Based On Machine Vision

Shellfish production sorting required a lot of manpower. The development of shellfishindustry is hampered seriously due to the poor working environment, low productivity, laborshortages, rising wages and so on. Efficient and accurate technology is needed to complete thework of sorting shellfish with increasing requirements of production efficiency and aquaticproducts quality. This topic is a sub-project of cooperative research project with enterpriseswhich is named as ‘equipment research and development for automatic shellfish sorting’. Byextracting the robustness shellfish classification features in the images gathering by camera fromthe conveyor belt, the shellfishes are classified and located by applying the intelligent algorithm.The result can help the real classification for shellfish using mechanism system. The main workpresented in the dissertation can be summarized as follows:Scallops image processing and edge detection method were researched. First, the targetwas highlighted by preprocessing the image, and then the canny operator was used to extract thetarget edge information.Automatic identification was realized by using the BP neural network which trained by theedge feature of scallop extracted from image. Experiment proved that the proposed method weremore efficient comparing with HU moment invariant features method.The methods of scallops’ identification using fuzzy theory were study in order to improvethe recognition efficiency. Fuzzy recognition systems were established for scallops identificationusing the classification fuzzy rules established according to the extracted features from the image.The experimental result proved that the proposed method was effective.By the method of neural network, mapping relationship between image coordinates and theactual work space were found to realize the aim of scallops positioning. The input of the neuralnetwork was the center coordinates of scallop image, and the output was the coordinates of workspace. The finding of the mapping relationship laid a foundation for controlling the mechanicalsystems to complete the classification.