Study On Tuna Shoal Identification System Based On Remote Sensing Image Of Unmanned Aerial Vehicle

Tuna seine fishery is one of the fishing methods with the highest requirements for technology and equipment in the world's pelagic fishery.The level of fishing technology and equipment of tuna seine is embodiment of the comprehensive strength for oceanic fishing countries.Since 2001,Chinese Mainland has been involved in tuna purse-seine fishery in the Western and Central Pacific Ocean.In the following years,the tuna seine catch yield has been continuously increasing,and the gap with developed countries such as the European Union,the United States and Japan has gradually narrowed.At present,tuna purse-seine fishery are gradually shifting from FAD fishing to free-fish fishing as the tuna Regional Fishery Management Organizations(t RFMOs)are increasingly strict with the use of fishing aggregation devices(FADs).Fish searching of tuna is an important part of tuna seine fishery and its technical level has an important impact on the fishing efficiency and economic benefits.Aiming at the problem of high cost and high risk for helicopter fish searching in tuna purse-seine fishery,this study focused on the development of image recognition system for tuna fish intelligent detection.The remote sensing image of the sea surface is obtained with wide visual angle and high resolution through the over-the-horizon high-definition cameras on UAV,to effectively reduce the production cost and operation safety risk.(1)According to the principle of "acquisition-processing-recognition",the technical framework of the computer intelligent recognition of tuna school image was established based on image acquisition,image preprocessing,image feature extraction,feature library construction,image matching and recognition.(2)The average coefficient method(W1=0.299,W2=0.587,W3=0.144)can be used to convert RGB three-channel color image into one-channel grayscale image.Non-local Means method can be effectively employed for noise reduction through the ubiquitous redundant information in natural images.(3)It is showed that the ORB algorithm has an obvious comparative advantage inimage recognition of tuna shoals for the main indicators,such as operation speed,rotation robustness,fuzzy transform robustness,brightness transform robustness and feature point proportion in feature area.(4)An image recognition model of tuna shoal was constructed based on the Brute-force matcher matching method and the KNN nearest neighbor algorithm.It is showed that the recognition accuracy of the characteristic points of tuna fish is about60% by the computer simulation.