Detection Of The Deep-Sea Plankton Community In Marine Ecosystem Based On Optical Flow

With the development of society,human are developing and using more and more various resources.With the gradual depletion of non-renewable resources,mankind has shifted the focus of obtaining resources to the deep sea.In addition to deep-sea petroleum gas,natural gas hydrate and other non-biological energy sources,there are also a large number of unknown biological resources in the seafloor.Creatures are ubiquitous in the ocean,and different environments have created biological communities with different types and structures.The distribution pattern of various ocean organisms interacts and develops in harmony with the environment.This thesis will study the detection of deep-sea plankton.There are many species and large quantities of plankton,and it is one of the main members of deep-sea creatures.Its research is of great significance to fishery production and basic theories of marine science.Therefore,how to study deep sea plankton quickly and effectively has become a hotspot for researchers.In the deep sea environment,the white sediments are often confused with plankton on the seafloor.The complex background and uneven illumination increase the difficulty of plankton detection.In consequence,the requirements for object detection algorithms based on visual features are very strict.In the thesis,the detection of plankton in the deep-sea environment is mainly studied by improving the HS optical flow algorithm,and at the same time,the classic object detection algorithms and theories are introduced.Aiming at the complex background,uneven illumination and background movement in plankton detection,and based on some classic theories and algorithms,an improved HS optical flow method is proposed,which is applied to plankton detection in deep-sea environments.The main research content of this thesis is divided into the following points:(1)The algorithms of plankton detection and target detection based on machine vision and image processing are studied,and several classical algorithms and research theories are briefly summarized.(2)An improved optical flow method for plankton detection is proposed and proved theoretically.When solving the optical flow,the gray gradient feature of the image is extracted,and the particularity of the pixel gradient feature at the location of plankton is found,and this feature is used for the object detection of plankton.(3)Propose adaptive threshold method in plankton detection algorithm to the complex background and uneven illumination.According to the different image characteristics in video data,the background type of the image is determined,so as to adaptively select the best threshold combination.(4)The efficiency of plankton detection is greatly improved by using the methods of interframe desampling and intrapixel desampling of video sequences.The experiment shows that the intrapixel desampling method can improve the detection efficiency while maintaining a high accuracy and recall rate.Through the above four aspects of research and innovation,this paper has made a multi-directional and multi-level detection and research of plankton in the video sequence under the deep-sea environment,which lays a foundation for further exploration of deep-sea plankton.