Font Size: a A A

Study On Fish Density Optical Detection System

Posted on:2019-04-08Degree:MasterType:Thesis
Country:ChinaCandidate:M M JiangFull Text:PDF
GTID:2393330596963475Subject:Optical Engineering
Abstract/Summary:PDF Full Text Request
China is a big country in fishery and aquaculture and have abundant freshwater resources and freshwater fish resources.Modern fishery farming is inseparable from systematic management.Online monitoring systems based on pattern recognition,machine vision,artificial intelligence,etc.can exactly meet the needs of industrial aquaculture.The online video surveillance system analyzes and processes the video signals captured by the camera to realize the identification and localization of the farmed fish,and then obtain the size and quantity information of the fish.This system can not only reduce labor expenditures in fishery farming,but also respond to emergencies in a timely manner.Therefore,the research in this paper is of great significance.Based on machine vision,this article has conducted in-depth research on key technologies such as image acquisition,image processing and image transmission in fish intelligent monitoring.The main research content is as follows:First,the research background and significance of the research were introduced,the status of fishery development was briefly described,on this basis,the scheme of using computer vision technology to monitor the density of fish in pastures is analyzed.Secondly,the binocular stereo vision technology is used to calculate the size of the fish target,and a series of processes of double target fixation,distortion elimination,binocular correction and stereo matching are performed,and the three-dimensional position information of the target in the image is calculated according to the triangulation principle.According to the obtained three-dimensional coordinates of the head and tail position of the fish body,the length of the fish body is calculated,and the error value of the measurement result si about 15%.Again,the related theories of convolutional neural network and Faster R-CNN model are introduced.The fisher image data set is trained by the Faster R-CNN model,and the fish image is detected by the trained model.The number of fish is calculated by the class target.The experimental results show that the model has a good recognition effect,and the correct rate of the test result is close to 90%.Finally,a wireless video transmission system based on Raspberry Pi is designed,including a series of processes such as video format conversion encoding,WiFi wireless transmission,and PC display.The relevant technologies are introduced in detail.Finally,the entire system is debugged and operated,and the video stream can be output smoothly,achieving the desired effect.
Keywords/Search Tags:Fish density, Binocular stereo vision, Faster R-CNN, Video transmission
PDF Full Text Request
Related items