Font Size: a A A

Recognition And Detection Method Of Raft Aquaculture Area Based On Deep Learning

Posted on:2020-12-29Degree:MasterType:Thesis
Country:ChinaCandidate:Y ZhongFull Text:PDF
GTID:2492306305486364Subject:Computer technology
Abstract/Summary:PDF Full Text Request
In recent years,with the rapid development of marine aquaculture industry in China,it is becoming more and more important to make statistics on marine aquaculture information.The traditional field investigation is not suitable for large-scale development because it is very time-consuming and man-made.The extraction of marine aquaculture information through remote sensing technology has become one of the important research directions in the marine environment.When extracting the aquaculture area from remote sensing images,the visual interpretation method is time consuming and laborious,while the traditional machine learning method has poor generalization ability and fitting ability to remote sensing data.In order to overcome these two problems,this paper proposes a method for automatically extracting the raft aquaculture area based on the fully convolutional neural network.The method belongs to the pixel level classification,which uses multiple convolution layers,pooling layer and nonlinear ReLU function to form a deep network.The method will extract the nonlinear and unchanging deep features that can effectively improving the accuracy of the raft aquaculture identification in remote sensing image.At the same time,the L2 regularization and dropout strategy will be added to the neural network model to avoid overfitting.The research area of the raft aquaculture will select the remote sensing of Lianyungang offshore area,which was taken using the GF-1 satellite.The experimental results show that the method proposed in this paper effectively identifies and extracts the aquaculture area,and the F1 score of the identified raft aquaculture area reaches the highest 83.9%.The method of automatically extracting the aquaculture area based on the full convolutional neural network belongs to the end-to-end classification method.In practical applications,it is necessary to estimate the area and quantity of the raft aquaculture areas.The target detection method is used to detect the raft aquaculture areas of remote sensing image,effectively detect the area of aquaculture and count the quantity of aquaculture areas.In the experiment,the detection of the aquaculture area was carried out in the experimental study area using the You Only Look Once target detection method.The experimental results show that the method has an overall F1 score of 97.4%in the test cockroach culture area.
Keywords/Search Tags:High resolution remote sensing image, Marine aquaculture, Feature extraction, Full convolutional neural network, Classification, Target detection, You Only Look Once(YOLO)
PDF Full Text Request
Related items