Linear Target Detection Algorithm For Large Scale Remote Sensing Image

With the rapid development of remote sensing technology,the resolution of remote sensing images returned by various remote sensing satellites is getting higher and higher.For the reason that remote sensing image carries more and more information,it is possible to realize for identifying typical targets from the remote sensing image.Airport area is an important large artificial target among them,and discerning it has always been one of the important research directions in the field of remote sensing image recognition.Due to its unique function that the airport area is large,the airport area is far away from the city center.In order to accurately discern the airport area from the remote sensing image of large scene,based on the remote sensing image pretreatment method and according to the texture features of the airport area,the paper adopts the improved maximum inter-class variance method to segment the remote sensing image and extract the suspected airport area.The Res Net50 network is as the basic network,and the fine-turning mechanism is introduced to establish the residual network,which was used to screen the suspected airport areas.The improved edge marshalling algorithm is used to detect the runway position of the area identified as an airport.The main research contents are as follows:1.Use the improved maximum between-class variance method to segment large remote sensing images.Based on the background analysis of the remote sensing image and the characteristics of the target in the airport area,the remote sensing image is segmented by the improved maximum variance method,and the suspected airport area is extracted by the minimum external moment algorithm.2.Build a residual network to filtrate airport areas based on fine-tuning mechanism.The background of a large amount of remote sensing images is complex,and a fine-tuning mechanism based residual network is designed for the regions similar to airports,including but not limited to highways,bridges,coastlines and large factories.The network can effectively identify the airport area and eliminate interference for subsequent airport runway detection.3.Improve the edge marshalling algorithm.In view of the problem of offsets and over connections in Hough transformation,a runway recognition algorithm for marshalling along the airport runway edge was designed.This algorithm uses probability Hough to estimate the airport runway direction,and marshals the airport area edge along the airport runway direction.The improved edge marshalling algorithm can effectively reduce the number of marshalling and the efficiency of linear connection complexity.By dwelling on the algorithm of airport area recognition and runway detection in a lot of remote sensing images,this dissertation summarizes the research content and proposes the next research direction.