Study On Object Oriented Vegetation Information Extraction From Urban Visible Light Aerial Image

Vegetation is an important part of the community environment and has important ecological functions such as dust retention,alleviating the greenhouse effect,and improving air quality.Remote sensing images have become a common method for extracting and monitoring regional vegetation information.UAV aerial images generally only contain three bands of red,green,and blue,which can obtain a large range of ground information in a short period of time.In the classification and rapid update of information It has great advantages and has become the main data source for rapid and small-scale information extraction.In urban areas,affected by tall features(such as office buildings,buildings,etc.),part of the vegetation information will be covered by the shaded area.The existence of the shaded area will affect the accuracy of vegetation information extraction.Therefore,this paper studies from two parts: the non-shaded area and the shaded area.The vegetation information extraction method integrates the vegetation information extraction results in the non-shaded areas and the vegetation extraction results in the non-shaded areas to improve the vegetation information extraction accuracy of the urban visible light aerial image as a whole.The extraction of vegetation information in non-shaded areas currently mainly uses visible light vegetation index.The visible light vegetation index is effective in extracting vegetation information in large vegetation areas or simple ground features,but the effect of extracting vegetation information in complex backgrounds is not ideal;shadow areas The extraction of vegetation information needs to solve two problems,namely,shadow detection and shadow elimination.Based on shadow restoration,an appropriate method is selected to extract vegetation information in the shadow area;this paper focuses on the above-mentioned problems in the extraction of vegetation information.The specific research results are as follows :(1)Extraction of vegetation information in non-shaded areas.Aiming at the problem of low discrimination between vegetation and water bodies,vegetation and shadows when extracting vegetation information from the visible light vegetation index,a method of combining the knowledge rules of the visible light vegetation index,brightness value and texture variance information with the object as a unit is proposed.The experimental results show the method in this paper.The effect of distinguishing vegetation and water bodies,vegetation and shadows is better.The overall accuracy of vegetation extraction exceeds 90%,the kappa coefficient is close to 0.8,and the corresponding misclassification and omission accuracy are lower,and the vegetation extraction results are more complete.(2)Shadow detection.This paper is based on the color invariant model method for shadow detection.Aiming at the problem of color space that water and blue objects are easily detected as shadow areas,an object-oriented RGB-C1C2C3 method for shadow detection is proposed.The experiment proves that object-oriented RGB-The C1C2C3 method has a better shadow extraction effect,and will not misdetect water and bluish objects as shadow areas.The shadow accuracy and overall accuracy can reach 90%,and the shadow area extraction effect is better.(3)Shadow elimination.This paper is based on the color constancy theory to eliminate shadows,and conducts shadow elimination experiments on several commonly used color constancy algorithms in the study area.The results prove that the Shades of Gray algorithm has a better shadow elimination effect.After using this method to eliminate the shadows,the spectrum of the vegetation area is inter-spectral.The relationship information is better restored,which is beneficial to the extraction of vegetation information in the shadow area.(4)Extraction of vegetation information in shaded areas.On the basis of shadow detection and elimination,the spectral response curve of the ground objects after shadow elimination is analyzed,and the visible light vegetation index VDVI is selected to extract the vegetation information in the shadow area at the pixel level and the object level respectively.The experiment proves that the object-oriented visible light vegetation index method is adopted.The effect of extracting the vegetation information in the shadow area is better,and it can extract the vegetation information in the shadow area to a certain extent.