Research On Crop Area Extraction Method Based On GF-1 Remote Sensing Image

In the third national agricultural census work as the background,combined with the problems existing in the current crop area measurement method,this paper tries to seek a comprehensive,accurate,rapid and effective access to large area crop spatial distribution and the method of planting area of information,to develop a plan of the agricultural development of science to provide a reasonable data basis.In order to obtain more effective research conclusions,this paper USES the three research area images and phenological data as the basic data.Because of GF-1 panchromatic and multi-spectral data can be synchronized and the precision of matching is poorer,therefore,choose fusion is correct,before registration of remote sensing data quickly handle way 4 band orthogonal projection as results in the study area are obtained.Based on the research area,artificial visual interpretation method,image cl assification based on pixel and object oriented image classification are used to extract remote sensing of crop area in the research area.In the process of extr acting crop area by object oriented classification,the optimal segmentation para meter setting determines the accuracy of target information extraction.Firstly,t he optimal segmentation scale of images was determined by means of the mea n standard deviation of brightness,and the optimal parameters of shape factor and compactness were selected by the maximum area method.From different a ngles,it is possible to choose the characteristics of various types of ground ob jects and match the feature combination to set up multiple multi-scale segment ation rule sets to classify images.In addition,the results of sample survey data and artificial visual interpretation are used for reference,and error matrix analysis and area interpretation accuracy and measurement time comparison are carried out for the results of target information extraction.Object-oriented image classification and artificial visual interpretation method to obtain the results of the spatial distribution,the classification accuracy and interpretation accuracy,the overall accuracy can reach more than 90%,than the image classification based on pixels,the acquired results precision is much higher,can meet the demand of actual production application.In the work efficiency,with the increase of crop area,the advantage of object oriented image classification in measuring speed is obvious.To sum up,using the classification results of the object-oriented image classification is closer to the real surface features,and overcomes the phenomenon of "salt and pepper,this method can be used as a quick access to a wide range of crop area of information a good choice.