Research On The Extraction Method Of Maize Planting Distribution Information Based On The UAV Remote Sensing System

In order to achieve the precision agriculture, we must obtain farmland feature information precisely and quickly. This paper describes an information extraction method of jointing stage maize by using the high-resolution visible images which were obtained by the UAV(Unmanned Aerial Vehicle) remote sensing system. It included the following steps. First, selected the study area and obtained ground survey data. Second, using UAV remote sensing system to obtain the pictures of the study area, meanwhile mosaicking and preprocessing the pictures which included radiation correction, geometric correction and geographic registration the images. Finally, to dig out the method of extracting maize planting information by using the preprocessed images. Specifically, confirmed the methods of extracting maize planting distribution information, and evaluated it. The specific steps were described as follow:(1) This study adopted the T-EZ fixed-wing UAV to gain pictures in bayinnaoer city land of Inner Mongolia. And preprocessed the aerial images, which included mosaic, radiation correction, geometric correction and geographic registration.(2) Dug out the feature parameters and confirmed the method. To extract maize distribution information, we needed to research the feature parameters, and found out or created parameters to extract maize distribution information. Finally, mean of green, homogeneity of blue were chosen and TLVI was created.(3) The operation process of extracting Maize distribution information. To extract information, the ENVI5.1 software had been Used. Specific operation: Used the the feature parameters to layer the process of information extraction, used threshold method to complete the classification of each layer, used construction and application of mask to combine the classification results of each layer, that was the preliminary operation. After preliminary operation, we could find that the wheat plot and saplings plot still remain as small patches. We used ARCGIS to remove small patches by using shape and area parameters, then finished the preliminary extraction. After preliminary extraction we could find that in the extraction maize plot some maize information was separated. We could also use ARCGIS to keep the small patches, change the reserved small patches’ information and overlay them to the early extracted results to finish extracting the information of maize.(4) Precision evaluation and validation. Randomly selected three research areas to verify the accuracy of the research methods in this paper. Used visual interpreting results as standard and area extraction error as indicator to evaluate maize information extraction accuracy of the study area and other three validation area, the information extraction for error were 8%, 13.98%, 16.26% and 20%. Meanwhile,using error range to evaluate the four areas’ extraction precision,we can find the method this paper proposed achieved the overall level of the low altitude remote sensing classification.