Crop area ground sample survey using Google Earth image-aided

Abstract: By using Google Earth (GE) image revised by differential global positioning system (DGPS) actual measurement points, this paper conducts a ground sample survey of crop planting areas, and compares the difference in survey accuracy and efficiency between this method and the method completely using GPS field measurement. The study area is the Agricultural High-tech Industrial Park of Chinese Academy of Agricultural Sciences (Wanzhuang) and its surrounding area with the area of 3.1 km × 2.0 km. The paper defines the data from the different GE image sources. The images downloaded based on GE Client COM API programming are defined as A-level data, the images revised by online GE images are defined as B-level data, and the images revised by DGPS actual measurement points are defined as C-level data. Compared with the checkpoints of DGPS actual measurement, A-level data of the GE images with spatial resolution of over 0.5 m have a mean square error of 232.7 m in X and Y directions, and for B-level and C-level data it is 5.4 m and 1.0 m, respectively. The B-level data meet the requirement that "The mean square error in planimetric position of 1:25000 should be no more than 8.75 m", and the C-level data meet the demand that "the mean square error in planimetric position of 1:10000 flat ground should be no more than 3.5 m", which are specified in the Digital Aerophotogrammetry Aerial Trigonometric Survey Specifications. Choosing the samples with 3 structure levels, i.e. simple, medium and complex level in the Langfang survey area, the area measurement accuracy of B-level and C-level data is measured, and the average errors are 0.108% and 0.018% respectively through the comparison with DGPS actual measurement areas. The larger the crop area, the higher the accuracy of the measurement. The survey meets the accuracy requirement of large scale ground sample survey. With respect to GE online coordinate, the average minimal mean square error of B-level data is 0.5 m, and the average maximum mean square error is 1.0 m. The average mean square error in XY direction is 0.5 m, and its relative accuracy meets the requirement of mapping specification of 1:2 000. If only the ground sample changing trend among different years is described without considering the absolute calibration accuracy, applying the B-level data GE images has a stronger charting ability. In order to evaluate ground survey efficiency based on this method, it is applied in the winter wheat ground survey of central plain area of Shaanxi Province in 2015, and the contrast test is conducted with conventional GPS field measurement method. The result of the contrast test shows that, the GE image-aided survey has reduced the time by 64.2% and the walking distance by 82.5% based on GPS measurement method. As a result, the survey efficiency is improved by 73.3%. Ground sample survey is an important link for crop area survey with remote sensing monitoring. Usually, it is achieved by using DGPS field measurement, which is a time and effort consuming job in a large scale crop area survey. The GE image-aided ground sample survey method proposed in this paper can save the time and labor. With the use of more and more available high spatial resolution satellite images and aerial survey images, this method will have a broader application prospect.