Background Soil erosion has a great correlation with land use, and land use directly affects the form, spatial distribution and intensity of soil erosion. High resolution remote sensing image is the main information source of land use status survey, and it is the foundation and precondition to accurately grasp the dynamic soil erosion. The purpose of this study is to improve the interpretation accuracy of land use types for high-resolution images.

Methods Comprehensively supporting soil erosion dynamic monitoring and soil and water conservation planning in county-level, this paper took Huailai county of Hebei province as the study case, and land use information was obtained based on high-resolution image interpretation and field sampling survey. The land use results by the field sampling survey were taken as actual values, and the types (Cultivated land, garden plot, forest land, shrub land, grass land, and other), area and distribution of land use by remote sensing survey were compared with the sampling survey unit analyzed by land use change transfer matrix at the two levels of the whole county.

Results 1) The proportion of cultivated land in remote sensing survey and field sampling survey was 16% and 17.1% respectively, with a difference of 1.1%; the proportion of garden plot was 21.8% and 10.7% respectively, with a difference of 11.1%; the proportion of forest land was 16.5% and 8.9% respectively, with a difference of 7.6%; the proportion of shrub land was 14.6% and 33.9% respectively, with a difference of 19.3%; the proportion of grass land was 18.4% and 15.5% respectively, with a difference of 2.9%; the proportion of other land was 12.7% and 13.9%, with a difference of 1.2%. In addition to the far different proportion of garden plot and shrub land, the proportion of other land uses was quite similar, therefore, the land use structure obtained by remote sensing survey and field sampling survey were in good consistency. 2)The precision of cultivated land, garden plot and grass land was higher by remote sensing, 68.2% and 64.2%, respectively, and the forest land was in second, 78.6%, followed by shrub land and other land, 68.2% and 64.2%, respectively. 3)The interpretation accuracy of shrub land by remote sensing was low, only 68.2%, and it was because shrub land and grass land were easily confused so that led to misjudgment, the proportion of the shrub land misjudged to the grass land reached 30.5%.

Conclusions Both methods can be used as reliable means of obtaining land use information, moreover, high-resolution remote sensing images have high value in the investigation and mapping of large scale soil erosion, the misjudgment between the two of the shrub land and the grass land also needs to be paid more attention to the interpretation, the accuracy of interpretation can be improved from the analysis of texture and shade, the increase of field investigation and image processing technology.