A Multilevel Stratified Spatial Sampling Approach for the Quality Assessment of Remote-Sensing-Derived Products

With the advent of new remote sensors, the number and volume of remote-sensing data and its derived products, which are regarded as typical big data, have grown exponentially. However, it remains a significant challenge to evaluate the quality of these big remote-sensing data and their derived products. Spatial sampling is necessary for the quality assessment of remote-sensing data and the derived products. This paper proposes an approach of multilevel stratified spatial sampling for the quality assessment of remote-sensing-derived products, with the aim of resolving the issue of the quality inspection of remote sensing big data and the derived products. The proposed multilevel stratified strategy: 1) makes full use of the prior knowledge of the data set; 2) selects a sample subset to get an unbiased estimator for the quality; 3) aims to acquire knowledge about the entire product; and 4) makes an evaluation based on statistical inference. The proposed method improves the sampling accuracy without increasing the inspection cost, and the whole procedure is repeatable and easily adopted for the quality inspection of remote-sensing-derived products and other geospatial data.