A novel quality control model of rainfall estimation with videos – A survey based on multi-surveillance cameras.

• Accuracy enhancement for current surveillance camera-based rainfall estimation. • A quality control model for rainfall data obtained by cameras. • A solution to the critical problems of regional rainfall data production. • A novel high-resolution, low-cost, ground-level rainfall monitoring network. The widespread use of surveillance cameras has become an emerging means for rainfall observations. With the advantages of high spatial–temporal resolution, rainfall information obtained from surveillance videos is highly suitable for meteorological-related research and has bright prospects. However, due to the complex and variable monitoring scenarios, the quality of the rainfall data estimated by each camera is always inconsistent, resulting in low practical value. Dense ground-level surveillance cameras have temporal and spatial correlations that can be used to improve the accuracy of rainfall estimation through mutual verification. In this study, we first introduce camera parameters to refine the spatial volume of rainfall (SVoR) 1 1 Most acronyms used in this study appear in the Appendix 1. perceived by cameras to improve the accuracy of rainfall intensity (RI) estimation. Next, a novel quality control (QC) model of rainfall estimation with multi-surveillance camera collaboration that takes the rainfall observations of all cameras as input is proposed. (i) We build a reliability evaluation (RE) model for the estimation of RI in accordance with raindrop imagery features to provide a reference for the subsequent correction of RI estimation; (ii) inspired by the first law of geography (Tobler, 1970), we then construct a spatial–temporal consistency filter and a situation consistency filter by using the spatial–temporal constraints between cameras to coarsely evaluate the RI values; and (iii) the correlation between cameras is calculated based on the fuzzy method to further build a correlation filter for the fine-grained correction of RI values. Experiments show that our method can effectively eliminate RI outliers and improve the accuracy and reliability of rainfall estimation results. Moreover, our method is highly suitable for heavy and violent rainfall application scenarios and can provide high-resolution rainfall data support for flooding warnings and simulations in urban areas. [ABSTRACT FROM AUTHOR]