Abstract: To solve the problem of ephemeral lights source and background noise for NPP/VⅡRS images, the multi-temporal nighttime light remote sensing images calibration method is proposed based on the radiation characteristics of monthly images. The radiation difference of the images is amplified by the spike characterization. The high-pass convolution operation is used to distinguish the stable pixel from the abnormal pixel. On the basis of this, an adaptive threshold segmentation method called the maximum entropy is used to resegment the aliasing noise at the convolution boundary. Through analysis of experimental results, the abnormal fluctuation of the original image is corrected, and the abnormal pixels are eliminated. After calibration, the correlation between VⅡRS data and quarterly economic data is stronger. The lack of micro-scale research on nighttime light remote sensing data is compensated, and the dynamic assessment for socio-economic data is provided with new ways.

Key words: NPP/VIIRS, image calibration, maximum entropy thresholds, result evaluation, nighttime light remote sensing