The Research On Aerial Image Dehazing Fused With Mid-Wave Infrared Band

The development of remote sensing technology in recent years has led to great increase in the quantity of satellite images.Many remote sensing applications benefit from high resolution images due to the better representability of ground features.Since the instinct of the remote sensing,the remote sensing imaging is sensitive to atmospheric conditions during the data acquisition process.The existence of the haze severely degrade the quality of the remote sensing images,which hinders the effectiveness and availability of remote sensing data.Haze removal not only recovers a high quality of the visible images,but also improves the capability for practical usage.It is highly desired in remote sensing image applications.The innovation and the main research content are as follows:1.According to the status that there is not profound research in the mid-wave infrared area.We acquired the visible and mid-wave infrared orthophoto by our own airborne mid-wave infrared imaging system.Based on the analysis of mid-wave radiation characteristics,we explore the utility of aerial mid-wave infrared image for the land cover classification.2.We firstly review the prevailing dehazing algorithms in computer vision and their limitation in the application of remote sensing image dehazing.Then this article proposed a single aerial image dehazing algorithm based on the convolutional regression and hazy image degradation model.Firstly,using synthetic images series with different transmissions,we generate the training data to determine the coefficients of the convolutional net,which is used to estimate the transmission map of the hazy image.Then,the transmission map could be refined by guided filter to retrieve the hazefree image based on the dehazing model.Comparing to the existing approaches,our method works well in the estimation of the transmission map.3.Based on the different characteristic of visible and mid-wave electromagnetic wave(the wave length of mid-wave infrared is much longer than that of visible band),we proposed an airborne visible image dehazing algorithms fused with mid-wave infrared data.