At-Sensor Radiometric Correction of a Multispectral Camera (RedEdge) for sUAS Vegetation Mapping

Rapid advancement of drone technology enables small unmanned aircraft systems (sUAS) for quantitative applications in public and private sectors. The drone-mounted 5-band MicaSense RedEdge cameras, for example, have been popularly adopted in the agroindustry for assessment of crop healthiness. The camera extracts surface reflectance by referring to a pre-calibrated reflectance panel (CRP). This study tests the performance of a Matrace100/RedEdge-M camera in extracting surface reflectance orthoimages. Exploring multiple flights and field experiments, an at-sensor radiometric correction model was developed that integrated the default CRP and a Downwelling Light Sensor (DLS). Results at three vegetated sites reveal that the current CRP-only RedEdge-M correction procedure works fine except the NIR band, and the performance is less stable on cloudy days affected by sun diurnal, weather, and ground variations. The proposed radiometric correction model effectively reduces these local impacts to the extracted surface reflectance. Results also reveal that the Normalized Difference Vegetation Index (NDVI) from the RedEdge orthoimage is prone to overestimation and saturation in vegetated fields. Taking advantage of the camera’s red edge band centered at 717 nm, this study proposes a red edge NDVI (ReNDVI). The non-vegetation can be easily excluded with ReNDVI < 0.1. For vegetation, the ReNDVI provides reasonable values in a wider histogram than NDVI. It could be better applied to assess vegetation healthiness across the site.

Design of Process Products Based on Image Processing Multimode Interaction

With the development of modern science and technology and more and more image processing systems, related technologies are becoming more and more complex. The application of image processing technology can be seen in various fields of society, such as medical field, aerospace field and life, and entertainment field. Due to the increasing amount of information on the picture, the requirements for the speed and clarity of image processing are also increasing. The existence of various external factors will lead to the production of image products and objects between the error and distortion problems. In order to make the process product design more authentic and reliable, this paper studies the process product design based on image processing multimode interaction. It uses radiometric correction and geometric correction to process distorted images and uses GPU parallel computing technology to accelerate the correction process. In this paper, this technology is applied to the visual recognition of welding robot, and the experiment shows that the product produced by the image processed by this module can obviously reduce the error.

A fast radiometric correction method for Sentinel-2 satellite images

Purpose Radiometric calibration is a method that estimates the reflection of the target from the measured input radiation. The purpose of this study is to radiometrically calibrate three spectral bands of Sentinel-2A, including green, red and infrared. For this purpose, Landsat-8 OLI data are used. Because they have bands with the same wavelength range and they have the same structure. As a result, Landsat-8 OLI is appropriate for relative radiometric calibration. Design/methodology/approach The method used in this study is radiometric calibration uncorrected data from a sensor with corrected data from another sensor. Also, another aim of this study is a comparison between radiometric correction data and data that, in addition to radiometric correction, has been sharpened with panchromatic data. In this method, both of them have been used for radiometric calibration. Calibration coefficients have been obtained using the first-order polynomial equation. Findings This study showed that the corrected data has more valid answers than corrected and sharpened data. This method studied three land-cover types, including soil, water and vegetation, which it obtained the most accurate coefficients of calibration for soil class because R-square in all three bands was above 88%, and the root mean square error in all three bands was below 0.01. In the case of water and vegetation classes, only results of red and infrared bands were suitable. Originality/value For validating this method, the radiometric correction module of SNAP software was used. According to the results, the coefficient of radiometric calibration of the Landsat-8 sensor was very close to the coefficients obtained from the corrected data by SNAP.

Matrix SegNet: A Practical Deep Learning Framework for Landslide Mapping from Images of Different Areas with Different Spatial Resolutions

Practical landslide inventory maps covering large-scale areas are essential in emergency response and geohazard analysis. Recently proposed techniques in landslide detection generally focused on landslides in pure vegetation backgrounds and image radiometric correction. There are still challenges in regard to robust methods that automatically detect landslides from images with multiple platforms and without radiometric correction. It is a significant issue in practical application. In order to detect landslides from images over different large-scale areas with different spatial resolutions, this paper proposes a two-branch Matrix SegNet to semantically segment input images by change detection. The Matrix SegNet learns landslide features in multiple scales and aspect ratios. The pre- and post- event images are captured directly from Google Earth, without radiometric correction. To evaluate the proposed framework, we conducted landslide detection in four study areas with two different spatial resolutions. Moreover, two other widely used frameworks: U-Net and SegNet, were adapted to detect landslides via the same data by change detection. The experiments show that our model improves the performance largely in terms of recall, precision, F1-score, and IOU. It is a good starting point to develop a practical, deep learning landslide detection framework for large scale application, using images from different areas, with different spatial resolutions.

AUTOMATED REFLECTANCE TARGET DETECTION FOR AUTOMATED VICARIOUS RADIOMETRIC CORRECTION OF UAV IMAGES

Recently, with increasing use of unmanned aerial vehicle (UAV), radiometric calibration of UAV images has become an important pre-processing step for application such as vegetation mapping, crop field monitoring, etc. In order to obtain accurate spectral reflectance, some UAVs measure irradiance at the time of image acquisition. However, most of UAV systems do not have such irradiance sensors. In these cases, vicarious radiometric correction method has to be used. Digital numbers (DNs) of imaged ground reflectance targets are measured and spectral reflectance is acquired from with known reflectance values of the targets. For automated vicarious calibration, a technique for automatically detecting image location of ground reflectance targets has been developed. In this study, we report an improved version of automated reflectance target detection and a new semi-automatic reflectance target detection developed. Test results showed that among the 14 reflectance targets, 13 targets were detected with the automatic target detection method. The undetected target was extracted by the proposed semi-automatic target detect method. Additional test was conducted on the remaining targets to confirm the applicability of our semi-automatic target detection method. As a result, other targets were also detected. The proposed automated and semi-automated target detection method can be used for automated vicarious calibration of UAV images.

Radiometric correction and normalization of airborne LiDAR intensity data for land cover classification

Airborne Light Detection And Ranging (LiDAR) has been used extensively to model the topography of the Earth surface by emitting laser pulses and measuring the distance (range) between the LiDAR sensor and the illuminated object as well as the backscattered laser energy (intensity). Nowadays, airborne LiDAR systems operating in near-infrared spectrum are also gaining a high level of interest for surface classification and object recognition. Nevertheless, due to the system- and environmental- induced distortions, airborne LiDAR intensity data requires certain correction and normalization schemes to maximize the benefits from the collected data. The first part of the thesis presents a correction model for airborne LiDAR intensity data based on the radar (range) equation. To fill the gap in current research, the thesis introduces a set of correction parameters considering the attenuation due to atmospheric absorption and scattering which have not been previously considered. The thesis further derives a set of equations to compute the laser incidence angle based on the LiDAR data point cloud and GPS trajectory. In the second part of the thesis, a normalization model is proposed to adjust the radiometric misalignment amongst overlapping airborne LiDAR intensity data. The model is built upon the use of a Gaussian mixture modeling technique for fitting the intensity histogram which can then be partitioned into several sub-histograms. Finally, sub-histogram equalization is applied to calibrate the LiDAR intensity data. To evaluate the effects of the proposed methods, a LiDAR dataset covering an urban area with three different scans was used for experimental testing. The results showed that the coefficient of variance of five land cover features were significantly reduced by 70% to 82% and 33% to 80% after radiometric correction and radiometric normalization, respectively. Land cover classification was conducted on the LiDAR intensity data where accuracy improvements of up to 15% and 16.5% were found on the classification results using the radiometrically corrected intensity data, and radiometrically corrected and normalized intensity data, respectively. With the improved land cover homogeneity and classification accuracy, the effectiveness of the proposed approach was demonstrated. The outcome of the thesis fills the gap in existing airborne LiDAR research and paves the way for the future development of LiDAR data processing system.

Radiometric correction and normalization of airborne LiDAR intensity data for land cover classification

Airborne Light Detection And Ranging (LiDAR) has been used extensively to model the topography of the Earth surface by emitting laser pulses and measuring the distance (range) between the LiDAR sensor and the illuminated object as well as the backscattered laser energy (intensity). Nowadays, airborne LiDAR systems operating in near-infrared spectrum are also gaining a high level of interest for surface classification and object recognition. Nevertheless, due to the system- and environmental- induced distortions, airborne LiDAR intensity data requires certain correction and normalization schemes to maximize the benefits from the collected data. The first part of the thesis presents a correction model for airborne LiDAR intensity data based on the radar (range) equation. To fill the gap in current research, the thesis introduces a set of correction parameters considering the attenuation due to atmospheric absorption and scattering which have not been previously considered. The thesis further derives a set of equations to compute the laser incidence angle based on the LiDAR data point cloud and GPS trajectory. In the second part of the thesis, a normalization model is proposed to adjust the radiometric misalignment amongst overlapping airborne LiDAR intensity data. The model is built upon the use of a Gaussian mixture modeling technique for fitting the intensity histogram which can then be partitioned into several sub-histograms. Finally, sub-histogram equalization is applied to calibrate the LiDAR intensity data. To evaluate the effects of the proposed methods, a LiDAR dataset covering an urban area with three different scans was used for experimental testing. The results showed that the coefficient of variance of five land cover features were significantly reduced by 70% to 82% and 33% to 80% after radiometric correction and radiometric normalization, respectively. Land cover classification was conducted on the LiDAR intensity data where accuracy improvements of up to 15% and 16.5% were found on the classification results using the radiometrically corrected intensity data, and radiometrically corrected and normalized intensity data, respectively. With the improved land cover homogeneity and classification accuracy, the effectiveness of the proposed approach was demonstrated. The outcome of the thesis fills the gap in existing airborne LiDAR research and paves the way for the future development of LiDAR data processing system.