In recent days, phenotyping of various crops is gaining widespread popularity due to its ability to recognize variations in the effects of different genotypes of a particular crop in terms of its growth, yield, biomass, and so on. Such an application requires extensive data collection and analysis with a high spatial and temporal resolution, which can be attained using multiple sensors onboard Unmanned Aerial Vehicles (UAVs). In this study, we focus on harnessing information from a variety of sensors, such as RGB cameras, LiDAR units, and push-broom hyperspectral sensors – Short-wave Infrared (SWIR) and Visible Near Infrared (VNIR). The major challenge that needs to be overcome in this regard is to ensure an accurate integration of information captured across several days from the different sensor modalities. Moreover, the payload constraint for UAVs restrain us from mounting all the sensors simultaneously during a single flight mission, thus entailing the need for data capture from different sensors mounted on separate platforms that are flown individually over the agricultural field of interest. The first step towards integration of different data modalities is the generation of georeferenced products from each of the flight missions, which is accomplished with the help of Global Navigation Satellite Systems (GNSS) and Inertial Navigation Systems (INS) mounted on the UAVs that are time-synchronized with the onboard LiDAR units, cameras and/or hyperspectral sensors. Furthermore, an accurate georeferencing is achieved by developing robust calibration approaches dedicated towards accurate estimation of mounting parameters of the involved sensors. Finally, the geometric and spectral characteristics, such as canopy cover and leaf count, derived from the different sensors are used to devise a model to analyze the phenotypic traits of crops. The preliminary results indicate that the proposed calibration techniques can attain an accuracy of upto 3 cm.
Classification of multi-source data has recently gained significant attention, as accuracies can often be improved by incorporating complementary information extracted in single and multi-sensor scenarios. Supervised approaches to classification of multi-source remote sensing data are dependent on the availability of representative labeled data, which are often limited relative to the dimensionality of the data for training. To address this problem, in this paper, we propose a new framework in which active learning (AL) and semi-supervised learning (SSL) strategies are combined for multi-source classification of hyperspectral images. First, the spatial-spectral features are represented via the redundant discrete wavelet transform (RDWT). Then, the spatial context provided by the hierarchical segmentation algorithm (HSEG) in conjunction with an unsupervised pruning strategy is exploited to combine AL and SSL. Finally, SVM classification is performed due to the high dimensionality of the feature space. The proposed framework is validated with two benchmark hyperspectral data sets. Higher classification accuracies are obtained by the proposed framework with respect to other state-of-the-art active learning classification approaches.
In response to the 2010 Haiti earthquake, the ALIRT ladar system was tasked with collecting surveys to
support disaster relief efforts. Standard methodologies to classify the ladar data as ground, vegetation, or
man-made features failed to produce an accurate representation of the underlying terrain surface. The majority
of these methods rely primarily on gradient- based operations that often perform well for areas with low
topographic relief, but often fail in areas of high topographic relief or dense urban environments. An
alternative approach based on a adaptive lower envelope follower (ALEF) with an adaptive gradient operation
for accommodating local slope and roughness was investigated for recovering the ground surface from the
ladar data. This technique was successful for classifying terrain in the urban and rural areas of Haiti over
which the ALIRT data had been acquired.
Hyperspectral data can potentially provide greatly improved capability for discrimination between many land cover types, but new methods are required to process these data and extract the required information. Data sets are extremely large, and the data are not well distributed across these high dimensional spaces. The increased number and resolution of spectral bands, many of which are highly correlated, is problematic for supervised statistical classification techniques when the number of training samples is small relative to the dimension of the input vector. Selection of the most relevant subset of features is one means of mitigating these effects. A new algorithm based on the tabu search metaheuristic optimization technique was developed to perform subset feature selection and implemented within a binary hierarchical tree framework. Results obtained using the new approach were compared to those from a greedy common greedy selection technique and to a Fisher discriminant based feature extraction method, both of which were implemented in the same binary hierarchical tree classification scheme. The tabu search based method generally yielded higher classification accuracies with lower variability than these other methods in experiments using hyperspectral data acquired by the EO-1 Hyperion sensor over the Okavango Delta of Botswana.
Prediction of landcover type from airborne/spaceborne sensors is an important classification problem in remote sensing. Due to advances in sensor technology, it is now possible to acquire hyperspectral data simultaneously in more than 100 bands, each of which measures the integrated response of a target over a narrow window of the electromagnetic spectrum. The bands are ordered by their wavelengths and spectrally adjacent bands are generally statistically correlated. Using such high dimensional data for classification of landcover potentially provides greatly improved results. However, it is necessary to select bands that provide the best possible discrimination among the classes of interest. In this paper, we propose an efficient top-down multiresolution class-dependent feature extraction algorithm for hyperspectral data to be used with a pairwise classification scheme. First, the C class problem is divided into (C2) two class problems. Features for each pair of classes are extracted independently. The algorithm decomposes the bands recursively into groups of adjacent bands (subspaces) in a top-down fashion. The features extracted are specific to the pair of classes that are being distinguished and exploit the ordering information in the hyperspectral data. Experiments on a 183 band AVIRIS data set for a 12 class problem show significant improvements in both classification accuracies and the number of features required for all 66 pairs of classes.
Mapping landcover type from airborne/spaceborne sensors is an important classification problem in remote sensing. Due to advances in sensor technology, it is now possible to acquire hyperspectral data simultaneously in more than 100 bands, each of which measures the integrated response of a target over a narrow window of the electromagnetic spectrum. The bands are ordered by their wavelengths and spectrally adjacent bands are generally statistically correlated.
In recent years, with the development of satellite and computer technology, Earth observation and atmospheric research have become highly dependent on digital imagery. One of the primary interests in digital image processing is the development of robust methods to perform feature detection, extraction, and classification. Until recently, classification methods for cloud discrimination were mainly based on the spectral information of the imagery. However, because of the spectral similarities of certain features (such as ice clouds and snow) and the effects of atmospheric attenuation, multispectral rule-based classifications do not necessarily produce accurate feature discrimination. Spectral homogeneity of two different features within a scene can lead to misclassification. Furthermore, the opposite problem can occur when one feature exhibits different spectral signatures locally but is homogeneous in its cyclic spatial variation. The exploration of spatial information is often advantageous in these discrimination problems. A texture-based method for feature identification has been investigated. This method uses a set of localized spatial filters known as 2-D Gabor functions. Gabor filters can be described as a sinusoidal plane wave within a 2-D Gaussian envelope. The frequency and orientation of the sine plane and the width of the Gaussian envelope are determined by the Gabor parameters. These tunable channels yield joint optimal information both in the spatial and the frequency domains. The new method has been applied to the thermal channels of the NOAA Advanced Very High Resolution Radiometer data for cloud-type discrimination. Results show that additional texture information improves discrimination between cloud types (especially thin cirrus).
A new edge detection algorithm has been developed and implemented that has both good speed and accuracy properties. The accuracy of the approach is derived from scale adaptation through anisotropic diffusion. The speed of the filter is based on recursive filtering. Specifically, the algorithm is implemented through a decomposition of the recursive filter into a convolution of two independent half space filters. This further improves the speed of the recursive filter and facilitates scale adaptation. The resulting algorithm is 1 - 2 orders of magnitude faster than comparable Gaussian based frequency domain and spatial domain methods. The new filter is omni-directional and super-elongated. It is also contour following, has computational complexity which is independent of scale, and has no truncation noise. The algorithm has been implemented and successfully applied to SPOT XS and Landsat MSS and TM imagery as one component of a region based image segmentation scheme.
One of the primary interests in digital image processing is the development of robust methods to perform feature detection, extraction, and classification. Until recently, classification methods for cloud discrimination were mainly based on the spectral information of the imagery. However, because of the spectral similarities of certain features (such as ice clouds and snow) and the effects of atmospheric attenuation, multi-spectral rule based classifications do not necessarily produce accurate feature discrimination. Spectral homogeneity of two different features within a scene can lead to misclassification. Furthermore, the opposite problem can occur when one feature exhibits different spectral signatures locally but is homogeneous in its cyclic spatial variation. The exploration of spatial information is often advantageous in these discrimination problems. A texture-based method for feature identification has been investigated. This method uses a set of localized spatial filters known as two dimensional Gabor functions. Gabor filters can be described as a sinusoidal plane wave within a two-dimensional Gaussian envelope. The frequency and orientation of the sine plane and the width of the Gaussian envelope are determined by the Gabor parameters. These tunable channels yield joint optimal information both in the spatial and the frequency domains. The new method has been applied to the thermal channels of the NOAA-advanced very high resolution radiometer (AVHRR) data for cloud type discrimination.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.