High Spatial Resolution Image Data Research Articles

A dataset of UAV multispectral images for the grassland-livestock balance

The grassland-livestock balance system plays an important role in reconciling the protection of grassland resources with the sustainable development of animal husbandry. The accurate estimation of the grassland-livestock balance is the basis for the implementation of the grassland-livestock balance system. In order to explore the accurate calculation method of the important parameter, i.e. livestock grass demand in the calculation system, in this study, we set up an experimental area in the natural grassland of Hulunbuir, Inner Mongolia, to simulate a real grazing scene, and conducted 4 grazing experiments controlling the sheep number and grazing time. We obtained multispectral images of the experimental area before and after grazing were with the help of the UAV remote sensing, and carried out a quadrat survey. After data sorting and preprocessing, we produced this dataset of high spatial resolution image data and accurate quadrat investigation data. It can be applied to the research on the daily grass demand of grazing sheep on Hulunbuir meadow steppe in autumn. This dataset provides data support for the accurate estimation of local grassland-livestock balance, and is of great significance for the adaptive dynamic management of grass and livestock.

Guided Filtered Sparse Auto-Encoder for Accurate Crop Mapping from Multitemporal and Multispectral Imagery

Accurate crop mapping is a fundamental requirement in various agricultural applications, such as inventory, yield modeling, and resource management. However, it is challenging due to crop fields’ high spectral, spatial, and temporal variabilities. New technology in space-borne Earth observation systems has provided high spatial and temporal resolution image data as a valuable source of information, which can produce accurate crop maps through efficient analytical approaches. Spatial information has high importance in accurate crop mapping; a Window-based strategy is a common way to extract spatial information by considering neighbourhood information. However, crop field boundaries implicitly exist in image data and can be more helpful in identifying different crop types. This study proposes Guided Filtered Sparse Auto-Encoder (GFSAE) as a deep learning framework guided implicitly with field boundary information to produce accurate crop maps. The proposed GFSAE was evaluated over two time-series datasets of high-resolution PlanetScope (3 m) and RapidEye (5 m) imagery, and the results were compared against the usual Sparse Auto Encoder (SAE). The results show impressive improvements in terms of all performance metrics for both datasets (namely 3.69% in Overal Accuracy, 0.04 in Kappa, and 4.15% in F-score for the PlanetScope dataset, and 3.71% in OA, 0.05 in K, and 1.61% in F-score for RapidEye dataset). Comparing accuracy metrics in field boundary areas has also proved the superiority of GFSAE over the original classifier in classifying these areas. It is also appropriate to be used in field boundary delineation applications.

Mapping Surface Moisture of a Gold Heap Leach Pad at the El Gallo Mine Using a UAV and Thermal Imaging

An understanding of the spatial and temporal variations of surface moisture content over a heap leach pad (HLP) is essential for leaching production and to achieve a high ore recovery. The current practice of leach pad monitoring and data collection remains highly manual and labor-intensive, and exposes technical staff to hazardous material (i.e., cyanide solution). To address these challenges, we propose using unmanned aerial vehicles (UAVs) equipped with thermal imaging sensors to remotely obtain high temporal and spatial resolution image data for monitoring the surface moisture distribution over HLPs. A field study was conducted over a sprinkler-irrigated HLP at El Gallo gold mine in Sinaloa State, Mexico, and the acquired data were used to derive an empirical relationship between the surface moisture content and the remotely sensed surface temperature using linear regression. Moreover, the data were used to generate moisture distribution maps of the entire HLP surface. In situ samples were taken manually during the field experiments to measure the ground-truth material moisture at selected sampling locations. The results show a good agreement between the remote sensing method and the measured ground-truth samples.

Multivariate image fusion: A pipeline for hyperspectral data enhancement

Hyperspectral cameras provide high spectral resolution data, but their usual low spatial resolution when compared to color (RGB) instruments is still a limitation for more detailed studies. This article presents a simple yet powerful method for fusing co-registered high spatial and low spectral resolution image data – e.g. RGB – with low spatial and high spectral resolution data – Hyperspectral. The proposed method exploits the overlap in observed phenomena by the two cameras to create a model through least square projections. This yields two images: 1) A high-resolution image spatially correlated with the input RGB image but with more spectral information than just the 3 RGB bands. 2) A low-resolution image showing the spectral information what is spatially uncorrelated with the RGB image. We show results for semi-artificial benchmark datasets and a real-world application. Performance metrics indicate the method is well suited for data enhancement.

MALDI Matrix Application Utilizing a Modified 3D Printer for Accessible High Resolution Mass Spectrometry Imaging.

Successful matrix-assisted laser desorption ionization (MALDI) mass spectrometry imaging (MSI) relies on the selection of the most appropriate matrix and optimization of the matrix application parameters. In order to achieve reproducible high spatial-resolution imaging data, several commercially available automated matrix application platforms have become available. However, the high cost of these commercial matrix sprayers is restricting access into this emerging research field. Here, we report an automated platform for matrix deposition, employing a converted commercially available 3D printer ($300) and other parts commonly found in an analytical chemistry lab as a low-cost alternative to commercial sprayers. Using printed fluorescent rhodamine B microarrays and employing experimental design, the matrix deposition parameters were optimized to minimize surface analyte diffusion. Finally, the optimized matrix application method was applied to image three-dimensional MCF-7 cell culture spheroid sections (ca. 500 μm diameter tissue samples) and sections of mouse brain. Using this system, we demonstrate robust and reproducible observations of endogenous metabolite and steroid distributions with a high spatial resolution.

Woody Cover Estimates in Oklahoma and Texas Using a Multi-Sensor Calibration and Validation Approach

Woody cover encroachment/expansion/conversion is a complex phenomenon that has environmental and economic impacts around the world. This research demonstrates the development of highly accurate models for estimating percent woody cover using high spatial resolution image data in combination with multi-seasonal Landsat reflectance products. We use a classification and regression tree (CART) approach to classify woody cover using fine resolution multispectral National Agricultural Imaging Program (NAIP) data. A continuous classification and regression tree (Cubist) ingests the aggregated woody cover classification along with the seasonal Landsat data to create a continuous woody cover model. We applied the models, derived by Cubist, across several Landsat scenes to estimate the percentage of woody plant cover, within each Landsat pixel, over a larger regional extent. We measured an average absolute error of 12.1 percent and a correlation coefficient of 0.78 for the models performed. The method of modelling percent woody cover established in this manuscript outperforms currently available woody cover estimates including Landsat Vegetation Continuous Fields (VCF), on average by 26 percent, and Web-Enabled Landsat Data (WELD) products, on average by 16 percent, for the region of interest. Current woody cover products are also limited to certain years and not available pre-2000. This manuscript describes a novel Cubist-based technique to model woody cover for any area of the world, as long as fine (~1–2 m) spatial resolution and Landsat data are available.

Use of a semi-automated object based analysis to map benthic composition, Heron Reef, Southern Great Barrier Reef

ABSTRACTCoral reef scientists and managers require detailed information of reef environments to develop, implement and evaluate management strategies. Maps showing the distribution of corals, seagrass, algae, sand, rubble and rock are an essential component of this information. Using a semi-automated, object-based image analysis (OBIA) approach this study created a benthic community map for Heron Reef (area: approximately 27 km2), Southern Great Barrier Reef, Australia. Benthic cover information was acquired for calibration and validation via georeferenced photograph transects, which coincided with high spatial resolution Quickbird-2 satellite image capture in 2007. Through combination of these data sets and modification of a semi-automated OBIA classification scheme previously developed for seagrass environments to account for the heterogeneous nature of reef benthic communities, a benthic community map was created for Heron Reef with an overall accuracy of 61.6%. The robustness and repeatability of our method provide future opportunity to apply the semi-automated approach to a time series of field and high spatial resolution image data sets to create a time series of benthic community maps for Heron Reef. The results presented in this study therefore open up the opportunity to assess and understand the spatial and temporal dynamics of the reef environment.

A SPIRAL-BASED DOWNSCALING METHOD FOR GENERATING 30 M TIME SERIES IMAGE DATA

Abstract. The spatial detail and updating frequency of land cover data are important factors influencing land surface dynamic monitoring applications in high spatial resolution scale. However, the fragmentized patches and seasonal variable of some land cover types (e. g. small crop field, wetland) make it labor-intensive and difficult in the generation of land cover data. Utilizing the high spatial resolution multi-temporal image data is a possible solution. Unfortunately, the spatial and temporal resolution of available remote sensing data like Landsat or MODIS datasets can hardly satisfy the minimum mapping unit and frequency of current land cover mapping / updating at the same time. The generation of high resolution time series may be a compromise to cover the shortage in land cover updating process. One of popular way is to downscale multi-temporal MODIS data with other high spatial resolution auxiliary data like Landsat. But the usual manner of downscaling pixel based on a window may lead to the underdetermined problem in heterogeneous area, result in the uncertainty of some high spatial resolution pixels. Therefore, the downscaled multi-temporal data can hardly reach high spatial resolution as Landsat data. A spiral based method was introduced to downscale low spatial and high temporal resolution image data to high spatial and high temporal resolution image data. By the way of searching the similar pixels around the adjacent region based on the spiral, the pixel set was made up in the adjacent region pixel by pixel. The underdetermined problem is prevented to a large extent from solving the linear system when adopting the pixel set constructed. With the help of ordinary least squares, the method inverted the endmember values of linear system. The high spatial resolution image was reconstructed on the basis of high spatial resolution class map and the endmember values band by band. Then, the high spatial resolution time series was formed with these high spatial resolution images image by image. Simulated experiment and remote sensing image downscaling experiment were conducted. In simulated experiment, the 30 meters class map dataset Globeland30 was adopted to investigate the effect on avoid the underdetermined problem in downscaling procedure and a comparison between spiral and window was conducted. Further, the MODIS NDVI and Landsat image data was adopted to generate the 30m time series NDVI in remote sensing image downscaling experiment. Simulated experiment results showed that the proposed method had a robust performance in downscaling pixel in heterogeneous region and indicated that it was superior to the traditional window-based methods. The high resolution time series generated may be a benefit to the mapping and updating of land cover data.

Zero-Shot Scene Classification for High Spatial Resolution Remote Sensing Images

Due to the rapid technological development of various sensors, a huge volume of high spatial resolution (HSR) image data can now be acquired. How to efficiently recognize the scenes from such HSR image data has become a critical task. Conventional approaches to remote sensing scene classification only utilize information from HSR images. Therefore, they always need a large amount of labeled data and cannot recognize the images from an unseen scene class without any visual sample in the labeled data. To overcome this drawback, we propose a novel approach for recognizing images from unseen scene classes, i.e., zero-shot scene classification (ZSSC). In this approach, we first use the well-known natural language process model, word2vec, to map names of seen/unseen scene classes to semantic vectors. A semantic-directed graph is then constructed over the semantic vectors for describing the relationships between unseen classes and seen classes. To transfer knowledge from the images in seen classes to those in unseen classes, we make an initial label prediction on test images by an unsupervised domain adaptation model. With the semantic-directed graph and initial prediction, a label-propagation algorithm is then developed for ZSSC. By leveraging the visual similarity among images from the same scene class, a label refinement approach based on sparse learning is used to suppress the noise in the zero-shot classification results. Experimental results show that the proposed approach significantly outperforms the state-of-the-art approaches in ZSSC.

A Quantitative Assessment of Forest Cover Change in the Moulouya River Watershed (Morocco) by the Integration of a Subpixel-Based and Object-Based Analysis of Landsat Data

A quantitative assessment of forest cover change in the Moulouya River watershed (Morocco) was carried out by means of an innovative approach from atmospherically corrected reflectance Landsat images corresponding to 1984 (Landsat 5 Thematic Mapper) and 2013 (Landsat 8 Operational Land Imager). An object-based image analysis (OBIA) was undertaken to classify segmented objects as forested or non-forested within the 2013 Landsat orthomosaic. A Random Forest classifier was applied to a set of training data based on a features vector composed of different types of object features such as vegetation indices, mean spectral values and pixel-based fractional cover derived from probabilistic spectral mixture analysis). The very high spatial resolution image data of Google Earth 2013 were employed to train/validate the Random Forest classifier, ranking the NDVI vegetation index and the corresponding pixel-based percentages of photosynthetic vegetation and bare soil as the most statistically significant object features to extract forested and non-forested areas. Regarding classification accuracy, an overall accuracy of 92.34% was achieved. The previously developed classification scheme was applied to the 1984 Landsat data to extract the forest cover change between 1984 and 2013, showing a slight net increase of 5.3% (ca. 8800 ha) in forested areas for the whole region.

A high performance 3D cluster-based test of unsmoothed fMRI data

Cluster-size tests (CST) based on random field theory have been widely adopted in fMRI data analysis to detect brain activation. However, most existing approaches can be used appropriately only when the image is highly smoothed in the spatial domain. Unfortunately, spatial smoothing degrades spatial specificity. Recently, a threshold-free cluster enhancement technique was proposed which does not require spatial smoothing, but this method can be used only for group level analysis. Advances in imaging technology now yield high quality high spatial resolution imaging data in single subjects and an inference approach that retains the benefits of greater spatial resolution is called for. In this work, we present a new CST with a correction for voxelation to address this problem. The theoretical formulation of the new approach based on Gaussian random fields is developed to estimate statistical significance using 3D statistical parametric maps without assuming spatial smoothness. Simulated phantom and resting-state fMRI experimental data are then used to compare the voxelation-corrected procedure to the widely used standard random field theory. Unlike standard random field theory approaches, which require heavy spatial smoothing, the new approach has a higher sensitivity for localizing activation regions without the requirement of spatial smoothness.

Assessing stream bank condition using airborne LiDAR and high spatial resolution image data in temperate semirural areas in Victoria, Australia

Stream bank condition is an important physical form indicator for streams related to the environmental condition of riparian corridors. This research developed and applied an approach for mapping bank condition from airborne light detection and ranging (LiDAR) and high-spatial resolution optical image data in a temperate forest/woodland/urban environment. Field observations of bank condition were related to LiDAR and optical image-derived variables, including bank slope, plant projective cover, bank-full width, valley confinement, bank height, bank top crenulation, and ground vegetation cover. Image-based variables, showing correlation with the field measurements of stream bank condition, were used as input to a cumulative logistic regression model to estimate and map bank condition. The highest correlation was achieved between field-assessed bank condition and image-derived average bank slope (R2 1/4 0.60, n 1/4 41), ground vegetation cover (R2 1/4 0.43, n 1/4 41), bank width/height ratio (R2 1/4 0.41, n 1/4 41), and valley confinement (producer's accuracy 1/4 100%, n 1/4 9). Crossvalidation showed an average misclassification error of 0.95 from an ordinal scale from 0 to 4 using the developed model. This approach was developed to support the remotely sensed mapping of stream bank condition for 26,000 km of streams in Victoria, Australia, from 2010 to 2012.

THE DEVELOPMENT OF A FAMILY OF LIGHTWEIGHT AND WIDE SWATH UAV CAMERA SYSTEMS AROUND AN INNOVATIVE DUAL-SENSOR ON-SINGLE-CHIP DETECTOR

Abstract. Together with a Belgian industrial consortium VITO has developed the lightweight camera system MEDUSA. It combines high spatial resolution with a wide swath to support missions for large scale mapping and disaster monitoring applications. MEDUSA has been designed to be operated on a solar-powered unmanned aerial vehicle flying in the stratosphere. The camera system contains a custom designed CMOS imager with 2 sensors (each having 10000 × 1200 pixels) on 1 chip. One sensor is panchromatic, one is equipped with colour filters. The MEDUSA flight model camera has passed an extensive test campaign and is ready to conduct its maiden flight. First airborne test flights with an engineering model version of the camera have been executed to validate the functionality and the performance of the camera. An image stitching work flow has been developed in order to generate an image composite in near real time of the acquired images. The unique properties of the dual-sensor-on-single-chip detector triggered the development of 2 new camera designs which are currently in preparation. MEDUSA-low is a modified camera system optimised for compatibility with more conventional UAV systems with a payload capacity of 5–10 kg flying at an altitude around 1 km. Its camera acquires both panchromatic and colour images. The MEDUSA geospectral camera is an innovative hyperspectral imager which is equipped with a spatially varying spectral filter installed in front of one of the two sensors. It acquires both hyperspectral and broad band high spatial resolution image data from one and the same camera.

MODIS and ASTER synergy for characterizing thermal volcanic activity

Volcanic activity is diverse in its manifestations and spans wide temporal and spatial scales. Monitoring volcanoes with any single sensor system can only provide a limited perspective on the nature of such activity because of trade-offs between spatial, temporal and spectral resolution. Spaceborne observations of volcanoes are thus optimized by utilizing data from different and complementary remote sensing instruments. This study examines the combined use of the Moderate Resolution Imaging Spectroradiometer (MODIS) and the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) for analyzing thermal anomalies from four separate volcanoes: Erta 'Ale (Ethiopia), Kīlauea (Hawai'i), Láscar (Chile) and Kliuchevskoi (Russia). MODIS provides a high temporal resolution (daily) but low spatial resolution (~1km) dataset and is used to measure volcanic thermal radiance. ASTER provides relatively high spatial resolution (~90m) image data and is used to define the size of thermal features, albeit at a poorer temporal frequency. Erta 'Ale and Kīlauea have been erupting continuously over the 11-year study period and their thermal emission signatures reveal a persistent baseline on which occasional increases in thermal output, associated with more profuse effusions of lava, are superimposed. Kliuchevskoi also displays occasional peaks in thermal activity due to the eruption of lava however intervening periods are characterized by weak or absent thermal anomalies. Láscar, on the other hand, provides an example of volcanic activity that is difficult for MODIS-type sensors to detect. It is characterized by fumarolic activity that is too low in temperature and too spatially limited to be readily detectable with its coarser instantaneous field of view (i.e., pixel size). However, high resolution data from ASTER can be used to supplement and complement the MODIS dataset. Beyond characterizing the size of thermal anomalies ASTER can also provide constraints on their shape, location and orientation and is capable of detecting relatively subtle thermal anomalies given its high spatial resolution thermal infrared wavebands. These qualities permitted the identification of a thermal precursor to lava flows at Kliuchevskoi. The precursor, which was associated with a rising lava lake, is characterized as an increase in the size and intensity of ASTER-detected thermal anomalies situated in the active crater. Wavelet analysis of the MODIS dataset was also carried out. This revealed the time scales over which radiant output waxed and waned during individual eruptive episodes as well as the repose time between episodes. Such information might be useful in forecasting the likely duration and onset of future eruptions.

Feature selection based on max–min-associated indices for classification of remotely sensed imagery

This article proposes two novel feature selection methods for dimension reduction according to max–min-associated indices derived from Cramer's V-test coefficient. The proposed methods incrementally select features simultaneously satisfying the criteria of a statistically maximal association (A) between target labels and features and a minimal association (R) among selected features with respect to Cramer's V-test value. Two indices are developed by different combinations of the A and R conditions. One index is to maximize A/R and the other is to maximize A–λR, which are referred to as the MMAIQ and MMAIS methods, respectively. Since the proposed feature selection algorithms are feature filter methods, how to determine the best number of features is another important issue. This article adopts an information lost criterion by measuring the variation between χ2 and β statistics to optimize the number of features selected associated with the Gaussian maximal likelihood classifier (GMLC). To validate the proposed methods, experiments are conducted with both a hyperspectral image data set and a high spatial resolution image data set. The results demonstrate that the proposed methods can provide an effective tool for feature selection and improve classification accuracy significantly. Furthermore, the proposed methods with well-known feature selection methods, i.e. mutual information-based max-dependency criterion (mRMR) and sequential forward selection (SFS), are evaluated and compared. The experiments demonstrate that the proposed methods can offer better results in terms of kappa coefficient and overall classification accuracy measurements.

Forest structural diversity characterization in Mediterranean pines of central Spain with QuickBird-2 imagery and canonical correlation analysis

Variation in forest structure provides information on vegetation complexity and provides insights on biodiversity. Characterizing forest structural diversity with remotely sensed data supports reporting, monitoring, and policy development. We explored the relationship between forest structural diversity in Mediterranean pines of the Spanish Central Range and variables derived from imagery captured with a commercial high spatial resolution satellite (QuickBird-2; with pixels sided 2.4 m multispectral and 0.68 m panchromatic). To combine multiple aspects of tree conditions at a stand level, “structural diversity” was characterized at the plot level (N = 1022) as a linear combination of the median of absolute differences of individual trees’ bole diameter, height, and crown diameter measured on the field from the local median equivalents. Spectral reflectance variations in the visible and near-infrared, as well as image co-occurrence texture metrics from the panchromatic imagery at various window sizes were generated. All relationships to image-derived values were assessed against circular 0.3 ha areas corresponding with the field measured plots. Canonical correlation analysis aided in identification of combinations of reflectance and texture metrics most highly related with forest structural diversity (R = 0.89). Reflectance diversity was found to be more important than co-occurrence texture features in describing forest structural diversity when forest structure was limited (R = 0.47 vs. R = 0.39), whereas texture was more informative to the model when the forest structural diversity was high (R = 0.88 vs. R = 0.63), relating more complex forest conditions. Our results, although empirically defined by the local conditions and image acquisition characteristics, demonstrated the potential in high spatial resolution imagery for description of forest structural diversity in forests of the Mediterranean environment, especially important for Spain where a national high spatial resolution image data base has been collected.

Automatic Geographic Object Based Mapping of Streambed and Riparian Zone Extent from LiDAR Data in a Temperate Rural Urban Environment, Australia

This research presents a time-effective approach for mapping streambed and riparian zone extent from high spatial resolution LiDAR derived products, i.e., digital terrain model, terrain slope and plant projective cover. Geographic object based image analysis (GEOBIA) has proven useful for feature extraction from high spatial resolution image data because of the capacity to reduce effects of reflectance variations of pixels making up individual objects and to include contextual and shape information. This functionality increases the likelihood of developing transferable and automated mapping approaches. LiDAR data covered parts of the Werribee Catchment in Victoria, Australia, which is characterized by urban, agricultural, and forested land cover types. Field data of streamside vegetation structure and physical form properties were used for both calibration of the mapping routines and validation of the mapping results. To improve the transferability of the rule set, the GEOBIA approach was developed for an area representing different riparian zone environments, i.e., urbanized, agricultural and hilly forested areas. Results show that mapping streambed extent (R2 = 0.93, RMSE = 3.6 m, n = 35) and riparian zone extent (R2 = 0.74, RMSE = 3.9, n = 35) from LiDAR derived products can be automated using GEOBIA to enable derivation of spatial information in an accurate and time-effective manner suited for natural resource management agencies.

On the classification of remote sensing high spatial resolution image data

In this paper is investigated a methodology implementing an object-based approach to digital image classification using spectral and spatial attributes in a multiple-stage classifier structured as a binary tree. It is a well-established fact that object-based image classification is particularly appropriate when dealing with high spatial resolution image data. Following this approach, the image is initially segmented into objects that carry informational value. Next, spectral and spatial attributes are extracted from every object in the scene, and implemented into a classifier to produce a thematic map. As the combined number of spectral and spatial variables may become large compared to the number of available training samples, a reduction in the data dimensionality may be required whenever parametric classifiers are used, in order to mitigate the effects of the Hughes phenomenon. To this end the sequential feature selection (SFS) procedure is applied in a multiple-stage classifier structured as a binary tree. The advantage of a binary tree classifier lies in the fact that only one pair of classes is considered at each stage (node), allowing for an optimal selection of features. This proposed approach was tested using Quickbird image data covering an urban scene. The results are compared against results yielded by the traditional single-stage Gaussian maximum likelihood classifier. The results suggest the proposed methodology is adequate in the classification of high spatial resolution image data.

Object-based feature extraction using high spatial resolution satellite data of urban areas

Urban morphology is characterized by a complex and variable coexistence of diverse, spatially and spectrally heterogeneous objects. Built-up areas are among the most rapidly changing and expanding elements of the landscape. Thus, remote sensing becomes an essential field for up-to-date and area-wide data acquisition, especially in explosively sprawling cities of developing countries. The urban heterogeneity requires high spatial resolution image data for an accurate geometric differentiation of the small-scale physical features. This study proposes an object-based, multi-level, hierarchical classification framework combining shape, spectral, hierarchical and contextual information for the extraction of urban features. The particular focus is on high class accuracies and stable transferability by fast and easy adjustments on varying urban structures or sensor characteristics. The framework is based on a modular concept following a chronological workflow from a bottom-up segmentation optimization to a hierarchical, fuzzy-based decision fusion top-down classification. The workflow has been developed on IKONOS data for the megacity Istanbul, Turkey. Transferability is tested based on Quickbird data from the various urban structures of the incipient megacity Hyderabad, India. The validation of both land-cover classifications shows an overall accuracy of more than 81 percent.

Hybrid Segmentation – Artificial Neural Network Classification of High Resolution Hyperspectral Imagery for Site-Specific Herbicide Management in Agriculture

Site-Specific Herbicide Management (SSHM) in Precision Agriculture (PA) requires weed detection in crop fields for directed herbicide application instead of spraying entire fields. This has significant economic and environmental advantages for improved agricultural practices that are essential given forecast increases in global population and food production needs. In this study, a new hybrid segmentation - Artificial Neural Network (HS-ANN) method was compared to standard Maximum Likelihood Classification (MLC) for improving crop/weed species discrimination in SSHM/PA. Very high spatial resolution (1.25 mm) ground-based hyperspectral image data were acquired over field plots of canola, pea, and wheat crops seeded with two weed species (redroot pigweed, wild oat) in southern Alberta, Canada. The very high spatial and spectral resolution image data required development of a simple yet efficient vegetation index (Modified Chlorophyll Absorption in Reflectance Index (MCARI)) threshold segmentation to separate vegetation from soil for classification. The HSANN consistently outperformed MLC in both single date and multi-temporal classifications. Higher class accuracies were obtained with multi-temporally trained ANNs (84 to 92 percent overall), with improvements up to 31 percent over MLC. With advancements in imaging technology and computer processing speed, this HS-ANN method may constitute a viable farm option for real-time detection and mapping of weed species for SSHM in agriculture.