Pradeep K. Goel

Classification of hyperspectral data by decision trees and artificial neural networks to identify weed stress and nitrogen status of corn

This study evaluates the potential of decision tree classification algorithms for the classification of hyperspectral data, with the goal of discriminating between different growth scenarios in a cornfield. A comparison was also made between decision tree and artificial neural networks (ANNs) classification accuracies. In the summer of the year 2000, a two-factor field experiment representing different crop conditions was carried out. Corn was grown under four weed management strategies: no weed control, control of grasses, control of broadleaf weeds, and full weed control with nitrogen levels of 60, 120, and 250 N kg/ha. Hyperspectral data using a Compact Airborne Spectrographic Imager were acquired three times during the entire growing season. Decision tree technology was applied to classify different treatments based on the hyperspectral data. Various tree-growing mechanisms were used to improve the accuracy of classification. Misclassification rates of detecting all the combinations of different nitrogen and weed categories were 43, 32, and 40% for hyperspectral data sets obtained at the initial growth, the tasseling and the full maturity stages, respectively. However, satisfactory classification results were obtained when one factor (nitrogen or weed) was considered at a time. In this case, misclassification rates were only 22 and 18% for nitrogen and weeds, respectively, for the data obtained at the tasseling stage. Slightly better results were obtained by following the ANN approach. However, the advantage with the decision tree was the formulation of simple and clear classification rules. The highest accuracy was obtained for the data acquired at tasseling stage. The results indicate the potential of decision tree classification algorithms and ANN usage in the classification of hyperspectral data for crop condition assessment.

ESTIMATION OF CROP BIOPHYSICAL PARAMETERS THROUGH AIRBORNE AND FIELD HYPERSPECTRAL REMOTE SENSING

The potential of airborne hyperspectral remote sensing in crop monitoring and estimation of various biophysical parameters was examined in this study. A field experiment, consisting of four weed control strategies (no weed control, broadleaf control, grass control, and full weed control) as the main plot effect, factorially combined with three nitrogen (N) fertilization rates (60, 120, and 250 N kg ha–1), and replicated four times, was conducted. Hyperspectral data in 72 narrow wavebands (409 to 947 nm) from a Compact Airborne Spectrographic Imager (CASI) sensor were acquired 30 days after planting, at tasseling, and at the fully mature stage. In addition, measurements were made concurrently on various crop physiological parameters: leaf greenness (SPAD readings), leaf area index (LAI), plant height, leaf nitrogen content, leaf chlorophyll content, and associated factors such as soil moisture. Regression models were generated to estimate crop biophysical parameters and yield, in terms of reflectance at one or more wavebands, using the maximum r2 improvement criterion. The models that best represented the data had five wavebands as independent variables. Coefficients of determination (r2) were generally greater than 0.9, when based on the spectral data taken at the tasseling stage. Results were improved when normalized difference vegetation indices (NDVI) were used rather than the five–waveband reflectance values. The wavebands at 701 nm and 839 nm were the most prevalent in the NDVI–based models.

Potential of airborne hyperspectral remote sensing to detect nitrogen deficiency and weed infestation in corn

A compact airborne spectrographic imager (CASI) was used to obtain images over a field that had been set up to study the effects of various nitrogen application rates and weed control on corn (Zea mays). The objective was to determine to what extent the reflectances obtained in the 72 visible and near-infrared (NIR) wavebands (from 409 to 947 nm) might be related to differences associated with combinations of weed control (none, full, grasses only or broadleafs only) and nitrogen application rate (60, 120 or 250 kg/ha). Plots were arranged in split-plot experiment in completely randomized design at the McGill University Research Farm on Macdonald Campus, Ste Anne de Bellevue, Que., Canada. Weeding treatments were assigned to the main-plot units, and nitrogen rates to the sub-plot units. Three flights were made during the growing season. Data were analyzed for each flight and each band separately, then regrouped into series of neighboring bands yielding identical analyses with respect to the significance of the main effects and interactions on reflectance. The results indicate that the reflectance of corn is significantly influenced (α=0.05) at certain wavelengths by the presence of weeds, the nitrogen rates and their interaction. The influence of weeds was most easily observed in the data from the second flight (August 5, 2000), about 9 weeks after planting. The nitrogen effect was detectable in all the three flights. Differences in response due to nitrogen stress were most evident at 498 nm and in the band at 671 nm. In these bands, differences due to nitrogen levels were observed at all growth stages, and the presence of weeds had no interactive effect. Differences in other regions, whether related to nitrogen, weeds or the combination of the two, appeared to be dependent on the growth stage. Furthermore, results comparable to those of the hyperspectral sensor were obtained when a multispectral sensor was simulated, indicating little advantage of using the former.

APPLICATION OF DECISION TREE TECHNOLOGY FOR IMAGE CLASSIFICATION USING REMOTE SENSING DATA

Abstract Hyperspectral images of plots, cropped with silage or grain corn and cultivated with conventional tillage, reduced tillage, or no till, were classified using the classification and regression tree (C&RT) approach, an innovative intelligent computational algorithm in data mining. Each tillage/cropping combination was replicated three times, for a total of 18 plots. Five hyperspectral reflectance measurements per plot were taken randomly to obtain a total of 90 measurements. Images were taken on June 30, August 5, and August 25, 2000 to reflect three stages of crop development. Each measurement consisted of reflectances in 71 wave bands ranging from 400 to 950 nm. C&RT models were developed separately for the three observation dates, using the 71 reflectances as inputs to classify the image according to: (a) tillage practice, (b) residue level, (c) cropping practices, (d) tillage/cropping (residue) combination. C&RT models could generally distinguish tillage practices with a classification accuracy of 0.89 and residue levels with a classification accuracy of 0.98.

CLASSIFICATION ACCURACY OF DISCRIMINANT ANALYSIS, ARTIFICIAL NEURAL NETWORKS, AND DECISION TREES FOR WEED AND NITROGEN STRESS DETECTION IN CORN

Hyperspectral images of experimental plots, cropped with corn (Zea mays L.) and to which twelve combinations of three nitrogen application rates and four weed management strategies were applied, were obtained with a 72-waveband compact airborne spectrographic imager (CASI). The images were taken at three times during the 2000 growing season: early growth, tasseling, and full maturity. Nitrogen application rates were 60, 120, and 250 kg N ha-1. Weed controls were: none, control of grasses, control of broadleaf weeds, and full weed control. The objective of this study was to evaluate discriminant analysis as a tool for classifying images with respect to the nitrogen and weed management practices applied to the experimental plots, and to compare the classification accuracy of this technique with those obtained by artificial neural network (ANN) and decision tree (DT) algorithms on the same data. Significant wavebands were selected, among the 72 available, using the stepwise option of the STEPDISC procedure (SAS software). Classification accuracy was determined for the full set of selected wavebands and for subsets thereof, for three problems: distinguishing between the 12 combinations of factor levels, differentiating between nitrogen levels only, and separating weed controls only. Misclassification rates of images, taken at the initial growth stage, were substantially lower for each of these tasks (25%, 17%, and 13%, respectively) when discriminant analysis was used. The ANN approach was best for images taken at the tasseling and full maturity stages. However, from the precision-farming point of view, it is easier to apply site-specific remedies to weed and nitrogen stresses early in the season than when the corn crop has reached the tasseling stage, so the results obtained with the discriminant analysis are noteworthy.

DISCRIMINANT ANALYSIS OF HYPERSPECTRAL DATA FOR ASSESSING WATER AND NITROGEN STRESSES IN CORN

The development and implementation of both economically and environmentally sustainable precision crop management systems can be greatly enhanced through the use of remote sensing. In this study, the potential of narrow-waveband hyperspectral observations in the discrimination of nitrogen and water stresses in corn (Zea mays L.) was investigated. A field experiment was conducted in the summer of 2002 at the Macdonald Research Farm, McGill University, Ste-Anne-de-Bellevue, Quebec, Canada. Corn was grown in forty 9.0 × 10.0 m test plots laid out in a split-plot design with irrigation (non-irrigated, irrigated) as the main treatment and nitrogen fertilizer application rate (50, 100, 150, 200, and 250 kg ha-1) as the sub-treatment. Hyperspectral measurements in 2151 wavebands (350 to 2500 nm) were made with a field spectroradiometer during the entire growing season. Using a stepwise procedure, the most effective wavebands capable of discriminating treatment effects were selected. By applying a discrimination procedure with a well-chosen subset of the selected wavebands, treatments were correctly classified with more than 95% accuracy. Specific narrow wavebands, from different portions of the spectrum, allowed the discrimination of plots differing in their irrigation and nitrogen treatments. This study supports past work suggesting that greater spectral resolution should lead to more consistent relationships between the spectral data and different crop status indicators.

USE OF AIRBORNE MULTI-SPECTRAL IMAGERY FOR WEED DETECTION IN FIELD CROPS

In this article, the potential of multi-spectral airborne remote sensing is evaluated for the detection of weed infestation in corn (Zea mays L.) and soybean (Glycine max.) crops. A field plot experiment was laid out at the Lods Agronomy Research Center of Macdonald Campus, McGill University, Ste-Anne-de-Bellevue, Quebec, Canada. A multi-spectral image in 24 wavebands (475.12 nm to 910.01 nm wavelength range) was obtained using an airborne platform. Three weed treatments were selected to represent different weed conditions in corn and soybean, namely velvetleaf (Abutilon theophrasti Medic.), grasses, and mixed weeds. For the purpose of comparison, a treatment without weeds was also planted of each type of crop. Statistical analysis of radiance values recorded in different wavebands was performed to find the wavelength regions that were most useful for detecting different weed infestations. The results indicate that wavebands centered at 675.98 and 685.17 nm in the red region, and from 743.93 nm to 830.43 nm in the near-infrared, have good potential for distinguishing weeds in corn. For soybean, however, only one waveband (811.40 nm) was found to be useful. Efforts were also made to evaluate various ratios of radiance values recorded in red and near-infrared (NIR) wavebands for the detection of weeds. Much better results were obtained when ratios were used than with single wavebands. The results of this study will be helpful in selecting the most useful parts of the electromagnetic spectrum for the detection of weeds in corn and soybean fields.

Hyperspectral Image Classification to Detect Weed Infestations and Nitrogen Status in Corn

The potential of hyperspectral aerial imagery for the detection of weed infestation and nitrogen fertilization level in a corn (Zea mays L.) crop was evaluated. A Compact Airborne Spectrographic Imager (CASI) was used to acquire hyperspectral data over a field experiment laid out at the Lods Agronomy Research Centre of Macdonald Campus, McGill University, Quebec, Canada. Corn was grown under four weed management strategies (no weed control, control of grasses, control of broadleaf weeds, and full weed control) factorally combined with nitrogen fertilization rates of 60, 120, and 250 N kg/ha. The aerial image was acquired at the tasseling stage, which was 66 days after planting. For the classification of remote sensing imagery, various widely used supervised classification algorithms (maximum likelihood, minimum distance, Mahalanobis distance, parallelepiped, and binary coding) and more sophisticated classification approaches (spectral angle mapper and linear spectral unmixing) were investigated. It was difficult to distinguish the combined effect of both weed and nitrogen treatments simultaneously. However, higher classification accuracies were obtained when only one factor, either weed or nitrogen treatment, was considered. With different classifiers, depending on the factors considered for the classification, accuracies ranged from 65.84% to 99.46%. No single classifier was found useful for all the conditions.

Pollutants Removal by Vegetative Filter Strips Planted with Different Grasses

Over the last few years, increasing occurrence of deadly pathogens and presence of various pollutants (nutrients, pesticides, other chemicals, and sediments) above the prescribed limit in water systems, clearly indicate alarmingly deteriorating quality of water resources. As a result, farming systems that are known to be the main non-point or diffuse pollution source are being reviewed microscopically. Vegetative Filter Strip (VFS) is considered to be one of the best management practices (BMPs) for effective control sediment and nutrient transport over agricultural lands. Many laboratory and field scale studies have also indicated the limited usefulness of VFS to control movement of bacteria in surface runoff. However, design of VFS under field conditions still remains a challenge due to variation in upland hydrological parameters and factors effecting movement of pollutants through VFS such as type of vegetation cover and density, width of strip, and land slope. Determination of trapping efficiency of VFS for bacteria is more complex due to the complex interaction of various factors governing the die-of and re-growth of bacteria under field condition, and release of bacteria from soil reserve. An extensive field experiment is being conducted at the research farm of University of Guelph in Southern Ontario, Canada, to evaluate to effectiveness of VFS under different vegetation cover, ground slope, width of filter strip, and in various seasons. Concentration of sediment reduced an average by 88.3% and almost 94.3% sediment mass was trapped in various filter strips. Higher trapping efficiencies for mass were observed for sediment bound nutrients (94.5% and 93.9% for N and P, respectively) compared to soluble forms (57.0% and 77.3% for N and P, respectively). Results for bacteria (Total Coliforms, Fecal Coliforms, and E. Coli) through VFSs were encouraging but not conclusive. In the present paper, experiment and results of the study are presented and discussed in details.

DIFFERENTIATION OF CROP AND WEEDS BY DECISION-TREE ANALYSIS OF MULTI-SPECTRAL DATA

The purpose of this study was to use a data mining technique (i.e., decision trees) to classify multi-spectral images of experimental plots having different crop and weed populations. Eleven types of plots were prepared for this study. Eight types were seeded with corn or soybeans and were either: (1) weed-free, (2) co-populated by velvetleaf only, (3) co-populated with a mixture of grass species, or (4) co-populated with the predominant weed species of the regions. The other three types were as (2), (3), and (4) with neither corn nor soybeans. An aircraft-mounted pushbroom imaging spectrometer was used to obtain scans of the plots in one blue, five green, five red, and thirteen infrared bands. Eight classification problems involving different degrees of recognition complexity were set up. Each was tested using three different input types from the multi-spectral data. The three types of input were: (a) absolute values of radiance from the 24 wavebands; (b) vegetation index (VI), which consists of 12 inputs; and (c) normalized difference vegetation index (NDVI), which consists of 65 inputs. Results showed that the most complex classification problem (distinguishing between 11 crop/weed combinations) was best resolved using the NDVI inputs (success classification of 0.85 as compared with 0.79 and 0.55 for the absolute radiance and VI, respectively). Moreover, NDVI performed best as inputs in seven out of the eight problems.