Mohammad Ali Mahmoodi; Sohaila Momeni; Masoud Davari
Abstract
Introduction: Land use and Land cover (LULC) information has been identified as one of the crucial data components for a range of applications including global change studies, urban planning, agricultural crop characterization, and forest ecosystem classification. The derivation of such information increasingly ...
Read More
Introduction: Land use and Land cover (LULC) information has been identified as one of the crucial data components for a range of applications including global change studies, urban planning, agricultural crop characterization, and forest ecosystem classification. The derivation of such information increasingly relies on remote sensing technology due to its ability to acquire valuable spatiotemporal information on LULC. One of the major approaches to deriving LULC information from remotely sensed images is classification. Numerous image classification algorithms exist. Among the most popular are the maximum likelihood classifier (MLC), artificial neural network (ANN) classifiers and decision tree (DT) classifiers. Conventional parametric method like MLC is based on statistical theory and assumes a multivariate normal distribution for each class. In case of data that has non-normal distribution (which is common with LULC data), the parametric classifiers may fail since the inability to resolve interclass confusion. This inability is the major limitation of parametric classifiers. Nonparametric classifiers like ANNs and DTs, which do not rely on any assumptions for the class distributions of data, could overcome the aforementioned limitations of parametric classifiers. The support vector machines (SVMs), a nonparametric classifier, that has recently been used in numerous applications in image processing, represents a group of theoretically superior machine learning algorithms. The SVM employs optimization algorithms to locate the optimal boundaries between classes. It was found competitive with the best available classification methods, including ANN and DT classifiers. The classification accuracy of SVMs is based upon the choice of the classification strategy and kernel function. The objective of this study was to investigate the sensitivity of SVM architecture including classification strategy and kernel types to identify LULC information from Landsat Enhanced Thematic Mapper (ETM) remote sensing data in Gavshan dam watershed in west of Iran.
Materials and Methods: SVMs were used to classify orthocorrected Landsat ETM images of May, 2016. Image pre-processing such as atmospheric correction were conducted before utilization. Three classification strategies (One versus one, one versus all and ordinal) and three types of kernels (linear, polynomial and radial basis function) were used for the SVM classification. A total of 18 different models were developed and implemented for sensitivity analysis of SVM architecture. A two-layer feed-forward Perceptron network classifier with sigmoid hidden and softmax output neurons was also used for comparison. The network was trained using scaled conjugate gradient backpropagation algorithm. A total of 1320 ground control points were collected to train, validate and test the SVM and ANN models. Ground truth locations on each image were identified using the GPS coordinates for extracting spectral reflectance data of seven bands (Bands 1-7) of Landsat ETM images. The LULC class of each point was identified using land survey or Google earth images. The identified LULC classes were agriculture, buffer forests, orchard, ranges brush, range grasses, urban areas, roads and water.
Results and Discussion: The results suggest that the choice of classification strategy and kernel types play an important role on SVMs classification accuracy. Statistical evaluation of the SVM models against the ground control points showed that the one versus one classification strategy had the highest accuracy than the two other ones for any kernel function type and the polynomial kernel function had the highest accuracy than the two other kernels for any classification strategy. The SVM model with polynomial (n=3) kernel and one versus one classification strategy outperformed all SVMs models and gave the highest overall classification accuracy of 78.5 and Kappa coefficient of 68.5. The McNemar’s test clearly showed significant improvement of the best SVM model in comparison to the ANN model (P<0.001). Also, the user accuracy and producer accuracy achieved by best SVM model were higher than ANN model for all LULC classes. In both approaches water and agriculture categories have high accuracy while roads have low accuracy. The resulting LULC map indicated that most parts of the studied area (52.8%) have been assigned to the agriculture. The ranges brush and range grasses categories cover 12.5% and 26.8% of the watershed, respectively. Only about 2.7% of the watershed have been covered with trees.
Conclusions: This study suggests that the SVMs approach based on Landsat ETM bands may provide reliable and accurate LULC information even better that best ANN approaches. However, choice of classification strategy and kernel types play an important role on SVMs classification accuracy. Best model of polynomial kernel and one versus one classification strategy outperformed all SVMs and ANN models and gave the highest classification accuracy.
F. Mahmoodi; R. Jafari; H. Karimzadeh; N. Ramezani
Abstract
Introduction: Use of remote sensing for soil assessment and monitoring started with the launch of the first Landsat satellite. Since then many other polar orbiting Earth-observation satellites such as the Landsat series, have been launched and their imagery have been used for a wide range of soil mapping. ...
Read More
Introduction: Use of remote sensing for soil assessment and monitoring started with the launch of the first Landsat satellite. Since then many other polar orbiting Earth-observation satellites such as the Landsat series, have been launched and their imagery have been used for a wide range of soil mapping. The broad swaths and regular revisit frequencies of these multispectral satellites mean that they can be used to rapidly detect changes in soil properties. Arid and semi-arid lands cover more than 70 percent of Iran and are very prone to desertification. Due to the broadness, remoteness, and harsh condition of these lands, soil studies using ground-based techniques appear to be limited. Remote sensing imagery with its cost and time-effectiveness has been suggested and used as an alternative approach for more than four decades. Flood irrigation is one of the most common techniques in Isfahan province in which 70% of water is lost through evaporation. This system has increased soil salinization and desert-like conditions in the region. For principled decision making on agricultural product management, combating desertification and its consequences and better use of production resources to achieve sustainable development; understanding and knowledge of the origin, amount and area of salinity, the percentage of calcite, gypsum and other mineral of soil in each region is essential. Therefore, this study aimed to map the physical and chemical characteristics of soils in Vazaneh region of Isfahan province, Iran.
Materials and Methods : Varzaneh region with 75000 ha located in central Iran and lies between latitudes 3550234 N and 3594309 N and longitudes 626530 E to 658338 E. The climate in the study area is characterized by hot summers and cold winters. The mean daily maximum temperature ranges from 35°C in summer to approximately 17°C in winter and mean daily minimum temperature ranges from 5°C in summer to about -24.5°C in winter. The mean annual evaporation rate is 3265 mm. In this study, image processing techniquess including band combinations, Principal Component Analysis (PC1, PC2 and PC3), and classification were applied to a TM image to map different soil properties. In order to prepare the satellite image, geometric correction was performed. A 1:25,000 map (UTM 39) was used as a base to georegister the Landsat image. 40 Ground Control Points (GCPs) were selected throughout the map and image. Road intersections or other man-made features were appropriate targets for this purpose. The raw image was transformed to the georectified image using a first order polynomial, and then resampled using the nearest neighbour method to preserve radiometry. The final Root Mean Square (RMS) error for the selected points was 0.3 pixels. To establish relationships between image and field data, stratified random sampling techniques were used to collect 53 soil samples at the GPS (Global Positioning System) points. The continuous map of soil properties was achieved using simple and multiple linear regression models by averaging 9 image pixels around sampling sites. Different image spectral indices were used as independent variables and the dependent variables were field- based data.
Results and Discussion: The results of multiple regression analysis showed that the strongest relationships was between sandy soil and TM bands 1, 2, 3, 4, and 5, explaining up to 83% of variation in this component. The weakest relationship was found between CaCo3 and 3, 5, and 7 TM bands. In some cases, the multiple regressions was not an appropriate predicting model of soil properties, therefore, the TM and PC bands that had the highest relationship with field data (confidence level, 99%) based on simple regression were classified by the maximum likelihood algorithm. According to error matrix, the overall accuracy of classified maps was between 85 and 93% for chlorine (Cl) and silt componets, repectively.
Conclusions: The results indicated that the discretely classified maps had higher accuracy than regression models. Therefore, to have an overview of soil properties in the region, classification techniques appears to be more applicable than regression models. The findings of this study shows that the extracted maps of the physical and chemical characteristics of soils can be used as a suitable tool for field operations, cambating desertification and rehabilitation purposes and compared to maps that are created by traditional methods, our final maps have more economically and time saving advantages. Therefore, they can be used as an adjunct to field methods to aid the assessment and monitoring of soil condition in the arid regions of Isfahan province.
