Knowledge-based Decision Tree Research Articles

Discrimination and monitoring of rice cultural types using dense time series of Sentinel-1 SAR data

Details about the spatial extent of rice cultivation in different seasons are essential for ensuring global food security as well as addressing environmental issues and climate change. This study aims to introduce a novel approach to discriminate between different rice cultural types and subtypes and identify their optimum stages to represent crop growth profiles using Synthetic Aperture Radar (SAR) data. Dense time series Sentinel-1 SAR backscatter data were analyzed at 12 days intervals throughout the growth period of summer and kharif rice. Statistical analysis using ANOVA at 5% level of significance was conducted to identify the critical stages of discrimination for all cultural types and subtypes. Knowledge-based decision tree algorithm was employed to classify all major types of rice grown during summer and kharif season. A comprehensive analysis revealed significant variations in backscatter profiles among different rice types. The most critical period for effectively distinguishing between various rice cultural types and subtypes was found to be from transplanting to tillering stage. Decision tree algorithm was able to discriminate between rice cultural types and subtypes with high overall accuracy and Kappa coefficient, viz., 94.74% and 0.94 for summer rice and 91.80% and 0.90 for kharif rice, respectively. The findings of this study revealed that C-band co and cross-polarized Sentinel-1 data have the potential to detect differences between different cultural types and subtypes of rice and, therefore, will be helpful in crop growth monitoring and agro-ecological change detection in the cultivation pattern.

Discrimination of maize crop in a mixed Kharif crop scenario with synergism of multiparametric SAR and optical data

Accurate mapping of coarse cereals envisaged maize to be discriminated from co-existing Kharif crops using multiplatform, multiparametric SAR and Optical data in various polarization combination from Sentinel-1 and RADARSAT-2 in synergism with Sentinel-2, under machine learning algorithms viz. support vector machine (SVM) and random forest (RF) and knowledge-based Decision Tree (DT). SVM over-estimates the area of maize whereas RF and DT performed similarly (60–75%) but DT being closest to field conditions (68–78%). A 4 date Sentinel-1 under DT classified Soybean to 62% accuracy and Cotton to 76% on a regional scale in Maharashtra. Inclusion of 1-date optical image improved sugarcane accuracy. The similar-structured crops, such as Maize and Pearl Millet were discriminated with their unique phenological response in crop-calendar using quad-pol RADARSAT-2. With inclusion of crop responsive model-based polarimetric decomposition (Yamaguchi Volume and Double-Bounce) and Eigen-based parameters (Entropy and Alpha-Angle), maize classification accuracy elevated to 84%.

Evaluation of different machine learning algorithms for pearl millet discrimination using multi-sensor SAR data

This study aims to assess the potential of multi-temporal Sentinel-1 and RADARSAT-2 data for pearl millet discrimination using various classifiers—machine learning and knowledge-based decision tree (m-DT and k-DT), Support Vector Machine (SVM), and Random Forest (RF). Results show that the RF classifier outperformed the other classifiers in terms of overall accuracy (OA) and kappa coefficient. The highest Overall Accuracy (OA) of 89.1% and kappa coefficients of 0.822 and 0.814 were achieved by the RF classifier for both the Sentinel-1 and RADARSAT-2 data, respectively. The Variable Importance (VI) of the RF classifier revealed that the polarimetric parameters—volume scattering, double bounce, entropy and alpha angle were crucial for pearl millet discrimination. Polarimetric parameters, when ingested with the RF machine learning classifier, achieved better classification accuracies.

Mapping and Area Estimation of Mango Orchards of Lucknow Region by Applying Knowledge Based Decision Tree to Landsat 8 OLI Satellite Images

Mango is a very important fruit which is liked by majority of the population due to its nutritional value and excellent taste. India is the largest producer of mango in the world. Accurate information is required for policy decision making in terms of providing subsidy, area expansion, and crop insurance planning. Hence, this type of information may be retrieve through satellite images by using the image classification techniques, which are playing a crucial role in crop cover classification, yield prediction and crop monitoring etc. Classification of optical satellite images is still a challenging task due to effect of changing atmospheric conditions such as cloud, snow, haze, dust, fog, and rain etc. In this paper, knowledge based decision tree classification (DTC) has been proposed to classify the mango orchards of Lucknow district using multi-temporal Landsat 8 operational land imager (OLI) images from year 2015 to 2017 and further mango orchard area were also estimated. In order to develop the DTC, separability analysis for various land cover classes was carried out on different vegetation indices namely, normalized difference vegetation index (NDVI), modified normalized difference water index (MNDWI), and soil adjusted vegetation index (SAVI). In order to analyze the performance of DTC, most commonly used satellite image classifiers such as unsupervised classifier (i.e. ISODATA) and supervised classifier (i.e. Maximum Likelihood) have been used and it is observed that the proposed DTC outperformed these traditional classifiers. Also, accuracy assessment has been carried out to measure the performance of proposed DTC and it is observed that all of the three images from 2015 to 2017 are classified with high overall accuracy, which is ranging from 70.66% to 86.69%. Kappa Coefficient (KC) for all the three images ranged from 0.65 to 0.83, which indicates that classified images are highly acceptable for area estimation.

Development of a knowledge based decision tree classifier using hybrid polarimetric SAR observables

ABSTRACTFast, efficient and accurate classification of the land cover using Synthetic aperture radar (SAR) observables extracted from hybrid polarimetric SAR data is achieved in this research. The proposed knowledge-based tree classifier utilizes little apriori real-time field survey information along with just four features, namely, backscattering coefficient (σrh+σrv), scattering mechanism (α), diffuse scattering and odd bounce (surface) scattering feature from m- α decomposition method in a sequential manner. We have exploited the Separability Index (SI) criterion for identifying these 4 features among the available 16 features. The overall accuracy (OA), user accuracy (UA), producer accuracy (PA), kappa coefficient (κ), precision (P), recall (R) and score (F) of the proposed classifier underscores its merits. Further, for the sake of fair comparison with the existing approaches, we have built the layer stacked images using these four features and applied them to the supervised maximum likelihood estimator classifier (MLE) as well as the unsupervised k-means classifier. It is found that the proposed classifier has better performance in terms of OA, UA, PA and κ on different SAR data sets consisting of different areas.

Voxel-based 3D Point Cloud Semantic Segmentation: Unsupervised Geometric and Relationship Featuring vs Deep Learning Methods

Automation in point cloud data processing is central in knowledge discovery within decision-making systems. The definition of relevant features is often key for segmentation and classification, with automated workflows presenting the main challenges. In this paper, we propose a voxel-based feature engineering that better characterize point clusters and provide strong support to supervised or unsupervised classification. We provide different feature generalization levels to permit interoperable frameworks. First, we recommend a shape-based feature set (SF1) that only leverages the raw X, Y, Z attributes of any point cloud. Afterwards, we derive relationship and topology between voxel entities to obtain a three-dimensional (3D) structural connectivity feature set (SF2). Finally, we provide a knowledge-based decision tree to permit infrastructure-related classification. We study SF1/SF2 synergy on a new semantic segmentation framework for the constitution of a higher semantic representation of point clouds in relevant clusters. Finally, we benchmark the approach against novel and best-performing deep-learning methods while using the full S3DIS dataset. We highlight good performances, easy-integration, and high F1-score (> 85%) for planar-dominant classes that are comparable to state-of-the-art deep learning.

SURFACE WATER DYNAMICS OF INLAND WATER BODIES OF INDIA USING GOOGLE EARTH ENGINE

Abstract. Dynamics, distribution and quality of water has a direct impact on environment and its dependent human activities. Regular monitoring of these hydrological processes help in understanding water cycle and better management policy making. Recent increase in remote sensing satellites offer multiple observations with high spatial and temporal resolution, thus calling for extensive use of high end computational resources. Google Earth Engine(GEE) is an open Application Programing Interface (API), which offers free computational resources and satellite data on cloud computational platform minimising the users need for computational resources and data availability. Five year Landsat-8 imagery (2013–18) from GEE database has been used to study the surface water extent of large inland water bodies (surface area greater than 6000 ha) of India. We have used a pixel based classification system to delineate water and non-water pixels. A knowledge based Decision Tree (DT) model has been employed to cluster the classes according to Normalized Difference Vegetation Index (NDVI) and Modified Normalized Difference Water Index (MNDWI) distribution. We report an anomalous departure from the 5-year trend line suggesting that the maximum decrease of water extent was found in year 2015–2016. Analysis of the decay pattern of reservoirs can provide timely inputs for better policy making and management of water resources. To understand the decay pattern, a Modified Gaussian model fit on time series of surface extent helps to determine maximum water extent, peak extent day and storage cycle of the water body.

An object-based approach for tree species extraction from digital orthophoto maps

Abstract. Tree segmentation is an active and ongoing research area in the field of photogrammetry and remote sensing. It is more challenging due to both intra-class and inter-class similarities among various tree species. In this study, we exploited various statistical features for extraction of hazelnut trees from 1 : 5000 scaled digital orthophoto maps. Initially, the non-vegetation areas were eliminated using traditional normalized difference vegetation index (NDVI) followed by application of mean shift segmentation for transforming the pixels into meaningful homogeneous objects. In order to eliminate false positives, morphological opening and closing was employed on candidate objects. A number of heuristics were also derived to eliminate unwanted effects such as shadow and bounding box aspect ratios, before passing them into the classification stage. Finally, a knowledge based decision tree was constructed to distinguish the hazelnut trees from rest of objects which include manmade objects and other type of vegetation. We evaluated the proposed methodology on 10 sample orthophoto maps obtained from Giresun province in Turkey. The manually digitized hazelnut tree boundaries were taken as reference data for accuracy assessment. Both manually digitized and segmented tree borders were converted into binary images and the differences were calculated. According to the obtained results, the proposed methodology obtained an overall accuracy of more than 85 % for all sample images.

GEO-CEOS stage 4 validation of the Satellite Image Automatic Mapper lightweight computer program for ESA Earth observation level 2 product generation - Part 2: Validation.

ESA defines as Earth Observation (EO) Level 2 information product a multi-spectral (MS) image corrected for atmospheric, adjacency, and topographic effects, stacked with its data-derived scene classification map (SCM), whose legend includes quality layers cloud and cloud-shadow. No ESA EO Level 2 product has ever been systematically generated at the ground segment. To fill the information gap from EO big data to ESA EO Level 2 product in compliance with the GEO-CEOS stage 4 validation (Val) guidelines, an off-the-shelf Satellite Image Automatic Mapper (SIAM) lightweight computer program was selected to be validated by independent means on an annual 30 m resolution Web-Enabled Landsat Data (WELD) image composite time-series of the conterminous U.S. (CONUS) for the years 2006 to 2009. The SIAM core is a prior knowledge-based decision tree for MS reflectance space hyperpolyhedralization into static (non-adaptive to data) color names. For the sake of readability, this paper was split into two. The present Part 2—Validation—accomplishes a GEO-CEOS stage 4 Val of the test SIAM-WELD annual map time-series in comparison with a reference 30 m resolution 16-class USGS National Land Cover Data (NLCD) 2006 map. These test and reference map pairs feature the same spatial resolution and spatial extent, but their legends differ and must be harmonized, in agreement with the previous Part 1 - Theory. Conclusions are that SIAM systematically delivers an ESA EO Level 2 SCM product instantiation whose legend complies with the standard 2-level 4-class FAO Land Cover Classification System (LCCS) Dichotomous Phase (DP) taxonomy.

Recurrent Transformation of Prior Knowledge Based Model for Human Motion Recognition.

Motion related human activity recognition using wearable sensors can potentially enable various useful daily applications. So far, most studies view it as a stand-alone mathematical classification problem without considering the physical nature and temporal information of human motions. Consequently, they suffer from data dependencies and encounter the curse of dimension and the overfitting issue. Their models are hard to be intuitively understood. Given a specific motion set, if structured domain knowledge could be manually obtained, it could be used for better recognizing certain motions. In this study, we start from a deep analysis on natural physical properties and temporal recurrent transformation possibilities of human motions and then propose a useful Recurrent Transformation Prior Knowledge-based Decision Tree (RT-PKDT) model for recognition of specific human motions. RT-PKDT utilizes temporal information and hierarchical classification method, making the most of sensor streaming data and human knowledge to compensate the possible data inadequacy. The experiment results indicate that the proposed method performs superior to those adopted in related works, such as SVM, BP neural networks, and Bayesian Network, obtaining an accuracy of 96.68%.

GEO-CEOS stage 4 validation of the Satellite Image Automatic Mapper lightweight computer program for ESA Earth observation level 2 product generation - Part 1: Theory.

ESA defines as Earth Observation (EO) Level 2 information product a single-date multi-spectral (MS) image corrected for atmospheric, adjacency and topographic effects, stacked with its data-derived scene classification map (SCM), whose legend includes quality layers cloud and cloud-shadow. No ESA EO Level 2 product has ever been systematically generated at the ground segment. To fill the information gap from EO big data to ESA EO Level 2 product in compliance with the GEO-CEOS stage 4 validation (Val) guidelines, an off-the-shelf Satellite Image Automatic Mapper (SIAM) lightweight computer program was validated by independent means on an annual 30 m resolution Web-Enabled Landsat Data (WELD) image composite time-series of the conterminous U.S. (CONUS) for the years 2006–2009. The SIAM core is a prior knowledge-based decision tree for MS reflectance space hyperpolyhedralization into static color names. Typically, a vocabulary of MS color names in a MS data (hyper)cube and a dictionary of land cover (LC) class names in the scene-domain do not coincide and must be harmonized (reconciled). The present Part 1—Theory provides the multidisciplinary background of a priori color naming. The subsequent Part 2—Validation accomplishes a GEO-CEOS stage 4 Val of the test SIAM-WELD annual map time-series in comparison with a reference 30 m resolution 16-class USGS National Land Cover Data 2006 map, based on an original protocol for wall-to-wall thematic map quality assessment without sampling, where the test and reference maps feature the same spatial resolution and spatial extent, but whose legends differ and must be harmonized.

Knowledge-based decision tree approach for mapping spatial distribution of rice crop using C-band synthetic aperture radar-derived information

Updated and accurate information of rice-growing areas is vital for food security and investigating the environmental impact of rice ecosystems. The intent of this work is to explore the feasibility of dual-polarimetric C-band Radar Imaging Satellite-1 (RISAT-1) data in delineating rice crop fields from other land cover features. A two polarization combination of RISAT-1 backscatter, namely ratio (HH/HV) and difference (HH−HV), significantly enhanced the backscatter difference between rice and nonrice categories. With these inputs, a QUEST decision tree (DT) classifier is successfully employed to extract the spatial distribution of rice crop areas. The results showed the optimal polarization combination to be HH along with HH/HV and HH−HV for rice crop mapping with an accuracy of 88.57%. Results were further compared with a Landsat-8 operational land imager (OLI) optical sensor-derived rice crop map. Spatial agreement of almost 90% was achieved between outputs produced from Landsat-8 OLI and RISAT-1 data. The simplicity of the approach used in this work may serve as an effective tool for rice crop mapping.

Energy crop mapping with enhanced TM/MODIS time series in the BCAP agricultural lands

Since the mid-2000s, agricultural lands in the United States have been undergoing rapid change to meet the increasing bioenergy demand. In 2009 the USDA Biomass Crop Assistance Program (BCAP) was established. In its Project Area 1, land owners are financially supported to grow perennial prairie grasses (switchgrass) in their row-crop lands. To promote the program, this study tested the feasibility of biomass crop mapping based on unique timings of crop development. With a previously published data fusion algorithm - the Enhanced Spatial and Temporal Adaptive Reflectance Fusion Model (ESTARFM), a 10-day normalized difference vegetation index (NDVI) time series in 2007 was established by fusing MODIS reflectance into TM image series. Two critical dates - peak growing (PG) and peak drying (PD) - were extracted and a unique “PG-0-PD” timing sequence was defined for each crop. With a knowledge-based decision tree approach, the classification of enhanced TM/MODIS time series reached an overall accuracy of 76% against the USDA Crop Data layer (CDL). Especially, our results showed that winter wheat single cropping and wheat-soybean double cropping were much better classified, which may provide additional information for the CDL product. More importantly, this study extracted the first spatial layer of warm-season prairie grasses that have not been published in any national land cover products, which could serve as a base map for decision making of bioenergy land use in BCAP land.

Classification of land-cover types in muddy tidal flat wetlands using remote sensing data

Remote sensing classification of tidal flat wetlands is important for obtaining high-precision information on wetland features. In this study, Thematic Mapper (TM) images of the Yancheng National Reserves, Jiangsu Province, China, for the years of 1996, 2002, 2006, and 2010 were considered. First, the optimum combination of bands was chosen. Second, vegetation and nonvegetation regions of interest were established to investigate the spectral reflectance characteristics of the different ground objects. Then we used the knowledge-based decision tree method on different features, such as the normalized difference vegetation index and the spectral reflectance. In particular, the ancillary information is helpful to distinguish the vegetation classes. The results demonstrate that the classification system has advantages in identifying the types of vegetation in ecotones, and it is 4 percentage points higher than the maximum likelihood method in classification accuracy. This study is useful to discriminate vegetation, and it provides an important reference for the effective extraction of tidal flat land-cover information from TM images.

Ten-Year Landsat Classification of Deforestation and Forest Degradation in the Brazilian Amazon

Forest degradation in the Brazilian Amazon due to selective logging and forest fires may greatly increase the human footprint beyond outright deforestation. We demonstrate a method to quantify annual deforestation and degradation simultaneously across the entire region for the years 2000–2010 using high-resolution Landsat satellite imagery. Combining spectral mixture analysis, normalized difference fraction index, and knowledge-based decision tree classification, we mapped and assessed the accuracy to quantify forest (0.97), deforestation (0.85) and forest degradation (0.82) with an overall accuracy of 0.92. We show that 169,074 km2 of Amazonian forest was converted to human-dominated land uses, such as agriculture, from 2000 to 2010. In that same time frame, an additional 50,815 km2 of forest was directly altered by timber harvesting and/or fire, equivalent to 30% of the area converted by deforestation. While average annual outright deforestation declined by 46% between the first and second halves of the study period, annual forest degradation increased by 20%. Existing operational monitoring systems (PRODES: Monitoramento da Florestal Amazônica Brasileira por Satélite) report deforestation area to within 2% of our results, but do not account for the extensive forest degradation occurring throughout the region due to selective logging and forest fire. Annual monitoring of forest degradation across tropical forests is critical for developing land management policies as well as the monitoring of carbon stocks/emissions and protected areas.

Impacts of land use and land cover changes on evapotranspiration and runoff at Shalamulun River watershed, China

The study assesses the effect of land use and land cover changes (LUCC) on evapotranspiration and runoff in the Shalamulun River watershed of 2,453 km2 located in Inner Mongolia Autonomic Region of China. First, Landsat Thematic Mapper (TM) and Enhanced Thematic Mapper Plus (ETM+) images from 1987, 2001 and 2007 are used to quantify LUCC in the watershed. A knowledge-based decision tree (K-DT) classification technique is used to detect LUCC. By comparison of post-classification change among 1987, 2001 and 2007, the results showed significant modification and conversion of land use and cover of the watershed over the 20-year period 1987–2007. The results show that the forest area underwent the greatest change, decreasing by 159.2 km2 in the study period. At the same time, the area of farmland, barren land and residential land increased by 89.5, 46.4 and 25.3 km2, respectively. Subsequently, a two-source potential evapotranspiration (PET) model is used to estimate the potential evapotranspiration response to LUCC. Finally, the influence of LUCC on annual runoff is evaluated using a statistical method. LUCC potentially caused a decrease in annual PET and runoff. Meanwhile, the land use changes resulted in spatio-temporal variations of monthly PET in the growing season (April–September).

THE APPLICATION OF THE KNOWLEDGE-BASED DECISION TREE CLASSIFICATION METHOD TO THE EXTRACTION OF LAND TYPES IN MINING AREAS: A CASE STUDY OF DAYE AREA, HUBEI PROVINCE

Simple and clear, the knowledge-based decision tree classification method can choose the best band composition and characteristic parameters of different ground objects so as to get the highly accurate classification results. Based on the spectral characteristics and the spatial knowledge of the ground objects in Daye which served as a training area, the authors used image composite, iron oxide index, normalized difference vegetation index (NDVI) and digital elevation model and employed the decision tree classification method with multitemporal Landsat TM (ETM) images. The classification algorithm was applied to all the Landsat TM (ETM) data so as to detect temporal and spatial changes in the mining areas, which, in turn, were divided into ten classes. The characteristics of the highly accurate classification results enable us to perform highly accurate change detection and quantitative analysis of such features in different mining areas as waste,water bodies, change of land use, reclamation process and estimation of vegetation cover in affected places. From the change detection results, it is observed that the decreasing vegetation and land degradation caused by mining activities in the study area are serious, and that only about 35% of the abandoned mining area was reclaimed from 1986 to 2002.

Including the effects of filamentous bulking sludge during the simulation of wastewater treatment plants using a risk assessment model

The main objective of this paper is to demonstrate how including the occurrence of filamentous bulking sludge in a secondary clarifier model will affect the predicted process performance during the simulation of WWTPs. The IWA Benchmark Simulation Model No. 2 (BSM2) is hereby used as a simulation case study. Practically, the proposed approach includes a risk assessment model based on a knowledge-based decision tree to detect favourable conditions for the development of filamentous bulking sludge. Once such conditions are detected, the settling characteristics of the secondary clarifier model are automatically changed during the simulation by modifying the settling model parameters to mimic the effect of growth of filamentous bacteria. The simulation results demonstrate that including effects of filamentous bulking in the secondary clarifier model results in a more realistic plant performance. Particularly, during the periods when the conditions for the development of filamentous bulking sludge are favourable – leading to poor activated sludge compaction, low return and waste TSS concentrations and difficulties in maintaining the biomass in the aeration basins – a subsequent reduction in overall pollution removal efficiency is observed. Also, a scenario analysis is conducted to examine i) the influence of sludge retention time (SRT), the external recirculation flow rate ( Q r) and the air flow rate in the bioreactor (modelled as k La) as factors promoting bulking sludge, and ii) the effect on the model predictions when the settling properties are changed due to a possible proliferation of filamentous microorganisms. Finally, the potentially adverse effects of certain operational procedures are highlighted, since such effects are normally not considered by state-of-the-art models that do not include microbiology-related solids separation problems.