Minimum Distance Classifier Research Articles

Achieving wheat seedling freezing injury assessment during the seedling stage using Unmanned Ground Vehicle (UGV) and hyperspectral imaging technology

Freezing injury may cause irreversible damage to wheat (Triticum aestivum L) tissues and can significantly reduce yield and quality. Therefore, quick and non-destructively estimating the degree of frost damage for formulating anti-freezing protection strategies and preventing frost damage is very crucial. In this study, we obtained hyperspectral images of wheat leaves for accurate identification of frost damage. A remote-controlled Unmanned Ground Vehicle (UGV) equipped with an imaging spectral camera was used to capture the hyperspectral images of frost-damaged wheat leaves. We compared the efficiency of two methods (the one without removal of weeds, and the other is to remove the corresponding area of weeds from the hyperspectral image by Deeplab V3+) for estimation of wheat freezing damage degree by using four different algorithms; Support Vector Machine Classification (SVM), Mahalanobis Distance Classification (MaD), Minimum Distance Classification (MiD), and Maximum Likelihood Classification (ML). We found that, Deeplab V3+ can efficiently identify the weeds from hyperspectral images, as the overall accuracy (OA) values of different algorithms were high in images with weeds removal as compared to the values in weeds containing images. Further, applying ML model after weeds removal have high OA (93.26 %) as compared to the other models. Thus, using Deeplab V3+ and ML can be a potential approach to identify the freezing injury in wheat for sustainable agricultural productivity.

Radiometric calibration of UAV multispectral images under changing illumination conditions with a downwelling light sensor

AbstractUnmanned aerial vehicle (UAV)‐acquired multispectral images commonly suffer from radiometric inaccuracies due to changing illumination produced by intermittent cloud cover. This study addressed cloud shadow effects by integrating direct irradiance measurements from a downwelling light sensor into reflectance calculations. Two reflectance calibration methods were proposed as follows: (1) use of a single calibration reference panel for all images (Method 1) and (2) employment of a minimum distance classifier to assign color‐graded in‐field reflectance calibration targets to individual images based on their proximity in irradiance levels (Method 2). Images were acquired from 30 and 75 m flight altitudes with fixed‐exposure and auto‐exposure settings. Conventional photogrammetric calibration of UAV multispectral images inadequately addressed cloud shadows, resulting in orthomosaics with poor to moderate spatial uniformity (R2 = 0.70–0.92) and high mean absolute percentage error (MAPE = 13.52%–49.18%). The proposed calibration methods improved radiometric accuracy, yielding average R2 = 0.99 and 0.98 at 30 and 75 m images, respectively. Method 1 showed superior performance on the red‐edge and near infrared bands, and Method 2 worked best in the blue, green, and red bands. Post‐processing empirical line calibration of orthomosaics from Methods 1 and 2 reduced MAPE in reflectance estimation by at least 50% compared to conventional calibration. The clear sky reference histograms had lower Jensen–Shannon distance, higher Pearson's correlation, and improved intersection ratios with the histograms from the proposed methods, implying high similarity. Vegetation indices calculated with the proposed methods closely matched those from the reference orthomosaic, exhibiting significantly lower MAPE than conventionally calculated vegetation indices.

A Comparative Analysis Machine Learning Techniques for LULC Classification Using Landsat-8 Satellite Imagery

This study compares various categorization methods to assign land use and land cover (LULC) classes. Using Geographic Information Systems (GIS) and Remote Sensing (RS) to leverage the dynamic and complex area of LULC, this study examines the potential of different machine learning classification methods. Precise differentiation and classification of various land cover categories, such as green vegetation, urban areas, water bodies, dark green vegetation, and bare terrain, are made possible by the great spatial and spectral resolution of Landsat imagery. For efficient land management and planning, the integration of Landsat data with GIS and RS approaches provides insightful information about the distribution and temporal changes in LULC. This study uses four classifiers to explore the principles of supervised machine learning techniques and identify their drawbacks and advantages. Testing results show that the Support Vector Machine with four kernel linear-99.17%, radial basis (RBF)-99.11%, Sigmoid-99.11% and Polynomial-99.11% is a reliable option for LULC classification, outperforming than other classifiers in terms of accuracy, including the Minimum Distance Classifier (MD-93.47%), Maximum Likelihood Classifier (MLC- 98.98%), and Mahalanobis Distance Classifier (MH-97.83%). Among the tested classifiers, SVM with four kernels notably shows the highest accuracy. With their essential insights for well-informed decision-making towards sustainable development and resource utilization, our findings add to a thorough understanding of LULC dynamics. For accurate mapping and long-term monitoring of deviations in land cover (LC), the study emphasizes the value of using front-line classification systems in remote sensing applications.

HomeOSD: Appliance Operating-Status Detection Using mmWave Radar.

Within the context of a smart home, detecting the operating status of appliances in the environment plays a pivotal role, estimating power consumption, issuing overuse reminders, and identifying faults. The traditional contact-based approaches require equipment updates such as incorporating smart sockets or high-precision electric meters. Non-constant approaches involve the use of technologies like laser and Ultra-Wideband (UWB) radar. The former can only monitor one appliance at a time, and the latter is unable to detect appliances with extremely tiny vibrations and tends to be susceptible to interference from human activities. To address these challenges, we introduce HomeOSD, an advanced appliance status-detection system that uses mmWave radar. This innovative solution simultaneously tracks multiple appliances without human activity interference by measuring their extremely tiny vibrations. To reduce interference from other moving objects, like people, we introduce a Vibration-Intensity Metric based on periodic signal characteristics. We present the Adaptive Weighted Minimum Distance Classifier (AWMDC) to counteract appliance vibration fluctuations. Finally, we develop a system using a common mmWave radar and carry out real-world experiments to evaluate HomeOSD's performance. The detection accuracy is 95.58%, and the promising results demonstrate the feasibility and reliability of our proposed system.

Detecting Changes in Impervious Surfaces Using Multi-Sensor Satellite Imagery and Machine Learning Methodology in a Metropolitan Area

This study utilizes multi-sensor satellite images and machine learning methodology to analyze urban impervious surfaces, with a particular focus on Nanchang, Jiangxi Province, China. The results indicate that combining multiple optical satellite images (Landsat-8, CBERS-04) with a Synthetic Aperture Radar (SAR) image (Sentinel-1) enhances detection accuracy. The overall accuracy (OA) and kappa coefficients increased from 84.3% to 88.3% and from 89.21% to 92.55%, respectively, compared to the exclusive use of the Landsat-8 image. Notably, the Random Forest algorithm, with its unique dual-random sampling technique for fusing multi-sensor satellite data, outperforms other machine learning methods like Artificial Neural Networks (ANNs), Support Vector Machines (SVMs), Classification and Regression Trees (CARTs), Maximum Likelihood Classification (Max-Likelihood), and Minimum Distance Classification (Min-Distance) in impervious surface extraction efficiency. With additional satellite images from 2015, 2017, and 2020, the impervious surface changes are tracked in the Nanchang metropolitan region. From 2015 to 2021, they record a notable increase in impervious surfaces, signaling a quickened urban expansion. This study observes several impervious surface growth patterns, such as a tendency to concentrate near rivers, and larger areas in the east of Nanchang. While the expansion was mainly southward from 2015 to 2021, by 2021, the growth began spreading northward around the Gan River basin.

Minimum Distance Classification for Dynamical Systems and Sensors

Minimum Distance Classification for Dynamical Systems and Sensors

Two-Stage Feature Generator for Handwritten Digit Classification.

In this paper, a novel feature generator framework is proposed for handwritten digit classification. The proposed framework includes a two-stage cascaded feature generator. The first stage is based on principal component analysis (PCA), which generates projected data on principal components as features. The second one is constructed by a partially trained neural network (PTNN), which uses projected data as inputs and generates hidden layer outputs as features. The features obtained from the PCA and PTNN-based feature generator are tested on the MNIST and USPS datasets designed for handwritten digit sets. Minimum distance classifier (MDC) and support vector machine (SVM) methods are exploited as classifiers for the obtained features in association with this framework. The performance evaluation results show that the proposed framework outperforms the state-of-the-art techniques and achieves accuracies of 99.9815% and 99.9863% on the MNIST and USPS datasets, respectively. The results also show that the proposed framework achieves almost perfect accuracies, even with significantly small training data sizes.

Analysis of built-up classes in urbanised zones using radar images

This paper presents the results of a study to determine the potential of radar imaging to detect classes of built-up areas defined in the Urban Atlas (UA) spatial database. The classes are distinguished by function and building density. In addition to the reflectance value itself, characteristics such as building density or spatial layout can improve the identification of these classes. In order to increase the classification possibilities and better exploit the potential of radar imagery, a grey-level co-occurrence matrix (GLCM) was generated to analyse the texture of built-up classes. Two types of synthetic-aperture radar (SAR) images from different sensors were used as test data: Sentinel-1 and ICEYE, which were selected for their different setup configurations and parameters. Classification was carried out using the Random Forests (RF) and Minimum Distance (MD) methods. The use of the MD classifier resulted in an overall accuracy of 64% and 51% for Sentinel-1 and ICEYE, respectively. In ICEYE, individual objects (e.g. buildings) are better recognised than classes defined by their function or density, as in UA classes. Sentinel-1 performed better than ICEYE, with its texture images better complementing the features of urban area classes. This remains a significant challenge due to the complexity of urban areas in defining and characterising urban area classes. Automatic acquisition of training fields directly from UA is problematic and it is therefore advisable to independently obtain reference data for built-up area categories.

A Real-Time Object Counting and Collecting Device for Industrial Automation Process Using Machine Vision

An efficient product packing system through a simplified automatic counting and collecting process is a critical aspect of the manufacturing industry. In the present paper, we propose a real-time tool for simultaneous counting and collection of the objects into different bins using machine vision. The system employs a minimum distance classifier for object detection, Counting on the conveyor belt is done by tracking the Euclidian distances between the centroids of the objects in successive frames. Furthermore, the required number of object collection is carried out in two steps. In the first step, multiple objects nearest to the required count are collected. In the second step, the difference of objects that remained from the previous step are added one by one. Object collection is carried out by integrating software output to the hardware, which collects them in required numbers in the different bins. To meet the fast counting requirements, a high-speed conveyor system constituting vision sensor with pixel resolution of 1920x1200 and 41 frames per second frame rate is employed. The high computational power requirements for implementing the algorithm is accomplished by employing a dedicated graphics microcontroller, whose GPIO pins are used for interfacing the hardware for the object collection. The testing results of the proposed device with 30 variants of different plumbing industry objects confirm an accuracy of 100% for a period of 3 hours. Moreover, the proposed module is scalable and could boost the productivity of the company in a complex environment without affecting the accuracy of the system.

Hyperspectral Remote Sensing Images Feature Extraction Based on Spectral Fractional Differentiation

To extract effective features for the terrain classification of hyperspectral remote-sensing images (HRSIs), a spectral fractional-differentiation (SFD) feature of HRSIs is presented, and a criterion for selecting the fractional-differentiation order is also proposed based on maximizing data separability. The minimum distance (MD) classifier, support vector machine (SVM) classifier, K-nearest neighbor (K-NN) classifier, and logistic regression (LR) classifier are used to verify the effectiveness of the proposed SFD feature, respectively. The obtained SFD feature is sent to the full connected network (FCN) and 1-dimensionality convolutional neural network (1DCNN) for deep-feature extraction and classification, and the SFD-Spa feature cube containing spatial information is sent to the 3-dimensionality convolutional neural network (3DCNN) for deep-feature extraction and classification. The SFD-Spa feature after performing the principal component analysis (PCA) on spectral pixels is directly connected with the first principal component of the original data and sent to 3DCNNPCA and hybrid spectral net (HybridSN) models to extract deep features. Experiments on four real HRSIs using four traditional classifiers and five network models have shown that the extracted SFD feature can effectively improve the accuracy of terrain classification, and sending SFD feature to deep-learning environments can further improve the accuracy of terrain classification for HRSIs, especially in the case of small-size training samples.

TRP-Oriented Hyperspectral Remote Sensing Image Classification Using Entropy-Weighted Ensemble Algorithm

The problem that the randomly generated random projection matrix will lead to unstable classification results is addressed in this paper. To this end, a Tighter Random Projection-oriented entropy-weighted ensemble algorithm is proposed for classifying hyperspectral remote sensing images. In particular, this paper presents a random projection matrix selection strategy based on the separable information of a single class able to project the features of a certain class of objects. The projection result is measured by the degree of separability, thereby obtaining the low-dimensional image with optimal separability of the class. After projecting samples with the same random projection matrix, to calculate the distance matrix, the Minimum Distance classifier is devised, repeating for all classes. Finally, the weight of the distance matrix is considered in ensemble classification by using the information entropy. The proposed algorithm is tested on real hyperspectral remote sensing images. The experiments show an increase in both stability and performance.

Palmprint authentication using symbolic aggregate approximation features with improved enhancement techniques

The palmprint enhancement is a pre-processing stage of palmprint authentication. The accuracy of the recognition rate can be improved by incorporating effective palmprint enhancement methods. So far, only a little work has been done in palmprint enhancement. Little attention is given to incorporate the enhancement techniques in palmprint authentication system to achieve performance improvement. This work presents robust enhancement methods which provide better performance and accuracy. Palmprint enhancement using curvelet and Recursive Histogram Equalization (RHE) overcomes the drawbacks of the existing systems and makes palmprint recognition simpler and more accurate with enhanced palm image as input. Palmprint recognition system uses the Symbolic Aggregate Approximation (SAX) features from the enhanced palmprint image. The recognition rate is optimum when both curvelet and RHE methods are used for enhancement of palmprint. Here we compared different types of enhancement methods and we have obtained maximum recognition rate with a combination of recursive histogram equalization and curvelet transform. The performance of the palmprint authentication is measured by False Acceptance Rate (FAR), False Rejection Rate (FRR) and Total Success Rate (TSR). The comparison of recognition rate with enhancement is done using minimum distance classifier, Support Vector Machine, Random Forest and Bayesian. Random Forest classifier provides better results.

A Possible Land Cover EAGLE Approach to Overcome Remote Sensing Limitations in the Alps Based on Sentinel-1 and Sentinel-2: The Case of Aosta Valley (NW Italy)

Land cover (LC) maps are crucial to environmental modeling and define sustainable management and planning policies. The development of a land cover mapping continuous service according to the new EAGLE legend criteria has become of great interest to the public sector. In this work, a tentative approach to map land cover overcoming remote sensing (RS) limitations in the mountains according to the newest EAGLE guidelines was proposed. In order to reach this goal, the methodology has been developed in Aosta Valley, NW of Italy, due to its higher degree of geomorphological complexity. Copernicus Sentinel-1 and 2 data were adopted, exploiting the maximum potentialities and limits of both, and processed in Google Earth Engine and SNAP. Due to SAR geometrical distortions, these data were used only to refine the mapping of urban and water surfaces, while for other classes, composite and timeseries filtered and regularized stack from Sentinel-2 were used. GNSS ground truth data were adopted, with training and validation sets. Results showed that K-Nearest-Neighbor and Minimum Distance classification permit maximizing the accuracy and reducing errors. Therefore, a mixed hierarchical approach seems to be the best solution to create LC in mountain areas and strengthen local environmental modeling concerning land cover mapping.

A Scalable Earth Observation Service to Map Land Cover in Geomorphological Complex Areas beyond the Dynamic World: An Application in Aosta Valley (NW Italy)

Earth Observation services guarantee continuous land cover mapping and are becoming of great interest worldwide. The Google Earth Engine Dynamic World represents a planetary example. This work aims to develop a land cover mapping service in geomorphological complex areas in the Aosta Valley in NW Italy, according to the newest European EAGLE legend starting in the year 2020. Sentinel-2 data were processed in the Google Earth Engine, particularly the summer yearly median composite for each band and their standard deviation with multispectral indexes, which were used to perform a k-nearest neighbor classification. To better map some classes, a minimum distance classification involving NDVI and NDRE yearly filtered and regularized stacks were computed to map the agronomical classes. Furthermore, SAR Sentinel-1 SLC data were processed in the SNAP to map urban and water surfaces to improve optical classification. Additionally, deep learning and GIS updated datasets involving urban components were adopted beginning with an aerial orthophoto. GNSS ground truth data were used to define the training and the validation sets. In order to test the effectiveness of the implemented service and its methodology, the overall accuracy was compared to other approaches. A mixed hierarchical approach represented the best solution to effectively map geomorphological complex areas to overcome the remote sensing limitations. In conclusion, this service may help in the implementation of European and local policies concerning land cover surveys both at high spatial and temporal resolutions, empowering the technological transfer in alpine realities.

Efficient Classification of Satellite Image with Hybrid Approach Using CNN-CA

Today, satellite imagery is being utilized to help repair and restore societal issues caused by habitats for a variety of scientific studies. Water resource search, environmental protection simulations, meteorological analysis, and soil class analysis may all benefit from the satellite images. The categorization algorithms were used generally and the most appropriate strategies are also be used for analyzing the Satellite image. There are several normal classification mechanisms, such as optimum likelihood, parallel piping or minimum distance classification that have presented in some other existing technologies. But the traditional classification algorithm has some disadvantages. Convolutional neural network (CNN) classification based on CA was implemented in this article. Using the gray level Satellite image as the target and CNN image classification by the CA’s selfiteration mechanism and eventually explores the efficacy and viability of the proposed method in long-term satellite remote sensing image water body classification. Our findings indicate that the proposed method not only has rapid convergence speed, reliability but can also efficiently classify satellite remote sensing images with long-term sequence and reasonable applicability. The proposed technique acquires an accuracy of 91% which is maximum than conventional methods.

Vision-based gait analysis for real-time Parkinson disease identification and diagnosis system

ABSTRACT Computer-assisted Parkinson’s disease-specific gait pattern recognition has gained more attention in the past decade due to its extensive application. In this research study, vision-based gait feature extraction is obtained from the observed skeleton points to support the real-time Parkinson disease prediction and diagnosis in the smart healthcare environment. So, a novel kernel-based principal component analysis (KPCA) is introduced for establishing respective feature extraction and dimensionality reduction on the patient’s video data. In this research study, a vision-based Parkinson disease identification system (VPDIS) is developed with a feature-weighted minimum distance classifier model to support the clinical assessment of Parkinson’s disease. At the time of experimentation, a steady-state walking style of the patient was captured using the cameras fixed in the smart healthcare environment. Then, the accumulated walking frames from the remote patients were transformed into the required binary silhouettes for the sake of noise minimisation and compression purpose. The resulting experimentation shows that the proposed feature extraction approach has significant improvements on the recognition of target gait patterns from the video-based gait analysis of Parkinson’s and normal patients. Accordingly, the proposed VPDIS using feature-weighted minimum distance classifier model provides better prediction time and classification accuracy against the existing healthcare systems that is developed using support vector machine and ensemble learning classifier models.

Sentinel-2 Data for Land Use Mapping: Comparing Different Supervised Classifications in Semi-Arid Areas

Mapping and monitoring land use (LU) changes is one of the most effective ways to understand and manage land transformation. The main objectives of this study were to classify LU using supervised classification methods and to assess the effectiveness of various machine learning methods. The current investigation was conducted in the Nord-Est area of Tunisia, and an optical satellite image covering the study area was acquired from Sentinel-2. For LU mapping, we tested three machine learning models algorithms: Random Forest (RF), K-Dimensional Trees K-Nearest Neighbors (KDTree-KNN) and Minimum Distance Classification (MDC). According to our research, the RF classification provided a better result than other classification models. RF classification exhibited the best values of overall accuracy, kappa, recall, precision and RMSE, with 99.54%, 0.98%, 0.98%, 0.98% and 0.23%, respectively. However, low precision was observed for the MDC method (RMSE = 1.15). The results were more intriguing since they highlighted the value of the bare soil index as a covariate for LU mapping. Our results suggest that Sentinel-2 combined with RF classification is efficient for creating a LU map.

Greenhouses Detection in Guanzhong Plain, Shaanxi, China: Evaluation of Four Classification Methods in Google Earth Engine

Greenhouses used for agricultural production have been expanding around the world because it significantly increases crop yield. Meanwhile, it brings a series of environmental problems that should be considered in agricultural planning and management. The advent of the Google Earth Engine (GEE) cloud platform makes remote sensing image processing more convenient and efficient. It has been widely applied in multiple disciplines, but few studies have investigated the detection of greenhouses. In this research, four different classification methods were applied for comparing their performance in monitoring greenhouses in the Guanzhong Plain, Shaanxi, China using GEE: the Minimum Distance Classifier (MDC), the Support Vector Machine with three kernel functions (linear, SVM-L, polynomial, SVM-P, and radial basis function variations, SVM-R), the Classification and Regression Trees (CART), and the Random Forest (RF). Our results illustrate that these classification techniques’ overall accuracy is >84%. The most accurate classification results were obtained by the SVM-R classifier, with an overall accuracy of 94%, followed by the RF and CART classifier, while the MDC performed worst among these four classifiers. These results would be useful for greenhouse extraction in long time series and large-scale areas, which provides solid information for decision-makers and practitioners for agriculture planning and management.

Extraction of Mountain Grasslands in Yunnan, China, from Sentinel-2 Data during the Optimal Phenological Period Using Feature Optimization

The timely and accurate mapping of the spatial distribution of grasslands is crucial for maintaining grassland habitats and ensuring the sustainable utilization of resources. We used Google Earth Engine (GEE) and Sentinel-2 data for mountain grassland extraction in Yunnan, China. The differences in the normalized vegetation index in the time-series data of different ground objects were compared. February to March, during grassland senescence, was the optimum phenological stage for grassland extraction. The spectral, textural of Sentinel-2, and topographic features of the Shuttle Radar Topography Mission (SRTM) were used for the classification. The features were optimized using the recursive feature elimination (RFE) feature importance selection algorithm. The overall accuracy of the random forest (RF) classification algorithm was 91.2%, the producer’s accuracy of grassland was 96.7%, and the user’s accuracy of grassland was 89.4%, exceeding that of the cart classification (Cart), support vector machine (SVM), and minimum distance classification (MDC). The SWIR1 and elevation were the most important features. The results show that Yunnan has abundant grassland resources, accounting for 18.99% of the land area; most grasslands are located in the northwest at altitudes above 3200 m and in the Yuanjiang River regions. This study provides a new approach for feature optimization and grassland extraction in mountainous areas, as well as essential data for the further investigation, evaluation, protection, and utilization of grassland resources.

Crop type discrimination using Geo-Stat Endmember extraction and machine learning algorithms

The identification of crop diversity in today's world is very crucial to ensure adaptation of the crop with changing climate for better productivity as well as food security. Towards this, Hyperspectral Remote Sensing (HRS) is an efficient technique that offers the opportunity to discriminate crop types based on morphological as well as physiological features due to availability of contiguous spectral bands. The current work utilized the benefits of Airborne Visible Infrared Imaging Spectrometer- New Generation (AVIRIS-NG) data and explored the techniques for classification and identification of crop types. The endmembers were identified using the Geo-Stat Endmember Extraction (GSEE) algorithm for pure pixels identification and to generate the spectral library of the different crop types. Spectral feature comparison was done among AVIRIS-NG, Analytical Spectral Device (ASD)-Spectroradiometer and Continuum Removed (CR) spectra. The best-fit spectra obtained with the Reference ASD-Spectroradiometer and Pure Pixel spectral library were then used for crop discrimination using the ten supervised classifiers namely Spectral Angle Mapper (SAM), Spectral Information Divergence (SID), Support Vector Machine (SVM), Minimum Distance Classifier (MDC), Binary Encoding, deep learning-based Convolution Neural Network (CNN) and different algorithms of Ensemble learning such as Tree Bag, AdaBoost (Adaptive Boosting), Discriminant and RUSBoost (Random Under Sampling). In total, nine crop types were identified, namely, wheat, maize, tobacco, sorghum, linseed, castor, pigeon pea, fennel and chickpea. The performance evaluation of the classifiers was made using various metrics like Overall Accuracy, Kappa Coefficient, Precision, Recall and F1 score. The classifier 2D-CNN was found to be the best with Overall Accuracy, Kappa Coefficient, Precision, Recall and F1 score values of 89.065 %, 0.871, 87.565%, 89.541% and 88.678% respectively. The output of this work can be utilized for large scale mapping of crop types at the species level in a short interval of time with high accuracy.