Selection Of Training Samples Research Articles

Change Detection for Heterogeneous Remote Sensing Images with Improved Training of Hierarchical Extreme Learning Machine (HELM)

To solve the problems of susceptibility to image noise, subjectivity of training sample selection, and inefficiency of state-of-the-art change detection methods with heterogeneous images, this study proposes a post-classification change detection method for heterogeneous images with improved training of hierarchical extreme learning machine (HELM). After smoothing the images to suppress noise, a sample selection method is defined to train the HELM for each image, in which the feature extraction is respectively implemented for heterogeneous images and the parameters need not be fine-tuned. Then, the multi-temporal feature maps extracted from the trained HELM are segmented to obtain classification maps and then compared to generate a change map with changed types. The proposed method is validated experimentally by using one set of synthetic aperture radar (SAR) images obtained from Sentinel-1, one set of optical images acquired from Google Earth, and two sets of heterogeneous SAR and optical images. The results show that compared to state-of-the-art change detection methods, the proposed method can improve the accuracy of change detection by more than 8% in terms of the kappa coefficient and greatly reduce run time regardless of the type of images used. Such enhancement reflects the robustness and superiority of the proposed method.

Bagging-gradient boosting decision tree based milling cutter wear status prediction modelling

Aiming to monitor wear condition of milling cutters in time and provide tool change decisions to ensure manufacturing safety and product quality, a tool wear monitoring model based on Bagging-Gradient Boosting Decision Tree (Bagging-GBDT) is proposed. In order to avoid incomplete tool state information contained in a single domain feature parameter, a multi-domain combination method is used to extract candidate characteristic parameter sets from time domain, frequency domain, and time–frequency domain. Then top 21 significant features are screened by eXtreme Gradient Boosting selection method. Synthetic Minority Oversampling Technique technology is integrated during feature selection to overly sample feature vectors, so that wear condition categories can be well balanced. Bagging idea is then introduced for parallel calculation of the gradient boosting decision tree and to improve its generalization ability. A Bagging-GBDT milling cutter wear condition prediction model is constructed and verified by public ball-end milling data set. Experiments show that random features and training samples selection can effectively improve prediction performance and generalization ability of prediction model. Our Bagging-GBDT model gains F1 score of 0.99350, which is 0.2% and 13.2% higher than the random forest algorithm and basic GBDT model, respectively.

Iterative Training Sampling Coupled With Active Learning for Semisupervised Spectral–Spatial Hyperspectral Image Classification

Training sample selection is a great challenge for hyperspectral image classification (HSIC), specifically when only a very limited number of labeled data samples are available for training. Two recently developed concepts for training sample selection are of particular interest. One is active learning (AL), which augments labeled training samples by including unlabeled data samples as new training samples. The other is iterative training sampling (ITS), which expands data cubes by including additional spatial classification information into the training samples. This article combines AL and ITS simultaneously to derive a joint ITS–AL spectral–spatial (SS) classification approach, to be called ITS augmentation by AL spectral–spatial classification, referred to as ITSA-AL-SS, which can improve SS classification using either AL or ITS alone. The novel idea of ITSA-AL-SS is to take advantage of ITS to expand data cubes by incorporating additional spatial classification information iteratively into the AL-augmented unlabeled data samples to update the current training samples iteration by iteration in one-shot operation. As expected, ITSA-AL-SS is benefited from both ITS and AL to not only further improve classification accuracy but also reduce classification inconsistency.

A novel automatic phenology learning (APL) method of training sample selection using multiple datasets for time-series land cover mapping

The long record of Landsat imagery, which is the cornerstone of Earth observation, provides an opportunity to monitor land use and land cover (LULC) change and understand the interactions between the climate and earth system through time. A few change detection algorithms such as Continuous Change Detection and Classification (CCDC) have been developed to utilize all available Landsat images for change detection and characterization at local or global scales. However, the reliable, rapid, and reproducible collection of training samples have become a challenge for time series land cover classification at a large scale. To meet the challenge, we proposed an automatic phenology learning (APL) method with the assumption that the temporal profiles of samples within the same land cover type are the same or similar at a local scale to generate evenly distributed training samples automatically. We designed the method to build land cover patterns for each category based on consensus samples derived from multiple existing scientific datasets including LANDFIRE's (LF) Existing Vegetation Type (EVT), USGS National Land Cover Database (NLCD), National Agricultural Statistics Service (NASS) Cropland Data Layer (CDL), and National Wetlands Inventory (NWI). Then we calculated the Time-Weighted Dynamic Time Warping (twDTW) distance between any undefined samples and land cover patterns in the same geographical region as prior knowledge. Finally, we selected the optimal land cover category for each undefined sample from the land cover products based on the designed criteria iteratively using the twDTW distance as an indicator. The method was applied in the footprint of 10 selected Landsat Analysis Ready Data (ARD) tiles in the eastern and western conterminous United States (CONUS) to produce annual land cover maps from 1985 to 2017. The accuracy assessment and visual comparison revealed that the APL method can generate reliable training samples without any manual interpretation, producing better land cover results especially for the grass/shrub and wetland land cover classes. Applying the APL method, the overall accuracy of the annual land cover maps was improved by 2% over the accuracy of Land Change Monitoring, Assessment, and Projection (LCMAP) Collection 1.0 Science Products in the research regions. Our results also indicate that the APL method provides an approach for best use of different land cover products and meets the requirement of intensive sampling for training data collection.

Bayesian Support Vector Regression for Reliability-Based Design Optimization

Solving reliability-based design optimization (RBDO) by combining surrogate models is a powerful tool to deal with the output variation induced by uncertainties during actual engineering design. This paper aims to develop a strategy for solving RBDO problems by support vector regression (SVR) under the Bayesian inference, referred to as Bayesian SVR (BSVR). The BSVR model possesses the features of original SVR as well as providing the prediction variance to direct the sequential sampling process for probabilistic analysis in RBDO. In the meanwhile, a learning function combined with the sample pool truncation strategy is proposed to select the new training samples to adaptively update the BSVR model, which can accurately and efficiently approximate the important probabilistic constraint boundaries in the desired regions, and the newly selected training samples tend to be far away from the existing training points under the current design to avoid clustering. The computational capability and the good engineering applicability of the proposed method are verified by three numerical examples and two engineering applications about a stiffened rib of the wing edge and a latch-lock mechanism of the cabin hatch.

Training sample selection: Impact on screening automation in diagnostic test accuracy reviews.

When performing a systematic review, researchers screen the articles retrieved after a broad search strategy one by one, which is time‐consuming. Computerised support of this screening process has been applied with varying success. This is partly due to the dependency on large amounts of data to develop models that predict inclusion. In this paper, we present an approach to choose which data to use in model training and compare it with established approaches. We used a dataset of 50 Cochrane diagnostic test accuracy reviews, and each was used as a target review. From the remaining 49 reviews, we selected those that most closely resembled the target review's clinical topic using the cosine similarity metric. Included and excluded studies from these selected reviews were then used to develop our prediction models. The performance of models trained on the selected reviews was compared against models trained on studies from all available reviews. The prediction models performed best with a larger number of reviews in the training set and on target reviews that had a research subject similar to other reviews in the dataset. Our approach using cosine similarity may reduce computational costs for model training and the duration of the screening process.

Effectiveness of Training Sample and Features for Random Forest on Road Extraction from Unmanned Aerial Vehicle-Based Point Cloud

The accuracy of random forest (RF) classification depends on several inputs. In this study, two primary inputs—training sample and features—are evaluated for road classification from an unmanned aerial vehicle-based point cloud. Training sample selection is a challenging step since the machine learning stage of the RF classification depends greatly on it. That is, an imbalanced training sample might dramatically decrease classification accuracy. Various criteria are defined to generate different types of training samples to evaluate the effectiveness of the training sample. There are several point features that can be used in RF classification under different circumstances. More features might increase the classification accuracy, however, in that case, the processing time is also increased. Point features such as RGB (red/green/blue), surface normals, curvature, omnivariance, planarity, linearity, surface variance, anisotropy, verticality, and ground/non-ground class are investigated in this study. Different training samples and sets of features are used in the RF to extract the road surface. The experiment is conducted on a local road without a raised curb located on a relatively steep hill. The accuracy assessment is conducted by comparing the model classification results with the manually extracted road surface point cloud. It is found that the accuracy increases up to around 4%–13%, and 95% overall accuracy was obtained when using convenient training samples and features.

Irregular urban Expansion and Its Effects on Air Temperature over Baghdad City using Remote Sensing Technique

Heat island is known as the increases in air temperature through large and industrial cities compared to surrounding rural areas. In this study, remote sensing technology is used to monitor and track thermal variations within the city center of Baghdad through Landsat satellite images and for the period from 2000 to 2015. Several processors and treatments were applied on these images using GIS 10.6 and ERDAS 2014, such as image correction and extraction, supervised classification, and selection of training samples. Urban areas detection was resulted from the supervised classification linked to the temperature readings of the surface taken from the thermal bands of satellite images. The results showed that the surface temperature of the city of Baghdad increased by 8 degrees Celsius in 15 years. This is due to the increase in the expansion of the urban areas type of land use, where the human activity, especially after 2003, caused increased buildup area to about 198.41 km2. All these changes occurred at the expense of many green regions which were reduced, with the transformation of open and agricultural areas to residential, commercial, and industrial uses. Increases in surface temperature resulted increases in air temperature, where the minimum temperature showed larger increases relative to maximum temperature (about 1.44 and 0.76 ºC, respectively).

A training sample selection method based on united generalised inner product statistics for STAP

A training sample selection method based on united generalised inner product statistics for STAP

A generalised eigenvalue reweighting covariance matrix estimation algorithm for airborne STAP radar in complex environment

To improve the space-time adaptive processing (STAP) performance of airborne radar in complex environment, a generalised eigenvalue reweighting covariance matrix estimation algorithm called GERCM is proposed here. First, the interference plus noise (IPN) covariance matrix of cell under test (CUT) data is estimated by the selected target-free training samples around the CUT with the sample covariance matrix method. Then, with the component decompositions of the selected training samples and the assumption of approximately equal subspace, the IPN covariance matrix of CUT data is reformulated by the eigenvector matrix, eigenvalue matrix, and the eigenvalue reweighting vector. Subsequently, based on the modified covariance matching estimation criterion, the eigenvalue reweighting vector is estimated by solving the redesigned convex optimisation problem with the Lagrange dual method. Finally, the STAP weight vector is calculated to process the CUT data. The proposed algorithm can obtain a relatively accurate IPN covariance matrix of CUT data by sufficiently utilising the non-homogeneous training samples and can effectively protect the moving targets in CUT data, which can be applied to airborne radar with arbitrary array structure and antenna configuration. Simulation results and performance analyses based on the multi-channel airborne radar measurement data demonstrate the effectiveness of the proposed GERCM algorithm.

ANALYSIS OF THE STATE OF PLANT COMMUNITIES IN THE DOLINSKY NATURE SANCTUARY (SAKHALIN) ACCORDING TO SATELLITE SURVEYS

The determination ratio of timber tree in the composition of stands makes it possible to judge about geoecological state of the vegetation cover. It be noted heterogeneous studying of forests in different regions of the country, in connection with which the research of specially protected natural areas (SPNA), which includes the Dolinsky Nature Sanctuary, with an insufficient number of employees in the forest industry, becomes especially relevant. Automated decoding of remote sensing data serves as an aid to ground-based observations. Work to identify the species composition of forests according to aerospace surveys on the territory of Sakhalin Island previously carried out in small volumes. The Sentinel-2 satellite images of the territory of the Dolinsky Nature Sanctuary, obtained in 2018-2021, analyzed in order to identify the classification features of various plant communities characteristic of the southern part of Sakhalin. On basis of visual interpretation, reference sites for geobotanical studies were identified, which with a high degree of confidence attributed to specific plant formations: fir-spruce, stone-birch, willow-alder river valleys, bamboo, deforested areas and plantings of pine and larch crops. Further work carried out using the methods of geoinformation analysis in ArcGIS. The optimal combinations of channels in the formation of synthesized image composites for the selection of training samples for each plant formation were determined when performing vegetation classification. In case of difficulties with the separation of some formations (larch - pine) in the images, the work carried out based on the materials of multi-season surveys. The obtained characteristics of the training samples can be used for further research in the Dolinsky Nature Sanctuary and beyond.

Object Detection in Drone Imagery via Sample Balance Strategies and Local Feature Enhancement

With the advent of drones, new potential applications have emerged for the unconstrained analysis of images and videos from aerial view cameras. Despite the tremendous success of the generic object detection methods developed using ground-based photos, a considerable performance drop is observed when these same methods are directly applied to images captured by Unmanned Aerial Vehicles (UAVs). Usually, most of the work goes into improving the performance of the detector in aspects such as design loss, training sample selection, feature enhancement, and so forth. This paper proposes a detection framework based on an anchor-free detector with several modules, including a sample balance strategies module and super-resolved generated feature module, to improve performance. We proposed the sample balance strategies module to optimize the imbalance among training samples, especially the imbalance between positive and negative, and easy and hard samples. Due to the high frequencies and noisy representation of the small objects in images captured by drones, the detection task is extraordinarily challenging. However, when compared with other algorithms of this kind, our method achieves better results. We also propose a super-resolved generated GAN (Generative Adversarial Network) module with center-ness weights to effectively enhance the local feature map. Finally, we demonstrate our method’s effectiveness with the proposed modules by carrying out a state-of-the-art performance on Visdrone2020 benchmarks.

New active learning algorithms for near-infrared spectroscopy in agricultural applications

Abstract The selection of training data determines the quality of a chemometric calibration model. In order to cover the entire parameter space of known influencing parameters, an experimental design is usually created. Nevertheless, even with a carefully prepared Design of Experiment (DoE), redundant reference analyses are often performed during the analysis of agricultural products. Because the number of possible reference analyses is usually very limited, the presented active learning approaches are intended to provide a tool for better selection of training samples.

Short-term Photovoltaic Power Prediction Based on IFCM and BA-Elman

The traditional fuzzy C-means clustering algorithm (FCM) based on Euclidean distance is only applicable to clustering of spherical structures. When applied to text clustering of photovoltaic data to select similar days, it fails to take into account the difference in the importance of meteorological factors on photovoltaic power, resulting in a decrease in the accuracy and efficiency of data clustering. To solve the above problems, a short-term photovoltaic power prediction method based on improved fuzzy C-means clustering (IFCM) and bat algorithm optimization Elman neural network (BA-Elman) combination model was proposed. First, the improved fuzzy C-means clustering algorithm is used to select training samples with higher similarity to the forecast day, and then the Elman neural network prediction model optimized by bat algorithm trained by selected training samples. Finally, according to the actual data of a photovoltaic power station in Qinghai Province, a simulation experiment is carried out to verify the effectiveness of the method and model.

Covariance Matrix Whitening-Based Training Sample Selection Method for Airborne Radar

As training samples are not always target-free in space-time processing for airborne radar, the traditional methods usually use the sample covariance matrix (SCM) as the test covariance matrix (TCM) to censor contaminated training samples. However, the SCM cannot represent the property of the cell under test (CUT) accurately, resulting in low selection efficiency. To deal with this problem, this letter proposes a novel training sample selection method based on covariance matrix whitening. Specifically, we utilize the reconstructed subaperture's clutter covariance matrix (RSCCM) of the CUT as the TCM. The RSCCM is only determined by the CUT and can characterize the CUT directly. Then, we use the RSCCM to whiten the subaperture's covariance matrix of the training sample. A criterion for selecting the training samples is derived based on the maximum eigenvalue of the whitened subaperture's covariance matrix, which is related to the energy of the outliers and more stable than the statistic of the generalized inner product method. Simulations are conducted to evaluate the performance of the proposed method.

Class label altering fuzzy min-max network and its application to histopathology image database

Hyperbox classifier is efficiently implemented using fuzzy min max neural network, where the input patterns present in the training phase place a vital role. In the training phase, a set of hyperboxes are constructed which are used to classify a testing pattern. A better selection of input patterns of the training set will generate an efficient set of hyperboxes. Inappropriate selection of training samples leads to an erroneous set of hyperboxes, which will degrade the accuracy rate. Therefore, instead of using a single training set an extra training set, secondary training set, may be used to reshuffle the hyperboxes generated during the primary training set. In this paper, we have used a secondary training set to update the hyperboxes generated during the primary training phase. When a secondary training set is used, two cases are arised. First, change the class label of some inefficient hyperboxes created during the primary training phase. Second, fix the class label of efficient hyperboxes to the class allotted during the primary training set. By using the above secondary training set mechanism, a novel class label altering fuzzy min max (CLAFMM) network is proposed to alter the class labels depending on the secondary training set. Experimental results prove that the proposed approach provides more accuracy rate than the FMM, Enhanced FMM and K-nearest FMM. Simultaneously, the proposed approach reduces the complexity of the network by reducing the number of hyperboxes generated by the above said state-of-the-art methods. The proposed method is also applied to the breast cancer histopathological images to identify the best magnifying factor for classification.

SELECTING TRAINING SAMPLES AUTOMATICALLY FROM VHR SATELLITE IMAGES FOR IMAGE CLASSIFICATION

Classification is one of the most significant phases for remote sensing image interpretation. All the supervised classifiers need sufficient and efficient training samples, which are usually selected manually and labeled by visual inspection or field survey. Selecting training samples manually requires more time and human effort. A new method is proposed for automatic selection of training samples from a Very High Resolution (VHR) satellite image. The proposed method is tested for selecting training samples automatically for standard supervised pixel-based classification methods instead of manual samples selection. The proposed method uses a set of indices with specific thresholds to identify the training areas for each class. A certain part of each index histogram can be chosen for each class and consider as training samples. Automatic training samples are compared with manual samples for three standard classification methods. The average accuracy achieved by the proposed automatic sample selection is promising; 76.56% for maximum likelihood classifier, 74.06% for spectral correlation mapper classifier, and 70.00% for spectral angle mapper classifier. Although their accuracy scores are slightly nearby classification with manually selected samples by an average of 1.74% for maximum likelihood, 2.44% for spectral correlation mapper, and 3.75% for spectral angle mapper classifier.

Deep change feature analysis network for observing changes of land use or natural environment

Change detection on surface of earth plays an important role in global-scale pattern of climate and biogeochemistry of the world, which helps to comprehend the connections and associations between human and nature. Remote Sensing and Geographic Information Systems can possibly provide accurate data in regards to land use and land cover changes. However, pixel-based change detection methods are limited in suppressing outliers and noise; they often fail to process remote sensing images with high spatial-/spectral-resolution. To conquer these drawbacks, a superpixel-level change detection and analysis method is proposed in this paper. Superpixels are the atomic regions gathering pixels with similar property, which will be more efficient and robust than pixels. Deep neural network is a powerful feature learning and classification tool, it can represent superpixel abstractly and classify them robustly. The learning progress of deep architectures includes unsupervised sample selection and supervised feature learning, unsupervised progress aims at selecting training samples for deep neural network, supervised progress aims at learning the representation of superpixels and fine-tuning the whole network to finish classification. Experimental results on multi-temporal images have demonstrated that the proposed approach can handle the task of change detection and analysis effectively and accurately.

Training Sample Selection Strategy Applied to CNN in Magneto-Thermal Coupled Analysis

Deep learning (DL) has attracted more and more attention in computational electromagnetism. Particularly, the convolutional neural network (CNN) is one of the most popular learning models in DL due to its excellent capacity for feature extraction and convergence. The efficiency of CNN mainly depends on how many training samples are needed to effectively converge the network. The sample preparation process often involves a lot of numerical computations, which can be very expensive and time-consuming. In this article, based on the traditional DL network training procedure, two different approaches, namely adding smart training samples and reference samples, are proposed to help the DL network converge. The smart sample selection is based on a greedy algorithm, which can be applied for both training and reference samples. The influences of these two approaches on the CNN training process are investigated by an example of the coupled magneto-thermal computation applied to a transformer. Numerical results show that the two proposed approaches can significantly help the network to converge and improve the efficiency of the DL model.

A Knowledge-Based Auxiliary Channel STAP for Target Detection in Shipborne HFSWR

The broadened first-order sea clutter in shipborne high frequency surface wave radar (HFSWR), which will mask the targets with low radial velocity, is a kind of classical space–time coupled clutter. Space–time adaptive processing (STAP) has been proven to be an effective clutter suppression algorithm for space-time coupled clutter. To further improve the efficiency of clutter suppression, a STAP method based on a generalized sidelobe canceller (GSC) structure, named as the auxiliary channel STAP, was introduced into shipborne HFSWR. To obtain precise clutter information for the clutter covariance matrix (CCM) estimation, an approach based on the prior knowledge to auxiliary channel selection is proposed. Auxiliary channels are selected along the clutter ridge of the first-order sea clutter, whose distribution can be determined by the system parameters and regarded as pre-knowledge. To deal with the heterogeneity of the spreading first-order sea clutter, an innovative training samples selection approach according to the Riemannian distance is presented. The range cells that had shorter Riemannian distances to the cell under test (CUT) were chosen as training samples. Experimental results with measured data verified the effectiveness of the proposed algorithm, and the comparison with the existing clutter suppression algorithms showed the superiority of the algorithm.