Search

Timely and accurate detection of crop water stress is vital for precision irrigation. Whether the accuracy of the prevailing diagnosis of crop water stress using vegetation indices (VIs) and spectral reflectance can be improved still remains to be investigated. The crop surface characteristics such as grayscale or color vary under different water stress, so in this study one more variable, image texture, was utilized together to diagnose water stress. For this end, the canopy image of winter wheat in bloom was obtained by unmanned aerial vehicle (UAV) equipped with multispectral sensor, and the effect of soil background was eliminated using vegetation index threshold method. On this basis, Grey level co-occurrence matrix (GLCM) was used to calculate the mean (MEA), variance (VAR), homogeneity (HOM), contrast (CON), dissimilarity (DIS), entropy (ENT), second moment (SEC) and correlation (COR) of the image texture under different spatial resolutions (0.008 m, 0.01 m, 0.02 m, 0.05 m, 0.1 m and 0.2 m). Next, the canopy vegetation indices were obtained by mathematical transformation of canopy reflectance, and then sensitive image texture and vegetation indices by full subset regression method. Finally, Cubist, BPNN (Back Propagation Neural Network) and ELM (Extreme Learning Machine) methods were adopted to build the estimation models of the stomatal conductance (Gs) of winter wheat (between the sensitive image texture and Gs, and between vegetation index and Gs), and the water stress map was plotted based on the optimal Gs estimation model. The result showed: (i) the image texture obtained from the high-resolution multispectral image had a high correlation with Gs, and the image texture (VAR, HOM, CON, DIS, ENT and SEC) at 550 nm had the most significant correlation; (ii) the higher the ground resolution, the higher the correlation between the Gs and the image texture, the vegetation indices, respectively. The image texture with a ground resolution of 0.008 m combined with VIs and Gs had the highest correlation, and combining image texture and vegetation index can significantly improve the estimation accuracy of winter wheat Gs; (iii) Among the three estimation models, the BPNN model constructed by combining the image texture and VIs (MEA, VAR, ENT, DWSI and EXG) had the best estimation performance (Calibration: R c 2 = 0.899, RMSE c = 0.01, MAE c = 0.006; Validation: R c 2 = 0.834, RMSE v =;0.018, MAE v = 0.014), and an accurate estimation could even be achieved at a lower Gs value. Compared with the BPNN model solely based on VIs or image texture, the R c 2 of the BPNN model based on the combined variables increased by 24% and 22.48%, respectively. Therefore, combining UAV multispectral image texture and VIs to estimate Gs provides a feasible and accurate method for water stress diagnosis of winter wheat. • The incorporation of texture improved the accuracy of water stress diagnoses with VIs. • BPNN model with VIs and texture provided the highest accuracy of Gs estimation. • UAV provides a feasible and accurate method for water stress diagnosis of winter wheat. • Removing soil background and screening methods improve the quality of remote sensing data. [ABSTRACT FROM AUTHOR]

• Propose a computer vision method for detecting asphalt pavement segregation. • Employ image texture analysis for characterizing pavement surface condition. • Extreme gradient boosting machine and deep neural network are used for classification. • Attractive repulsive center-symmetric local binary pattern is used for texture computation. • XGBoost has achieved the best detection accuracy with accuracy rate = 0.95. Aggregate segregation is a major form of defect that accelerates the pavement deterioration rate. Therefore, asphalt pavement segregation needs to be detected accurately and early during the quality survey process. This study proposes and verifies a computer vision based method for automatic identification of aggregate segregation. The new method includes Extreme Gradient Boosting Machine integrated with Attractive Repulsive Center-Symmetric Local Binary Pattern (ARCSLBP-XGBoost) and Deep Convolutional Neural Network (DCNN). Experimental results obtained from a repetitive random data sampling process with 20 runs show that the ARCSLBP-XGBoost is a capable approach for detecting asphalt pavement segregation with outstanding performance measurement metrics (classification accuracy rate = 0.95, precision = 0.93, recall = 0.98, and F1 score = 0.95). [ABSTRACT FROM AUTHOR]

• Cotton water status was estimated using vegetation indices and textural information. • Combining vegetation indices with textural information enhanced estimation accuracy. • The XGBoost_VIs + TFs + TIs model performed the best in estimating CWSI. • The estimated CWSI predictive map could diagnose cotton canopy water stress. Precise and timely estimation of crop canopy water status is of great importance for precision irrigation. The unmanned aerial vehicle (UAV)-based remote sensing technology has become increasingly popular for crop canopy water status estimation. Nevertheless, the capability of vegetation indices (VIs) along with textural information from high-resolution imagery for estimating cotton canopy water status has been rarely explored. The VIs, texture features (TFs), and texture indices (TIs) were obtained from UAV multispectral images of cotton with different irrigation levels and nitrogen rates in the southern Xinjiang of China. The performances of three machine learning models, i.e. support vector machine (SVM), back-propagation neural network (BPNN) and extreme gradient boosting (XGBoost) were evaluated for estimating canopy equivalent water thickness (CEWT) and crop water stress index (CWSI) throughout the cotton growing season. The results showed that combining vegetation indices and textural information significantly improved the estimation accuracy of models compared to vegetation indices or textural information alone. The XGBoost_VIs + TFs model exhibited the best accuracy in estimating CEWT (R2 = 0.75, RMSE = 0.01 cm, RE = 19.46 % at upper half-canopy level, and R2 = 0.65, RMSE = 0.02 cm, RE = 24.59 % at all-canopy level), while the XGBoost_VIs + TFs + TIs model performed best in predicting CWSI among the models (R2 = 0.90, RMSE = 0.05, RE = 5.84 %). Although CEWT estimation was fair to some extent, CWSI estimation was more applicable for diagnosing cotton water stress. The CWSI maps created from the optimal XGBoost_VIs + TFs + TIs model intuitively reflected the cotton canopy water status under various irrigation levels and nitrogen rates, which could help farmers implement timely and precision irrigation in cotton production. [ABSTRACT FROM AUTHOR]

Recently, a concept for a new approach for rapid identification of uranium ore concentrate (UOC) powders using colour, textural and spectroscopy signatures was developed using a sample dataset consisting of 79 industrial uranium ore concentrate powders produced by different production routes at different facilities. The samples were first grouped into six different predefined colour categories using a colour-based classification model previously developed. In this study, a machine learning approach was used to develop supervised texture- and spectroscopy-based classification models for each colour category of uranium ore concentrates. Scanning electron microscopy images of the UOC samples were collected at 100x, 250x, 500x/1000× magnifications. A total of 524 texture features describing sample morphology were extracted from each image using First Order Statistics (FOS), Grey Level Co-Occurrence Matrix (GLCM), Grey Level Run Length Matrix (GLRLM), Grey Level Size Zone Matrix (GLSZM), Locally Binary Pattern (LBP) and Angle Measure Technique (AMT) textural feature extraction algorithms. Hyperspectral images of the samples were collected in the wavelength range 900–1700 nm and the mean spectrum of each sample was calculated, giving the spectral features of the UOCs. The texture and spectral features were used as input to the machine learning approach which consisted of a feature reduction step to identify discriminatory features and a classification step to provide the texture- and spectroscopy-based models. An average prediction accuracy of 82% and 97% were obtained for the texture- and spectral-based classification models, respectively, on hold-out test sets. The feature reduction step identified two new nuclear forensics signatures, consisting of 3–7 texture (morphological) characteristics and 2–6 characteristic wavelengths, respectively. This new approach enables rapid classification of unknown uranium ore concentrate powders using only their textural and hyperspectral images. By combining the classification results from the two models, the most probable production process and chemical composition can be identified, and the provenience of the seized sample can be narrowed down to a few matching possibilities. Moreover, after the image acquisition, the models can provide results in less than 2 min speeding up the investigation procedure, a key factor when a nuclear security event occurs. • A new approach to classify seized UOCs powders by colour, texture and spectroscopy data is proposed. • 79 commercial UOCs samples were used as training and test samples. • ML approach was used to test several classifiers and to carry out feature reduction/selection. • LDA was selected as the most promising classifier to develop the texture/spectroscopy-based classification models. • An average prediction accuracy of 82% and 97% were obtained for the two models respectively. [ABSTRACT FROM AUTHOR]

Defrost control using digital image processing is a potential, economical, and energy-saving solution for air-source heat pump or refrigeration system, comparing with currently studied or used direct/indirect measuring methods. However, under complex operating condition of different shooting angles, lighting conditions, and pixel level, which are common in refrigeration system, the accuracy of frost state recognition with existing digital image processing methods decrease dramatically. Therefore, a new method based on optimized gray level co-occurrence matrix and new characterization parameter combination to extract image texture characteristics, combining extreme learning machine algorithm (OGLCM-ELM), is proposed for the first time. An experimental rig for evaporator frosting image under different lighting intensity, shooting angle and pixel level is set up. The collected experimental image data are divided into 3 classifications (Frostless, light frost, and heavy frost) or 2 classifications ((Frostless + light frost, and heavy frost)). The results show for ternary classification, OGLCM-ELM reveals significantly higher recognition accuracy then existing methods, average accuracy is as higher as 98.14%. It is also 5.28% and 3.14% higher than those of OGLCM-SVM, OGLCM-BP. Other performance parameters, precision, recall, F1-score, and calculating time are also totally better than other methods. For binary classification. the average accuracy of OGLCM-ELM even reaches 99% under complex operating condition, indicating it is a practical and potential technology for defrost control. [ABSTRACT FROM AUTHOR]

• Texture features of hydrophobized concrete contain the information about their absorptivity. • Machine learning algorithms are used to estimate the absorptivity of hydrophobic concrete. • Water absorption of concrete is reliably estimated from conventional SEM. • Application of image texture analysis to characterize the condition of hydrophobic concrete surfaces. The work concerned the study of surface hydrophobized concrete's physical and mechanical properties. An aqueous emulsion based on nano-silicates (A1) and an oligomeric propylsilicate/silicate (A2) concentrate in three dilution states (100%, 70%, and 50%) were used as surface modification agents. The scanning electron microscopy (SEM) images determined three classes of water absorption (WA). A predictive modeling process was performed to automatically identify 1 of the 3 water absorption classes. For the best model, a classification accuracy of 96% was obtained. After 14 days of testing, the hydrophobization efficiency was still high, over 54% for A1 and 45% for A2 for 100% concentration. The samples achieved the best frost resistance with agents A1 and A2 in a 70% concentration. Experimental studies have confirmed the close relationship between concretes' water absorptivity and their surfaces' SEM images. No similar studies of this type are known. [ABSTRACT FROM AUTHOR]

• An automated, machine-learning based, material-testing fracture characterization process is elaborated. • Image-based, highly-significant texture features are identified and classified. • Neural network, high accuracy material-testing fracture classifiers are trained. • The applicability of the methodology in different ductile fracture experiments is illustrated. Correctly determining the onset of fracture is crucial when performing mechanical experiments. Commonly carried out by visual inspection, here an image-based machine learning approach is proposed to classify cracked and un-cracked specimens. It yields the potential to objectify and automate crack detection, thereby removing sources of uncertainty and error from the post-processing of experiments. More than 30'000 speckle-pattern images obtained from 77 experiments on three specimen geometries are evaluated. They comprise uniaxial tension, notched tension as well as axisymmetric V-bending experiments. Statistical texture features are extracted from all images. They include both first-order (variance, skewness, kurtosis) and higher-order statistical texture features, i.e. Haralick features. The discriminatory power of the texture information is evaluated based on the Fisher's Discriminant Ratio and feature correlations are identified and quantified. Image texture feature subsets of high discriminatory power are used to parse neural network architectures of different complexities from simple perceptron to feed-forward and cascade neural networks. It is found that a small subset of the investigated texture features is highly significant for all experiments. Using this feature subset in conjunction with multi-layer, non-linear and low complexity feed-forward network architectures classification accuracies in the order of 99% are obtained. At the same time, it is shown that linear classifiers are not sufficient to robustly distinguish the state of the specimens, even when high discriminatory power features are used. Graphical Abstract: [Display omitted] [ABSTRACT FROM AUTHOR]