Gray Level-gradient Co-occurrence Matrix Research Articles

Early Detection of Verticillium Wilt in Cotton by Using Hyperspectral Imaging Combined with Recurrence Plots

Cotton is susceptible to Verticillium wilt (VW) during its growth. Early and accurate detection of VW can facilitate targeted pesticide treatment and reduce the potential spread of the disease. However, accurately detecting VW in cotton before symptoms appear (the asymptomatic period) after infection by Verticillium dahliae remains challenging. This study proposes an early detection method for cotton wilt disease using hyperspectral imaging and recurrence plots (RP) combined with machine learning techniques. First, spectral curves were collected and analyzed under three conditions of cotton plants: healthy, asymptomatic, and symptomatic. Then, the one-dimensional spectral curve was transformed into two-dimensional recurrence plots to enhance the detail differences in the original spectral curve of cotton plants in various states. Hyperspectral recurrence plots contain rich texture information; fifteen texture features were extracted from the spectral recurrence plots using the Gray-Level Gradient Co-occurrence Matrix (GLGCM). Eleven of these texture features showed a strong correlation with the class labels of the cotton plants. In order to reduce redundant information between features, principal component analysis (PCA) was used to extract the first five principal components, which explained 99.02% of the information from the 11 features. The final principal component dataset was then input into KNN, SVM, ELM, and XGBoost classifiers to assess the accuracy of early detection of VW in cotton. The results showed that the XGBoost model, based on the first five principal components obtained from the texture features, achieved accuracy, precision, recall, and F1-score of 96.3%, 95.6%, 96%, and 95.8%, demonstrating a high classification capability. The results of this study confirm the feasibility of converting spectral curves into recurrence plots and extracting image texture features for the accurate identification of VW in cotton during the asymptomatic period. This method also provides a new strategy for early disease detection of cotton and other plants in the future.

Coal Gangue Recognition during Coal Preparation Using an Adaptive Boosting Algorithm

The recognition of coal and gangue is the premise and foundation of coal gangue intelligent sorting. Adaptive boosting (AdaBoost) algorithm-based coal gangue identification has not been studied in depth. This paper proposed a coal gangue image recognition algorithm and a strong classifier based on the AdaBoost algorithm with a genetic algorithm (GA)-optimized support vector machine (SVM). One thousand coal gangue images were collected on-site and expanded to five thousand via rotation and exposure adjustment. The 12 gray-level gradient co-occurrence matrix texture features of the images were extracted to construct a feature vector, establishing the training dataset and test dataset. Selection of the SVM kernel function, the GA optimization parameter setting, and the base classifier number was discussed. The coal gangue image recognition effects of the AdaB-GA-SVM classifier and the other strong classifiers with different base SVM classifiers were investigated. The results indicated that the recognition accuracy of GA-SVM was the best when the kernel function of SVM was RBF and the population number, crossover probability, and mutation probability were 80, 0.9, and 0.005, respectively. The AdaB-GA-SVM classifier has excellent identification and effective classification performance with the highest accuracy of 95%, a precision rate of 92.8%, recall rate of 97.3%, and KS values of 0.79.

Correlations between B-mode ultrasound image texture features and tissue temperatures in hyperthermia.

PurposeThe noninvasive monitoring of mild hyperthermia or thermal ablation is important to guarantee therapeutic safety and efficacy. The potential of ultrasound B-mode image texture features in monitoring temperature or coagulation zones studied in this article.Materials and methodsThe experiments carried out on eighteen in vitro porcine liver samples heated from 20°C to 60°C in the water bath. The ultrasound radiofrequency signal at different temperature collected to reconstruct B-mode ultrasound images. The texture features based on gray level histogram (GLH), gray level co-occurrence matrix (GLCM), and gray level-gradient co-occurrence matrix (GGCM) extracted, respectively. Accordingly, we analyze the correlations between these texture features and temperature based on the experiment results.ResultsThe results showed that five texture feature parameters closely related to temperature, including mean gray scale of GLH, homogeneity of GLCM, hybrid entropy, inverse difference moment, and correlation of GGCM. Some of these feature parameters have correlation coefficients larger than 0.9 within the temperature range of 20°C to 60°C.ConclusionsThe above-mentioned five feature parameters expected to apply for noninvasive monitoring of MH or TA.

Dynamic PET Imaging Using Dual Texture Features.

Purpose: This study aims to explore the impact of adding texture features in dynamic positron emission tomography (PET) reconstruction of imaging results.Methods: We have improved a reconstruction method that combines radiological dual texture features. In this method, multiple short time frames are added to obtain composite frames, and the image reconstructed by composite frames is used as the prior image. We extract texture features from prior images by using the gray level-gradient cooccurrence matrix (GGCM) and gray-level run length matrix (GLRLM). The prior information contains the intensity of the prior image, the inverse difference moment of the GGCM and the long-run low gray-level emphasis of the GLRLM.Results: The computer simulation results show that, compared with the traditional maximum likelihood, the proposed method obtains a higher signal-to-noise ratio (SNR) in the image obtained by dynamic PET reconstruction. Compared with similar methods, the proposed algorithm has a better normalized mean squared error (NMSE) and contrast recovery coefficient (CRC) at the tumor in the reconstructed image. Simulation studies on clinical patient images show that this method is also more accurate for reconstructing high-uptake lesions.Conclusion: By adding texture features to dynamic PET reconstruction, the reconstructed images are more accurate at the tumor.

Inferior alveolar canal segmentation based on cone-beam computed tomography.

The shape and position of the inferior alveolar canal (IAC) are analyzed to effectively reduce the risk of iatrogenic injury based on cone-beam computer tomography (CBCT). To assist dental clinicians to make better use of the IAC information, we propose an IAC segmentation method based on CBCT images. In this paper, CBCT images are first preprocessed by the Hounsfield unit values clipping and gray normalization. Secondly, based on the multi-plane reconstruction (MPR) and curved surface reconstruction, the curved MPR image sets are generated by the smooth dental arch curve with a sampling distance of 1.00pixels. Then, the K-means clustering algorithm is used to cluster the texture parameters of the gray level-gradient co-occurrence matrix enhanced by the gradient directions to improve the image contrast of the IAC. Finally, the IAC edges are roughly segmented by the 2D line-tracking method, and smoothed by the fourth-order polynomial to obtain the final segmentation result. Twenty-one real clinical dental CBCT datasets were used to test the proposed method. The manual segmentation results of two specialized dental clinicians were used as quantitative evaluation criteria. The dice similarity index (DSI), average symmetric surface distance (ASSD), and mean curve distance (MCD) of the left IAC are 0.93 (SD=0.01), 0.16mm (SD=0.05mm), and 1.59mm (SD=0.25mm), respectively; the DSI, ASSD, and MCD of the right IAC are 0.93 (SD=0.02), 0.16mm (SD=0.05mm), and 1.60mm (SD=0.30mm), respectively. The proposed method provides an effective image enhancement and segmentation solution to analyze the shape and position of the IAC. Experimental results show that the relationships between the IAC and other structures can be accurately reflected in the panoramic images without superimposition and geometric distortion, and the smooth edges of the IAC can be segmented.

Early detection of moldy apple core based on time-frequency images of vibro-acoustic signals

A nondestructive vibro-acoustic device with piezoelectric transducers was employed to detect the moldy core of apples. In this study, 15 texture features were extracted from the time-frequency images of acoustic signals using the gray-level gradient co-occurrence matrix method. A total of 11 texture features were selected and subsequently concentrated in the first four principal components. The principal component dataset was fed into three supervised classifiers, i.e., a back-propagation neural network (BP-NN), support vector machine (SVM), and an extreme learning machine (ELM) to identify slightly diseased apples. The ELM model achieved a relatively higher classification accuracy of 92.3 % for the slightly diseased apples with moldy-core degree less than 6.5 %. The overall classification accuracy reached to 93.9 %. Consequently, the proposed method can be applied for the early detection of apples with moldy core.

Feasibility of Authenticating Mutton Geographical Origin and Breed Via Hyperspectral Imaging with Effective Variables of Multiple Features

Muttons of specific geographical origin and breed are nutritious and flavorful and they have high economic value. However, substitution and adulteration are pervasive in the market. In this study, we explore hyperspectral imaging for authenticating mutton geographical origin and breed. We extracted the reflectance spectra of 400–1000 nm from the region of interest (ROI) of hyperspectral images, and we selected effective wavelengths through principal component analysis. Then, we acquired textural features from grayscale ROI images using a gray-level gradient co-occurrence matrix and screened effective variables through correlation analysis. With various features, we used support vector machines, random forest (RF), K-nearest neighbor, and convolutional neural network to develop models to identify mutton geographical origin and breed. Optimal identification was achieved using RF with the spectra of effective wavelengths and effective variables of textural features. Classification accuracies of 99.54% and 95.67% for calibration and prediction sets, respectively, were obtained. From the confusion matrix, we authenticated each mutton type with high accuracy. Hence, HSI with effective variables of multiple features coupled with RF can identify mutton geographical origin and breed accurately and provide a potential reference for a simple system that integrates multi-wavelength spectroscopy and digital imaging.

Weed and Corn Seedling Detection in Field Based on Multi Feature Fusion and Support Vector Machine.

Detection of weeds and crops is the key step for precision spraying using the spraying herbicide robot and precise fertilization for the agriculture machine in the field. On the basis of k-mean clustering image segmentation using color information and connected region analysis, a method combining multi feature fusion and support vector machine (SVM) was proposed to identify and detect the position of corn seedlings and weeds, to reduce the harm of weeds on corn growth, and to achieve accurate fertilization, thereby realizing precise weeding or fertilizing. First, the image dataset for weed and corn seedling classification in the corn seedling stage was established. Second, many different features of corn seedlings and weeds were extracted, and dimensionality was reduced by principal component analysis, including the histogram of oriented gradient feature, rotation invariant local binary pattern (LBP) feature, Hu invariant moment feature, Gabor feature, gray level co-occurrence matrix, and gray level-gradient co-occurrence matrix. Then, the classifier training based on SVM was conducted to obtain the recognition model for corn seedlings and weeds. The comprehensive recognition performance of single feature or different fusion strategies for six features is compared and analyzed, and the optimal feature fusion strategy is obtained. Finally, by utilizing the actual corn seedling field images, the proposed weed and corn seedling detection method effect was tested. LAB color space and K-means clustering were used to achieve image segmentation. Connected component analysis was adopted to remove small objects. The previously trained recognition model was utilized to identify and label each connected region to identify and detect weeds and corn seedlings. The experimental results showed that the fusion feature combination of rotation invariant LBP feature and gray level-gradient co-occurrence matrix based on SVM classifier obtained the highest classification accuracy and accurately detected all kinds of weeds and corn seedlings. It provided information on weed and crop positions to the spraying herbicide robot for accurate spraying or to the precise fertilization machine for accurate fertilizing.

Differentiating novel coronavirus pneumonia from general pneumonia based on machine learning

BackgroundChest CT screening as supplementary means is crucial in diagnosing novel coronavirus pneumonia (COVID-19) with high sensitivity and popularity. Machine learning was adept in discovering intricate structures from CT images and achieved expert-level performance in medical image analysis.MethodsAn integrated machine learning framework on chest CT images for differentiating COVID-19 from general pneumonia (GP) was developed and validated. Seventy-three confirmed COVID-19 cases were consecutively enrolled together with 27 confirmed general pneumonia patients from Ruian People’s Hospital, from January 2020 to March 2020. To accurately classify COVID-19, region of interest (ROI) delineation was implemented based on ground-glass opacities (GGOs) before feature extraction. Then, 34 statistical texture features of COVID-19 and GP ROI images were extracted, including 13 gray-level co-occurrence matrix (GLCM) features, 15 gray-level-gradient co-occurrence matrix (GLGCM) features and 6 histogram features. High-dimensional features impact the classification performance. Thus, ReliefF algorithm was leveraged to select features. The relevance of each feature was the average weights calculated by ReliefF in n times. Features with relevance larger than the empirically set threshold T were selected. After feature selection, the optimal feature set along with 4 other selected feature combinations for comparison were applied to the ensemble of bagged tree (EBT) and four other machine learning classifiers including support vector machine (SVM), logistic regression (LR), decision tree (DT), and K-nearest neighbor with Minkowski distance equal weight (KNN) using tenfold cross-validation.Results and conclusionsThe classification accuracy (ACC), sensitivity (SEN), specificity (SPE) of our proposed method yield 94.16%, 88.62% and 100.00%, respectively. The area under the receiver operating characteristic curve (AUC) was 0.99. The experimental results indicate that the EBT algorithm with statistical textural features based on GGOs for differentiating COVID-19 from general pneumonia achieved high transferability, efficiency, specificity, sensitivity, and impressive accuracy, which is beneficial for inexperienced doctors to more accurately diagnose COVID-19 and essential for controlling the spread of the disease.

Detection of Winding Faults Using Image Features and Binary Tree Support Vector Machine for Autotransformer

Autotransformer (AT) is the most core power supply equipment, and overvoltage and short circuit (SC) fault may lead to winding deformation, which will have a negative impact on its insulation and even affect the operation of a train. The frequency response analysis (FRA) is widely used for detecting winding faults in a transformer. However, the direct measure of FRA for each split winding fails because the split windings are adopted to satisfy the impedance requirement of a high-speed railway, where the windings are connected inside the tank. A novel fault interpretation method based on image features and binary tree support vector machine (SVM) is proposed, which can get the condition of three windings in one measurement. Winding faults caused by different windings are simulated, including SC defect, axial deformation, and series capacitance variation, and the FRA curves are measured under various faults. Then, the features of the gray-level gradient co-occurrence matrix and the gray-level difference statistics are got from the polar plot of FRA. Finally, the image features are used as the inputs to the binary tree SVM for fault type and faulty winding classification. The results show that the proposed method has high accuracy for identifying fault type and faulty winding in AT.

Roughness Analysis of Sea Surface From Visible Images by Texture

This paper presents a roughness analysis of sea surface from visible images by feature measurements of texture for the first time. The algorithms presented in this paper include six texture feature measurements of sea surface use gray level co-occurrence matrix, gray level-gradient co-occurrence matrix, Tamura texture feature, autocorrelation function, edge frequency and fractional Brownian motion autocorrelation. The empirical relationship between wind speeds (or sea surface roughness) and image texture roughness are estimated based on the extracted data. Our experiments have demonstrated that our texture methods and empirical relation between wind speeds and image texture roughness can potentially be used to analyze sea surface roughness from visible images.

Automatic Extraction of Water and Shadow from SAR Images Based on a Multi-Resolution Dense Encoder and Decoder Network.

The water and shadow areas in SAR images contain rich information for various applications, which cannot be extracted automatically and precisely at present. To handle this problem, a new framework called Multi-Resolution Dense Encoder and Decoder (MRDED) network is proposed, which integrates Convolutional Neural Network (CNN), Residual Network (ResNet), Dense Convolutional Network (DenseNet), Global Convolutional Network (GCN), and Convolutional Long Short-Term Memory (ConvLSTM). MRDED contains three parts: the Gray Level Gradient Co-occurrence Matrix (GLGCM), the Encoder network, and the Decoder network. GLGCM is used to extract low-level features, which are further processed by the Encoder. The Encoder network employs ResNet to extract features at different resolutions. There are two components of the Decoder network, namely, the Multi-level Features Extraction and Fusion (MFEF) and Score maps Fusion (SF). We implement two versions of MFEF, named MFEF1 and MFEF2, which generate separate score maps. The difference between them lies in that the Chained Residual Pooling (CRP) module is utilized in MFEF2, while ConvLSTM is adopted in MFEF1 to form the Improved Chained Residual Pooling (ICRP) module as the replacement. The two separate score maps generated by MFEF1 and MFEF2 are fused with different weights to produce the fused score map, which is further handled by the Softmax function to generate the final extraction results for water and shadow areas. To evaluate the proposed framework, MRDED is trained and tested with large SAR images. To further assess the classification performance, a total of eight different classification frameworks are compared with our proposed framework. MRDED outperformed by reaching 80.12% in Pixel Accuracy (PA) and 73.88% in Intersection of Union (IoU) for water, 88% in PA and 77.11% in IoU for shadow, and 95.16% in PA and 90.49% in IoU for background classification, respectively.

Analysis and detection of decayed blueberry by low field nuclear magnetic resonance and imaging

Decay is a major cause of quality loss and food safety concerns for blueberry fruit. Low-field nuclear magnetic resonance (LF-NMR) has been used to analyze and detect decayed blueberry fruit. According to NMR relaxation analysis, transverse relaxation time (T21, T22, T23) of decayed fruit increased to different degrees, while signal amplitude (A21, A22, A23) decreased to different degrees. Six relaxation features were input into BPNN model to identify the decay classes of fruit. The identification accuracy of training set was 86.7%, and that of the validation set was 90%. According to magnetic resonance imaging (MRI), twice threshold segmentation algorithm was proposed to segment the decayed region with dark color. A total of 15 features were extracted from gray histogram, gray level co-occurrence matrix (GLCM) and gray level-gradient co-occurrence matrix (GGCM) of the image for correlation analysis. Five features were then input into a back propagation neural network (BPNN) model. The identification accuracy of training set was 92.2%, and that of the validation set was 83.3%. Finally, 11 variables, including transverse relaxation features and image features, were input into the BPNN model. The identification accuracy of training set was 98.8%, and that of the validation set was 94.2%, which showed the highest identification accuracy. The results showed that LF-NMR and MRI are suitable for analyzing and detecting the decay disease of fruit, which provides a theoretical basis for nondestructive detection of fruit disease.

Rapid Estimation of Road Friction for Anti-Skid Autonomous Driving

The failure to estimate friction on the road is a significant cause of roadway departure crashes. It is even more relevant to autonomous vehicles (AVs) as they are currently not designed to measure the skid resistance of roads. Most traditional measurements are too inefficient for real-time AV control. Thus, in this paper, we propose a dynamic method to estimate pavement friction level using computer vision. We collect 100 sets of high-quality road images and their resistance values to train the model, and design two major methods for analysis: 1) texture identification, consisting of grayscale-enhanced local binary pattern and the gray-level gradient co-occurrence matrix and 2) a deep neural network based on domain knowledge (TLDKNet). We introduce two standards of classification for model training and propose three indices—underestimation error, overestimation error, and accuracy—to verify the performance of our algorithms. The results showed the great correlation between pavement texture and skid resistance. The TLDKNet yielded the best performance with an accuracy of 90.67% and only 2.67% underestimation error, showing that it is sufficiently conservative with regard to safety. Based on the proposed method of estimation, a framework for anti-skid driving control is developed regarding the car-following behavior and turning movements. The anti-skid control framework provides new insights into enhancing the AV safety performance.

A New Deep Learning Algorithm for SAR Scene Classification Based on Spatial Statistical Modeling and Features Re-Calibration.

Synthetic Aperture Radar (SAR) scene classification is challenging but widely applied, in which deep learning can play a pivotal role because of its hierarchical feature learning ability. In the paper, we propose a new scene classification framework, named Feature Recalibration Network with Multi-scale Spatial Features (FRN-MSF), to achieve high accuracy in SAR-based scene classification. First, a Multi-Scale Omnidirectional Gaussian Derivative Filter (MSOGDF) is constructed. Then, Multi-scale Spatial Features (MSF) of SAR scenes are generated by weighting MSOGDF, a Gray Level Gradient Co-occurrence Matrix (GLGCM) and Gabor transformation. These features were processed by the Feature Recalibration Network (FRN) to learn high-level features. In the network, the Depthwise Separable Convolution (DSC), Squeeze-and-Excitation (SE) Block and Convolution Neural Network (CNN) are integrated. Finally, these learned features will be classified by the Softmax function. Eleven types of SAR scenes obtained from four systems combining different bands and resolutions were trained and tested, and a mean accuracy of 98.18% was obtained. To validate the generality of FRN-MSF, five types of SAR scenes sampled from two additional large-scale Gaofen-3 and TerraSAR-X images were evaluated for classification. The mean accuracy of the five types reached 94.56%; while the mean accuracy for the same five types of the former tested 11 types of scene was 96%. The high accuracy indicates that the FRN-MSF is promising for SAR scene classification without losing generality.

Handling Water Reflections for Computer Vision in Aquaculture

Abstract. In aquaculture, almost all images collected of an aquaculture scene contain reflections, which often affect the results and accuracy of machine vision. Classifying these images and obtaining images of interest are key to subsequent image processing. The purpose of this study was to identify useful images and remove images that had a substantial effect on the results of image processing for computer vision in aquaculture. In this study, a method for classification of reflective frames based on image texture and a support vector machine (SVM) was proposed for an actual aquaculture site. Objectives of this study were to: (1) develop an algorithm to improve the speed of the method and to ensure that the method has a high classification accuracy, (2) design an algorithm to improve the intelligence and adaptability of the classification, and (3) demonstrate the performance of the method. The results show that the average classification accuracy, false positive rate, and false negative rate for two types of reflective frames (type I and II) were 96.34%, 4.65%, and 2.23%, respectively. In addition, the running time was very low (1.25 s). This strategy also displayed considerable adaptability and could be used to obtain useful images or remove images that have substantial effects on the accuracy of image processing results, thereby improving the applicability of computer vision in aquaculture. Keywords: Aquaculture, Genetic algorithm, Gray level-gradient co-occurrence matrix, Principal component analysis, Reflection frame, Support vector machine.

A Multi-Scale Learning Local Phase and Amplitude Blind Image Quality Assessment for Multiply Distorted Images

In the real application, digital images may undergo the process of acquisition, compression, and transmission, which causes the excess blurring, quantization, and noise. However, the metrics of image quality assessment for multiply distorted images are very limited. In this paper, we propose a new multi-scale learning quality-aware features blind image quality assessment algorithm for multiply distorted images by using both local phase and local amplitude. In the new model, a distorted image is decomposed into three scales by Gabor transform, and its phase congruency image (PCI), phase congruency covariance maximum image (PCCmax), and phase congruency covariance minimum image (PCCmin) are produced. Then, we calculate contrast sensitivity function and gray level-gradient co-occurrence matrix features from distorted image and its PCI, PCCmax, and PCCmin, and mean value of intensity of PCI, PCCmax, PCCmin, and overlapping blocked local amplitude of distorted image. At last, SVR is used to build the approximating function between these features and subjective mean opinion scores. Both local phase and local amplitude features are extracted at multi-scale images, which supply more flexibility than the previous single-scale methods in incorporating the variations of viewing scene. Comparative experiments between our proposed metric and the state-of-the-art full-reference and no-reference metrics are conducted on two newly released multiply distorted image databases (LIVEMD and MDID2013) that demonstrate the effectiveness of our proposed method.

Texture-based classification of different single liver lesion based on SPAIR T2W MRI images

BackgroundTo assess the feasibility of texture analysis (TA) based on spectral attenuated inversion-recovery T2 weighted magnetic resonance imaging (SPAIR T2W-MRI) for the classification of hepatic hemangioma (HH), hepatic metastases (HM) and hepatocellular carcinoma (HCC).MethodsThe SPAIR T2W-MRI data of 162 patients with HH (n=55), HM (n=67) and HCC (n=40) were retrospectively analyzed. We used two independent cohorts for training (n = 112 patients) and validation (n = 50 patients). The TA was performed and textual parameters derived from the gray level co-occurrence matrix (GLCM), gray level gradient co-occurrence matrix (GLGCM), gray-level run-length matrix (GLRLM), Gabor wavelet transform (GWTF), intensity-size-zone matrix (ISZM), and histogram features were calculated. The capacity of each parameter to classify three types of single liver lesions was assessed using the Kruskal-Wallis test. Specificity and sensitivity for each of the studied parameters were derived using ROC curves. Four supervised classification algorithms were trained with the most influential textural features in the classification of tumor types. The test datasets validated the reliability of the models.ResultsThe texture analyses showed that the HH versus HM, HM versus HCC, and HH versus HCC could be differentiated by 9, 16 and 10 feature parameters, respectively. The model’s misclassification rates were 11.7, 9.6 and 9.7% respectively. No texture feature was able to adequately distinguish among the three types of single liver lesions at the same time. The BP-ANN model had better predictive ability.ConclusionTexture features of SPAIR T2W-MRI can classify the three types of single liver lesions (HH, HM and HCC) and may serve as an adjunct tool for accurate diagnosis of these diseases.

Deep Supervised and Contractive Neural Network for SAR Image Classification

The classification of a synthetic aperture radar (SAR) image is a significant yet challenging task, due to the presence of speckle noises and the absence of effective feature representation. Inspired by deep learning technology, a novel deep supervised and contractive neural network (DSCNN) for SAR image classification is proposed to overcome these problems. In order to extract spatial features, a multiscale patch-based feature extraction model that consists of gray level-gradient co-occurrence matrix, Gabor, and histogram of oriented gradient descriptors is developed to obtain primitive features from the SAR image. Then, to get discriminative representation of initial features, the DSCNN network that comprises four layers of supervised and contractive autoencoders is proposed to optimize features for classification. The supervised penalty of the DSCNN can capture the relevant information between features and labels, and the contractive restriction aims to enhance the locally invariant and robustness of the encoding representation. Consequently, the DSCNN is able to produce effective representation of sample features and provide superb predictions of the class labels. Moreover, to restrain the influence of speckle noises, a graph-cut-based spatial regularization is adopted after classification to suppress misclassified pixels and smooth the results. Experiments on three SAR data sets demonstrate that the proposed method is able to yield superior classification performance compared with some related approaches.

Near-infrared imaging to quantify the feeding behavior of fish in aquaculture

In aquaculture, fish feeding behavior under culture conditions holds important information for the aquaculturist. In this study, near-infrared imaging was used to observe feeding processes of fish as a novel method for quantifying variations in fish feeding behavior. First, images of the fish feeding activity were collected using a near-infrared industrial camera installed at the top of the tank. A binary image of the fish was obtained following a series of steps such as image enhancement, background subtraction, and target extraction. Moreover, to eliminate the effects of splash and reflection on the result, a reflective frame classification and removal method based on the Support Vector Machine and Gray-Level Gradient Co-occurrence Matrix was proposed. Second, the centroid of the fish was calculated by the order moment, and then, the centroids were used as a vertex in Delaunay Triangulation. Finally, the flocking index of fish feeding behavior (FIFFB) was calculated to quantify the feeding behavior of a fish shoal according to the results of the Delaunay Triangulation, and the FIFFB values of the removed reflective frames were fitted by the Least Squares Polynomial Fitting method. The results show that variations in fish feeding behaviors can be accurately quantified and analyzed using the FIFFB values, for which the linear correlation coefficient versus expert manual scoring reached 0.945. This method provides an effective method to quantify fish behavior, which can be used to guide practice.