Image Classification Results Research Articles

A Multi-Level Non-Uniform Spatial Sampling Method for Accuracy Assessment of Remote Sensing Image Classification Results

Accuracy assessment of classification results has important significance for the application of remote sensing images, which can be achieved by sampling methods. However, the existing sampling methods either ignore spatial correlation or do not consider spatial heterogeneity. Here, we proposed a multi-level non-uniform spatial sampling method (MNSS) for the accuracy assessment of classification results. Taking the remote sensing image of Kobo Askov, Texas, USA, as an example, the classification result of this image was obtained by Support Vector Machine (SVM) classifier. In the proposed MNSS, the studied spatial region was zoned from high to low resolution based on the features of spatial correlation. Then, the sampling rate of each zone was deduced from the low to high resolution based on the spatial heterogeneity. Finally, the positions of sample points were allocated in each zone, and the classification results of the sample points were obtained. We also used other sampling methods, including a random sampling method (SRS), stratified sampling method (SS), and spatial sampling of the gray level co-occurrence matrix method (GLCM), to obtain the classification results of the sample points (2-m resolution). Five categories of ground objects in the same region were used as the ground truth data. We than calculated the overall accuracy, Kappa coefficient, producer accuracy, and user accuracy to estimate the accuracy of the classification results. The results showed that MNSS was the strictest inspection method as shown by the minimum value of accuracy. Moreover, MNSS overcame the shortcoming of SRS, which did not consider the spatial correlation of sample points, and overcame the shortcomings of SS and GLCM, which had redundant information between sample points. This paper proposes a novel sampling method for the accuracy assessment of classification results of remote sensing images.

Food image classification and image retrieval based on visual features and machine learning

Research on image retrieval and classification in the food field has become one of the more and more concerned research topics in the field of multimedia analysis and applications. In recent years, with the rapid development of the Internet industry and multimedia technology, image classification and retrieval technology has become a research hotspot at home and abroad. Traditional keyword-based image retrieval and image classification have been unable to meet people’s daily needs; so, image recognition methods based on image content came into being. The most representative of image feature description methods are mainly two aspects: image visual features and image abstract semantics extracted based on machine learning algorithms. These two algorithms have their own key points in describing images, which are difficult to achieve the desired results in image classification and image retrieval. Based on this, this paper proposes research on food image classification and image retrieval methods based on visual features and machine learning. This paper proposes a food image retrieval and classification method based on Faster R-CNN network. This paper selects food image sets from the visual gene database to fine-tune the Faster R-CNN network to ensure the accuracy of Faster R-CNN food area detection, and experimented on the Dish-233 food dataset, which is a subset of the dish dataset, including 233 dishes and 49,168 images. The experimental results in this paper show that the performance of this method is better than other methods in terms of image classification performance. Compared with CNN-GF, the performance is improved by 5%. In terms of image retrieval, this method also shows its superiority This proves that compared with other methods, the proposed method has more discriminative visual features, and its performance has been improved in food image retrieval and classification tasks.

A Comparative Study of Water Indices and Image Classification Algorithms for Mapping Inland Surface Water Bodies Using Landsat Imagery

A comparative study of water indices and image classification algorithms for mapping inland water bodies using Landsat imagery was carried out through obtaining 24 high-resolution (≤5 m) and cloud-free images archived in Google Earth with the same (or ±1 day) acquisition dates as the Landsat-8 OLI images over 24 selected lakes across the globe, and developing a method to generate the alternate ground truth data from the Google Earth images for properly evaluating the Landsat image classification results. In addition to the commonly used green band-based water indices, Landsat-8 OLI’s ultra-blue, blue, and red band-based water indices were also tested in this research. Two unsupervised (the zero-water index threshold H0 method and Otsu’s automatic threshold selection method) and one supervised (the k-nearest neighbor (KNN) method) image classification algorithms were employed for conducting the image classification. Through comparing a total of 2880 Landsat image classification results with the alternate ground truth data, this study showed that (1) it is not necessary to use some supervised image classification methods for extracting water bodies from Landsat imagery given the high computational cost associated with the supervised image classification algorithms; (2) the unsupervised classification algorithms such as the H0 and Otsu methods could achieve comparable accuracy as the KNN method, although the H0 method produced more large error outliers than the Otsu method, thus the Otsu method is better than the H0 method; and (3) the ultra-blue band-based AWEInsuB is the best water index for the H0 method, and the ultra-blue band-based MNDWI2uB is the best water index for both the Otsu and KNN methods.

Classification of Brain Tumors from MRI Images Using a Convolutional Neural Network

The classification of brain tumors is performed by biopsy, which is not usually conducted before definitive brain surgery. The improvement of technology and machine learning can help radiologists in tumor diagnostics without invasive measures. A machine-learning algorithm that has achieved substantial results in image segmentation and classification is the convolutional neural network (CNN). We present a new CNN architecture for brain tumor classification of three tumor types. The developed network is simpler than already-existing pre-trained networks, and it was tested on T1-weighted contrast-enhanced magnetic resonance images. The performance of the network was evaluated using four approaches: combinations of two 10-fold cross-validation methods and two databases. The generalization capability of the network was tested with one of the 10-fold methods, subject-wise cross-validation, and the improvement was tested by using an augmented image database. The best result for the 10-fold cross-validation method was obtained for the record-wise cross-validation for the augmented data set, and, in that case, the accuracy was 96.56%. With good generalization capability and good execution speed, the new developed CNN architecture could be used as an effective decision-support tool for radiologists in medical diagnostics.

Facial Beauty Prediction Based on Lighted Deep Convolution Neural Network with Feature Extraction Strengthened

Convolutionneural network (CNN) has significantly pushed forward machine vision, which has achieved very significant results in face recognition, image classification and objection detection, and provides a new method for facial beauty prediction (FBP). Although the approach is widely applied in FBP, the research progress in FBP is relatively slow compared with face recognition. The first one is that there is less public database for FBP, and experiments for FBP are tested on small-scale database. The second one is that evaluation of facial beauty is subjective and lack of criterion, and CNN model is hard to train. In view of the problems of FBP, we expand Largescale database of Asian women's face database (LSAFBD) with data augmentation. A lighted deep convolution neural network (LDCNN) for FBP including 5650K parameters is constructed by both Inception model of GoogleNet and Max-Feature-Max activation layer, which can extract multi-scale features of an image, get compacted presentation and reduce parameters. Experiments on LSAFBD show that our LDCNN model has advantages of simple structure, small-scale parameters and is suitable for small embedded devices, with the best classification accuracy of 63.5%, which outperforms the other published CNN models for FBP.

Evaluating the classification of images from geoscience papers using small data

Image classification becomes a very challenging task when it involves classes that have shared characteristics and few data are available for training the classifier. Considering this problem, in this work we adopt a case study based on images from geoscience papers and investigate how different features can be combined in order to improve image classification results. In our investigation, we present a tool for evaluating class separability based on the position of the samples in a two-dimensional map according to different features. Moreover, we investigate the usefulness of classifiers’ membership probabilities for our scenario, validating if they can be used as reliable measures of the confidence in the predicted labels. Our experimental results show that it is possible to take advantage of deep learning models’ ability to learn discriminating features from data and combine them with hand-crafted features to improve classification. With this feature combination, we trained a Support Vector Machine (SVM) classifier whose results are better than the ones achieved using only deep learning.

Semi-Supervised Breast Histological Image Classification by Node-Attention Graph Transfer Network

As a major cause of leading female death, breast cancer is often diagnosed by histological images which has been resolved by many deep learning methods with the assistance of large amounts of annotated data. However, their performances are severely limited by the lack of sufficient labeled data in clinical practice. This paper aims to relieve the annotating workload by a semi-supervised transfer learning algorithm to conduct knowledge distillation from a completely labeled source domain. To achieve this goal, we propose a node-attention graph transfer network to exploit the inherent correlation between individual samples by graph convolutional network, along with a cross-domain graph learning module to stimulate the graph construction in target domain. In the meanwhile, we design a node-attention mechanism to learn the individual contribution of each source image for target domain, which can further leverage the domain-gap by our node-attention transfer learning. Results of semi-supervised breast histological image classification with various scales of annotated training images are performable and further experiments demonstrate the significant contributions of each component we proposed.

Convolutional Neural Network Based on Complex Networks for Brain Tumor Image Classification With a Modified Activation Function

The diagnosis of brain tumor types generally depends on the clinical experience of doctors, and computer-assisted diagnosis improves the accuracy of diagnosing tumor types. Therefore, a convolutional neural network based on complex networks (CNNBCN) with a modified activation function for the magnetic resonance imaging classification of brain tumors is presented. The network structure is not manually designed and optimized, but is generated by randomly generated graph algorithms. These randomly generated graphs are mapped into a computable neural network by a network generator. The accuracy of the modified CNNBCN model for brain tumor classification reaches 95.49%, which is higher than several models presented by other works. In addition, the test loss of brain tumor classification of the modified CNNBCN model is lower than those of the ResNet, DenseNet and MobileNet models in the experiments. The modified CNNBCN model not only achieves satisfactory results in brain tumor image classification, but also enriches the methodology of neural network design.

Channel Expansion Convolutional Network for Image Classification

With the continuous evolution of research on convolutional neural networks, it is an efficient and fashionable method to introduce attention mechanism into the convolutional structure. The channel attention designed in SENet has made a great contribution to the promotion of the attention convolution model. However, our research found that SENet focuses on certain feature channels rather than objects in the channels. It will simultaneously enhance or weaken the target objects and background information in a certain channel. On the basis of the channel attention convolution network, we first perform channel sorting and group convolution on the feature map, and expand each group to $\beta $ times the original feature channel during the group convolution process to construct a channel expansion convolution network (CENet), where $\beta $ is an array used to represent the channel expansion coefficient. CENet captures the attention of objects in the feature channel while expanding the proportion of features in the relatively important channel. Furthermore, we improved the structure of CENet and merged it into the intra-layer multi-scale convolutional model to construct an object-level attention multi-scale convolutional neural network (OAMS-CNN). We have conducted a large number of experiments on four data sets, CIFAR-10, CIFAR-100, FGVC-Aircraft and Stanford Cars. The experimental results show that our proposed new object-level attention convolution model has achieved good image classification results.

A New Loss Function for CNN Classifier Based on Predefined Evenly-Distributed Class Centroids

With the development of convolutional neural networks (CNNs) in recent years, the network structure has become more and more complex and varied, and has achieved very good results in pattern recognition, image classification, object detection and tracking. For CNNs used for image classification, in addition to the network structure, more and more research is now focusing on the improvement of the loss function, so as to enlarge the inter-class feature differences, and reduce the intra-class feature variations as soon as possible. Besides the traditional Softmax, typical loss functions include L-Softmax, AM-Softmax, ArcFace, and Center loss, etc. Based on the concept of predefined evenly-distributed class centroids (PEDCC) in CSAE network, this paper proposes a PEDCC-based loss function called PEDCC-Loss, which can make the inter-class distance maximal and intra-class distance small enough in hidden feature space. Multiple experiments on image classification and face recognition have proved that our method achieve the best recognition accuracy, and network training is stable and easy to converge. Code is available in https://github.com/ZLeopard/PEDCC-Loss

Multiscale Dynamic Graph Convolutional Network for Hyperspectral Image Classification

Convolutional neural network (CNN) has demonstrated impressive ability to represent hyperspectral images and to achieve promising results in hyperspectral image classification. However, traditional CNN models can only operate convolution on regular square image regions with fixed size and weights, and thus, they cannot universally adapt to the distinct local regions with various object distributions and geometric appearances. Therefore, their classification performances are still to be improved, especially in class boundaries. To alleviate this shortcoming, we consider employing the recently proposed graph convolutional network (GCN) for hyperspectral image classification, as it can conduct the convolution on arbitrarily structured non-Euclidean data and is applicable to the irregular image regions represented by graph topological information. Different from the commonly used GCN models that work on a fixed graph, we enable the graph to be dynamically updated along with the graph convolution process so that these two steps can be benefited from each other to gradually produce the discriminative embedded features as well as a refined graph. Moreover, to comprehensively deploy the multiscale information inherited by hyperspectral images, we establish multiple input graphs with different neighborhood scales to extensively exploit the diversified spectral–spatial correlations at multiple scales. Therefore, our method is termed multiscale dynamic GCN (MDGCN). The experimental results on three typical benchmark data sets firmly demonstrate the superiority of the proposed MDGCN to other state-of-the-art methods in both qualitative and quantitative aspects.

Computing fractal descriptors of texture images using sliding boxes: An application to the identification of Brazilian plant species

This work proposes a new model based on fractal descriptors for the classification of grayscale texture images. The method consists of scanning the image with a sliding box and collecting statistical information about the pixel distribution. Varying the box size, an estimation of the fractality of the image can be obtained at different scales, providing a more complete description of how such parameter changes in each image. The same strategy is also applied to a especial encoding of the image based on local binary patterns. Descriptors both from the original image and from the local encoding are combined to provide even more precise and robust results in image classification. A statistical model based on the theory of sliding window detection probabilities and Markov transition processes is formulated to explain the effectiveness of the method. The descriptors were tested on the identification of Brazilian plant species using scanned images of the leaf surface. The classification accuracy was also verified on three benchmark databases (KTH-TIPS2-b, UIUC and UMD). The results obtained demonstrate the power of the proposed approach in texture classification and, in particular, in the practical problem of plant species identification.

Mental models of AI-based systems: User predictions and explanations of image classification results

Humans should be able work more effectively with artificial intelligence-based systems when they can predict likely failures and form useful mental models of how the systems work. We conducted a study of human’s mental models of artificial intelligence systems focusing on a high-performing image classifier. Participants viewed individual labeled images in one of three general classes and then tried to predict whether the classifier would label it correctly. Participants were able to begin performing this task at levels much better than chance, 69% correct. However, after 137 trials with feedback, their performance improved a small, but statistically significant amount to 73%. Analysis of these results and comments indicated that humans were using their own perceptions of the images as first-approximation proxies. ‘Projecting’ human characteristics onto a computer might be considered a cognitive bias, but in this task, the strategy seemed to yield good initial results. This might be called effective anthropomorphism. Participants sometimes used this strategy both implicitly and explicitly. The paper includes discussion of why this strategy might have worked better than alternatives, why further learning was quite difficult, and what assumptions about similarities between human perception and image classification systems may in fact be correct.

Study on spatial distribution of coral reefs in Nam Yet island by using GIS and remote sensing techniques

Research and application of GIS and remote sensing techniques combined with field survey in coastal areas of Nam Yet island had been carried out to establish the distribution map of submarine habitats. Depth-invariant index was used to correct water column’s effects on spectral reflectance of each habitat. The results of satellite image classification showed that area with well-developed coral at great depths accounted for 12%, area with well-developed coral at small depths accounted for 9%, area with poorly-developed coral accounted for 13%, dead coral area accounted for 15% and area of sand, grit, pebbles and weathered coral accounted for 51%. The assessment after classification showed that the overall accuracy of the satellite image interpretation process was 94% and the kappa coefficient was 0.93.

Intelligence artificielle et radiothérapie : quelles bases et quelles perspectives ?

Artificial intelligence is a highly polysemic term. In computer science, with the objective of being able to solve totally new problems in new contexts, artificial intelligence includes connectionism (neural networks) for learning and logics for reasoning. Artificial intelligence algorithms mimic tasks normally requiring human intelligence, like deduction, induction, and abduction. All apply to radiation oncology. Combined with radiomics, neural networks have obtained good results in image classification, natural language processing, phenotyping based on electronic health records, and adaptive radiation therapy. General adversial networks have been tested to generate synthetic data. Logics based systems have been developed for providing formal domain ontologies, supporting clinical decision and checking consistency of the systems. Artificial intelligence must integrate both deep learning and logic approaches to perform complex tasks and go beyond the so-called narrow artificial intelligence that is tailored to perform some highly specialized task. Combined together with mechanistic models, artificial intelligence has the potential to provide new tools such as digital twins for precision oncology.

GenSynth: a generative synthesis approach to learning generative machines for generate efficient neural networks

The tremendous potential exhibited by deep learning is often offset by architectural and computational complexity, making widespread deployment a challenge for edge scenarios such as mobile and other consumer devices. To tackle this challenge, we explore the following idea: Can we learn generative machines to automatically generate deep neural networks with efficient network architectures? In this study, we introduce the idea of generative synthesis, which is premised on the intricate interplay between a generator-inquisitor pair that work in tandem to garner insights and learn to generate highly efficient deep neural networks that best satisfies operational requirements. Experimental results for image classification, semantic segmentation, and object detection tasks illustrate the efficacy of generative synthesis (GenSynth) in producing generators that automatically generate highly efficient deep neural networks (which we nickname FermiNets with higher model efficiency and lower computational costs (reaching $\gt 10\times $>10× more efficient and fewer multiply-accumulate operations than several tested state-of-the-art networks), as well as higher energy efficiency (reaching $\gt 4\times $>4× improvements in image inferences per joule consumed on a Nvidia Tegra X2 mobile processor). As such, GenSynth can be a powerful, generalised approach for accelerating and improving the building of deep neural networks for on-device edge scenarios.

A Modeling and Measurement Approach for the Uncertainty of Features Extracted from Remote Sensing Images

The reliability of remote sensing (RS) image classification is crucial for applying RS image classification results. However, it has received minimal attention, especially the uncertainty of features extracted from RS images. The uncertainty of image features constantly accumulates, propagates, and ultimately affects the reliability and accuracy of image classification results. Thus, research on the quantitative modeling and measurement of the feature uncertainty of RS images is very necessary. To make up for the lack of research on quantitative modeling and measurement of uncertainty of image features, this study first investigates and summarizes the appearance characteristics of the feature uncertainty of RS images in geospatial and feature space domains based on the source and formation mechanisms of feature uncertainty. Then, a modeling and measurement approach for the uncertainty of image features is proposed on the basis of these characteristics. In this approach, a new Local Adaptive Multi-Feature Weighting Method based on Information Entropy and the Local Distribution Density of Points is proposed to model and measure the feature uncertainty of an image in the geospatial and feature space domains. In addition, a feature uncertainty index is also constructed to comprehensively describe and quantify the feature uncertainty, which can also be used to refine the classification map to improve its accuracy. Finally, we propose two effectiveness verification schemes in two perspectives, namely, statistical analysis and image classification, to verify the validity of the proposed approach. Experimental results on two real RS images confirm the validity of the proposed approach. Our study on the feature uncertainty of images may contribute to the development of uncertainty control methods or reliable classification schemes for RS images.

A Lightweight and Discriminative Model for Remote Sensing Scene Classification With Multidilation Pooling Module

With the growing spatial resolution of satellite images, high spatial resolution (HSR) remote sensing imagery scene classification has become a challenging task due to the highly complex geometrical structures and spatial patterns in HSR imagery. The key issue in scene classification is how to understand the semantic content of the images effectively, and researchers have been looking for ways to improve the process. Convolutional neural networks (CNNs), which have achieved amazing results in natural image classification, were introduced for remote sensing image scene classification. Most of the researches to date have improved the final classification accuracy by merging the features of CNNs. However, the entire models become relatively complex and cannot extract more effective features. To solve this problem, in this paper, we propose a lightweight and effective CNN which is capable of maintaining high accuracy. We use MobileNet V2 as a base network and introduce the dilated convolution and channel attention to extract discriminative features. To improve the performance of the CNN further, we also propose a multidilation pooling module to extract multiscale features. Experiments are performed on six datasets, and the results verify that our method can achieve higher accuracy compared to the current state-of-the-art methods.

An Uncertainty Descriptor for Quantitative Measurement of the Uncertainty of Remote Sensing Images

Reliable image classification results are crucial for the application of remote sensing images, but the reliability of image classification has received less attention. In particular, the inherent uncertainty of remote sensing images has been disregarded. The uncertainty of a remote sensing image accumulates and propagates continuously in the classification process and ultimately affects the reliability of the classification results. Therefore, quantitative description and investigation of the inherent uncertainty of remote sensing images are crucial in achieving reliable remote sensing image classification. In this study, we analyze the sources of uncertainty of remote sensing images in detail and propose a quantitative descriptor for measuring image uncertainty comprehensively and effectively. In addition, we also design two verification schemes to verify the validity of the proposed uncertainty descriptor. Finally, the validity of the proposed uncertainty descriptor is confirmed by experimental results on three real remote sensing images. Our study on the uncertainty of remote sensing images may help the development of uncertainty control methods and reliable classification schemes of remote sensing images.

Mapping of Urban Features of Nnewi Metropolis Using High Resolution Satellite Image and Support Vector Machine Classifier

When maps and master plan of an area are available, they will definitely guide in the urban development, especially as a working document for enforcing planning laws by both government and the private urban developers. However, such basic geospatial information is reasonably lacking in this study area and sequel to this, the researcher aims at Mapping of Urban Features (Natural and Man-Made), which includes the Vegetation, wetlands/water bodies, buildings/pavement, open/bare surfaces and farm lands captured in GeoEye-1, High Resolution Satellite Image (HRSI) of 2016 using Support Vector Machine Classifier (SVMC) with a view of developing a reliable urban land use and land cover map of the area, which will serve as a base map for land-use planning and monitoring for a variety of end-users. The objectives include: to identify and extract features/regions of interest (ROIs) in a subset HRSI of the study area, to perform supervised classification using SVM in ENVI Software. The methodology used include Image acquisition, Image enhancement, Image Sub-setting, Image masking, Extraction of Regions of Interests (ROIs) and its separability index analysis, supervised classification using SVMC, Post-processing Accuracy Assessment, and Preparation of maps. Environment for Visualizing Image (ENVI 5.1) incorporated with Interactive Data Language (IDL 8.3) software was used for image processing, masking, spatial data analysis and image classification. Meanwhile, Esri ArcGIS 10.2 was employed for database development and production of thematic maps. Microsoft Excel and word was used for statistical analysis and result presentation. The result of image classification using SVMC, Radial Basis Function (BRF) default kernel in ENVI 5.1 indicates that Nnewi-North L.G.A is having 13.52% of Built-up Areas, 24.23% of Vegetation, 22.05% of Water bodies, Farm lands is equal to 39.40% and open/bare surface is 0.81% and result of image classification was validated using Error Matrix and Kappa Coefficient which results revealed that (SVMC overall Accuracy =98.07% and Kappa Coefficient = 0.97. The result revealed that ‘Support Vector Machine Classifier’ is robust in extracting urban landscape from HRSI, especially Built-up areas and open/bare for every Remote Sensing Analysis. The research recommends that it is imperative to check for ‘ROIs index separability’ before using it for classification, also there is need for periodic urban LULC analysis to guide stakeholders in Planning, Monitoring, and Management of Urban Areas.. Keywords: Support Vector Machine Classifier, Extraction of Regions of Interests (ROIs), ROIs separability index analysis, High Resolution Satellite Image (HRSI), Urban Landuse and Landcover. DOI : 10.7176/JEES/9-6-11 Publication date :June 30 th 2019