Decision Fusion Method Research Articles

Fusion Strategy for Improving Medical Image Segmentation

In this paper, we combine decision fusion methods with four meta-heuristic algorithms (Particle Swarm Optimization (PSO) algorithm, Cuckoo search algorithm, modification of Cuckoo Search (CS McCulloch) algorithm and Genetic algorithm) in order to improve the image segmentation. The proposed technique based on fusing the data from Particle Swarm Optimization (PSO), Cuckoo search, modification of Cuckoo Search (CS McCulloch) and Genetic algorithms are obtained for improving magnetic resonance images (MRIs) segmentation. Four algorithms are used to compute the accuracy of each method while the outputs are passed to fusion methods. In order to obtain parts of the points that determine similar membership values, we apply the different rules of incorporation for these groups. The proposed approach is applied to challenging applications: MRI images, gray matter/white matter of brain segmentations and original black/white images Behavior of the proposed algorithm is provided by applying to different medical images. It is shown that the proposed method gives accurate results; due to the decision fusion produces the greatest improvement in classification accuracy.

A Web Back-End Database Leakage Incident Reconstruction Framework Over Unlabeled Logs

In this article, we propose a web back-end database leakage incident reconstruction framework (WeB-DLIR) over unlabeled logs, designed to improve the intelligence and automation of reconstructing web back-end database leakage incidents triggered by web-based attacks in unannotated logging environments. Using WeB-DLIR, analysts can reduce the manual workload of tracing and responding to data leakage incidents. Specifically, we first design web front-end and back-end anomaly identification methods based on neural network models with a pruning strategy and fine-grained grouping clustering analysis, respectively, for completely identifying web-related abnormal events in unlabeled logs. To remove redundant abnormal events and reduce subsequent inspection work for false alarm cases, we then propose an anomaly detection result decision fusion method (DFADR). Moreover, to visualize the attack chain reflected by abnormal events, based on the decision fusion results, we propose an attack graph modeling method that can reflect the basic process of data leakage from multiple perspectives. Finally, based on the modeling results, the topology of the data leakage scenario reconstruction can be completed by further auditing the relevant logs. Experimental results using real-world datasets show that the proposed WeB-DLIR is efficient and feasible for practical applications.

Bayesian Fusion Model Enhanced Codfish Classification Using Near Infrared and Raman Spectrum.

In this study, a Bayesian-based decision fusion technique was developed for the first time to quickly and non-destructively identify codfish using near infrared (NIRS) and Raman spectroscopy (RS). NIRS and RS spectra from 320 codfish samples were collected, and separate partial least squares discriminant analysis (PLS-DA) models were developed to establish the relationship between the raw data and cod identity for each spectral technique. Three decision fusion methods: decision fusion, data layer or feature layer, were tested and compared. The decision fusion model based on the Bayesian algorithm (NIRS-RS-B) was developed on the optimal discrimination features of NIRS and RS data (NIRS-RS) extracted by the PLS-DA method whereas the other fusion models followed conventional, non-Bayesian approaches. The Bayesian model showed enhanced classification metrics (92% sensitivity, 98% specificity, 98% accuracy) that were significantly superior to those demonstrated by any of other two spectroscopic methods (NIRS, RS) and the two data fusion methods (data layer fused, NIRS-RS-D, or feature layer fused, NIRS-RS-F). This novel proposed approach can provide an alternative classification for codfish and potentially other food speciation cases.

Localization and classification of space objects using EfficientDet detector for space situational awareness

Space situational awareness (SSA) systems play a significant role in space navigation missions. One of the most essential tasks of this system is to recognize space objects such as spacecrafts and debris for various purposes including active debris removal, on-orbit servicing, and satellite formation. The complexity of object recognition in space is due to several sensing conditions, including the variety of object sizes with high contrast, low signal-to-noise ratio, noisy backgrounds, and several orbital scenarios. Existing methods have targeted the classification of images containing space objects with complex backgrounds using various convolutional neural networks. These methods sometimes lose attention on the objects in these images, which leads to misclassification and low accuracy. This paper proposes a decision fusion method that involves training an EfficientDet model with an EfficientNet-v2 backbone to detect space objects. Furthermore, the detected objects were augmented by blurring and by adding noise, and were then passed into the EfficientNet-B4 model for training. The decisions from both models were fused to find the final category among 11 categories. The experiments were conducted by utilizing a recently developed space object dataset (SPARK) generated from realistic space simulation environments. The dataset consists of 11 categories of objects with 150,000 RGB images and 150,000 depth images. The proposed object detection solution yielded superior performance and its feasibility for use in real-world SSA systems was demonstrated. Results show significant improvement in accuracy (94%), and performance metric (1.9223%) for object classification and in mean precision (78.45%) and mean recall (92.00%) for object detection.

KIND: A Novel Image-Mutual-Information-Based Decision Fusion Method for Saturation Attack Detection in SD-IoT

Software-defined networking for IoT (SD-IoT), as an emerging architecture, is suffering from numerous security issues. Saturation attacks against the SDN switches and controllers are major security concerns in SD-IoT. When using the Dempster–Shafer evidence theory (DSE) to detect various saturation attacks, the simple mutual information calculation may cause information loss problem, leading to the decrease of detection accuracy. Therefore, how to avoid information loss problem to improve the detection performance is a key issue. Aiming to solve the above-mentioned issues, we propose KIND, a novel image mutual information-based decision fusion method for saturation attack detection in SD-IoT. The main idea of KIND is that it converts the probability matrix of binary classifiers to images and detects the saturation attacks by fusing these images. More specifically, when executing the evidence fusion, each evidence is converted to an image by a nonlinear transformation method. Each image is converted from the related probability-supported matrix. Then, the evidence can be fixed by comparing structural similarities among different images. After that, all evidence can be utilized to obtain the final detection results using the conducted combination rule. The evaluation results demonstrate that KIND can achieve high detection performance, which outperforms other state-of-the-art methods, in terms of TPR, TNR, FPR, FNR, accuracy, precision, recall, F-score, confusion matrix, ROC curves, PR curves, Chi-square test, correlation coefficient, and the quantitative strategy test. In conclusion, KIND can detect saturation attacks with high precision while causing acceptable storage overload.

Embedded AMIS-Deep Learning with Dialog-Based Object Query System for Multi-Class Tuberculosis Drug Response Classification.

A person infected with drug-resistant tuberculosis (DR-TB) is the one who does not respond to typical TB treatment. DR-TB necessitates a longer treatment period and a more difficult treatment protocol. In addition, it can spread and infect individuals in the same manner as regular TB, despite the fact that early detection of DR-TB could reduce the cost and length of TB treatment. This study provided a fast and effective classification scheme for the four subtypes of TB: Drug-sensitive tuberculosis (DS-TB), drug-resistant tuberculosis (DR-TB), multidrug-resistant tuberculosis (MDR-TB), and extensively drug-resistant tuberculosis (XDR-TB). The drug response classification system (DRCS) has been developed as a classification tool for DR-TB subtypes. As a classification method, ensemble deep learning (EDL) with two types of image preprocessing methods, four convolutional neural network (CNN) architectures, and three decision fusion methods have been created. Later, the model developed by EDL will be included in the dialog-based object query system (DBOQS), in order to enable the use of DRCS as the classification tool for DR-TB in assisting medical professionals with diagnosing DR-TB. EDL yields an improvement of 1.17-43.43% over the existing methods for classifying DR-TB, while compared with classic deep learning, it generates 31.25% more accuracy. DRCS was able to increase accuracy to 95.8% and user trust to 95.1%, and after the trial period, 99.70% of users were interested in continuing the utilization of the system as a supportive diagnostic tool.

Evaluation of Decision Fusions for Classifying Karst Wetland Vegetation Using One-Class and Multi-Class CNN Models with High-Resolution UAV Images

Combining deep learning and UAV images to map wetland vegetation distribution has received increasing attention from researchers. However, it is difficult for one multi-classification convolutional neural network (CNN) model to meet the accuracy requirements for the overall classification of multi-object types. To resolve these issues, this paper combined three decision fusion methods (Majority Voting Fusion, Average Probability Fusion, and Optimal Selection Fusion) with four CNNs, including SegNet, PSPNet, DeepLabV3+, and RAUNet, to construct different fusion classification models (FCMs) for mapping wetland vegetations in Huixian Karst National Wetland Park, Guilin, south China. We further evaluated the effect of one-class and multi-class FCMs on wetland vegetation classification using ultra-high-resolution UAV images and compared the performance of one-class classification (OCC) and multi-class classification (MCC) models for karst wetland vegetation. The results highlight that (1) the use of additional multi-dimensional UAV datasets achieved better classification performance for karst wetland vegetation using CNN models. The OCC models produced better classification results than MCC models, and the accuracy (average of IoU) difference between the two model types was 3.24–10.97%. (2) The integration of DSM and texture features improved the performance of FCMs with an increase in accuracy (MIoU) from 0.67% to 8.23% when compared to RGB-based karst wetland vegetation classifications. (3) The PSPNet algorithm achieved the optimal pixel-based classification in the CNN-based FCMs, while the DeepLabV3+ algorithm produced the best attribute-based classification performance. (4) Three decision fusions all improved the identification ability for karst wetland vegetation compared to single CNN models, which achieved the highest IoUs of 81.93% and 98.42% for Eichhornia crassipes and Nelumbo nucifera, respectively. (5) One-class FCMs achieved higher classification accuracy for karst wetland vegetation than multi-class FCMs, and the highest improvement in the IoU for karst herbaceous plants reached 22.09%.

A decision fusion method based on classification models for water quality monitoring.

Monitoring of water quality is one of the world's main intentions for countries. Classification techniques based on support vector machines (SVMs) and artificial neural network (ANN) has been widely used in several applications of water research. Water quality assessment with high accuracy and efficiency with innovational approaches permitted us to acquire additional knowledge and information to obtain an intelligent monitoring system. In this paper, we present the use of principal component analysis (PCA) combined with SVM and ANN with decision templates combination data fusion method. PCA was used for features selection from original database. The multi-layer perceptron network (MLP) and the one-against-all strategy for SVM method have been widely used. Decision templates are applied to increase the accuracy of the water quality classification. The specific classification approach was employed to assess the water quality of the Tilesdit dam in Algeria as a study area, defined with a dataset of eight physicochemical parameters collected in the period 2009-2018, such as temperature, pH, electrical conductivity, and turbidity. The selection of the excellent parameters of the used models can be improving the performance of classification process. In order to assess their results, an experiment step using collected dataset corresponding to the accuracy and running time of training and test phases, and robustness to noise, is carried out. Various scenarios are examined in comparative study to obtain the most results of decision step with and without feature selection of the input data. From the results, we found that the integration of SVM and ANN with PCA yields accuracy up than 98%. The combination by decision templates of two classifiers SVM and ANN with PCA yields an accuracy of 99.24% using k-fold cross-validation. The combination data fusion enhanced expressively the results of the proposed monitoring framework that had proven a considerable ability in surface water quality assessment.

Decision fusion of two sensors object classification based on the evidential reasoning rule

The application of multiple sensors for object classification is commonly used in unmanned intelligent systems such as autonomous driving and robots. Current researches based on evidential reasoning rules mainly consider multiple sensor fusion. The reliability is determined by the decisions of the majority of sensors and the weight is determined by historical accuracy. However, the existing researches using the evidential reasoning rule are not applicable when there are only two sensors working in some situations. This paper proposes a two sensors decision fusion method for object classification based on the evidential reasoning rule. The reliability of sensors is calculated with the Logistic model based on the differences of the classification decisions in a certain time span and the adaptive weights of sensors are calculated based on the coefficient of variation to adapt to different environments. The final decision is drawn out by fusing the classification decisions of two sensors based on the fusion rules of evidential reasoning rule. Comparative experimental studies of object classification decision fusion in autonomous driving are carried out with the Nuscenes dataset and the Waymo dataset. The results show that the method proposed can effectively improve the accuracy of the fusion when there are only two sensors.

MBES Seabed Sediment Classification Based on a Decision Fusion Method Using Deep Learning Model

High-precision habitat mapping can contribute to the identification and quantification of the human footprint on the seafloor. As a representative of seafloor habitats, seabed sediment classification is crucial for marine geological research, marine environment monitoring, marine engineering construction, and seabed biotic and abiotic resource assessment. Multibeam echo-sounding systems (MBES) have become the most popular tool in terms of acoustic equipment for seabed sediment classification. However, sonar images tend to consist of obvious noise and stripe interference. Furthermore, the low efficiency and high cost of seafloor field sampling leads to limited field samples. The factors above restrict high accuracy classification by a single classifier. To further investigate the classification techniques for seabed sediments, we developed a decision fusion algorithm based on voting strategies and fuzzy membership rules to integrate the merits of deep learning and shallow learning methods. First, in order to overcome the influence of obvious noise and the lack of training samples, we employed an effective deep learning framework, namely random patches network (RPNet), for classification. Then, to alleviate the over-smoothness and misclassifications of RPNet, the misclassified pixels with a lower fuzzy membership degree were rectified by other shallow learning classifiers, using the proposed decision fusion algorithm. The effectiveness of the proposed method was tested in two areas of Europe. The results show that RPNet outperforms other traditional classification methods, and the decision fusion framework further improves the accuracy compared with the results of a single classifier. Our experiments predict a promising prospect for efficiently mapping seafloor habitats through deep learning and multi-classifier combinations, even with few field samples.

A Deep Learning Method for Early Detection of Diabetic Foot Using Decision Fusion and Thermal Images

Diabetes mellitus (DM) is one of the major diseases that cause death worldwide and lead to complications of diabetic foot ulcers (DFU). Improper and late handling of a diabetic foot patient can result in an amputation of the patient’s foot. Early detection of DFU symptoms can be observed using thermal imaging with a computer-assisted classifier. Previous study of DFU detection using thermal image only achieved 97% of accuracy, and it has to be improved. This article proposes a novel framework for DFU classification based on thermal imaging using deep neural networks and decision fusion. Here, decision fusion combines the classification result from a parallel classifier. We used the convolutional neural network (CNN) model of ShuffleNet and MobileNetV2 as the baseline classifier. In developing the classifier model, firstly, the MobileNetV2 and ShuffleNet were trained using plantar thermogram datasets. Then, the classification results of those two models were fused using a novel decision fusion method to increase the accuracy rate. The proposed framework achieved 100% accuracy in classifying the DFU thermal images in binary classes of positive and negative cases. The accuracy of the proposed Decision Fusion (DF) was increased by about 3.4% from baseline ShuffleNet and MobileNetV2. Overall, the proposed framework outperformed in classifying the images compared with the state-of-the-art deep learning and the traditional machine-learning-based classifier.

A decision fusion method based on hyperspectral imaging and electronic nose techniques for moisture content prediction in frozen-thawed pork

This paper studies the feasibility of combining hyperspectral imaging (HSI) and electronic nose (E-nose) techniques to improve the prediction performance of moisture content (MC) in frozen-thawed pork. In this experiment, the data of 240 pork samples under different freeze-thaw cycles (0–5 times) were collected using HSI and E-nose technology. Afterward, the spectra and image (color and texture) information were extracted from the HSI sensor, while the odor information was extracted from the E-nose sensor. In this setting, an improved decision fusion method was proposed to fuse this information, thus detecting the MC in pork. Based on single information and its fusion at different fusion levels (pixel, feature, traditional decision fusion, and improved decision fusion), partial least squares regression (PLSR) prediction models were established and compared. The results show that the model of improved decision fusion based on HSI and E-nose fusion technology exhibits the best prediction capability with the determination coefficient of prediction set (R2p) of 0.9533 and root mean square error of prediction set (RMSEP) of 0.3869. This study demonstrates that the proposed method is an effective data fusion method, and the combination of HSI and E-nose technology improves the prediction performance of MC in frozen-thawed pork.

Research on Music Emotional Expression Based on Reinforcement Learning and Multimodal Information

With the continuous development of the research in the field of emotion analysis, music, as a common multimodal information carrier in people’s daily life, often transmits emotion through lyrics and melody, so it has been gradually incorporated into the research category of emotion analysis. The fusion classification model based on CNN-LSTM proposed in this paper effectively improves the accuracy of emotional classification of audio and lyrics. At the same time, in view of the problem that the traditional decision-level fusion method ignores the correlation between modes and the limitations of dataset, this paper further improves the existing Thayer dimension emotional decision fusion method, takes the audio energy axis data as the main discrimination basis, and improves the accuracy of decision fusion classification. Based on the results of music emotion analysis, this paper further carries out the task of music generation. Based on the feature that there is often consistent emotional expression between music words and songs, a dual Seq2Seq framework based on reinforcement learning is constructed. By introducing the reward value of emotional consistency and content fidelity, the output melody has the same emotion with the input lyrics and good results are achieved. Compared with the ordinary Seq2Seq, the accuracy of our proposed model is improved by about 1.1%. This shows that the accuracy of the model can be effectively improved by using reinforcement learning.

Deep Learning-Based Music Quality Analysis Model.

In order to build an efficient and effective deep much quality recognition model, a decision fusion method by leveraging the advantages of shallow learning and deep learning is formulated. In the literature, shallow learning is a traditional music-related quality recognition method, that is, artificial statistical feature extraction and recognition are designed. Meanwhile, our deep learning module leverages the so-called PCANET network to implement the feature extraction process, and subsequently takes the spectrogram describing the music-related information as the network input. First, a variety of task classifications for the music signal problem are divided. Afterward, the optimization and adoption of deep learning in the two major problems of music feature extraction and sequence modeling are introduced. Finally, a music application is presented to illustrate the practical application of deep learning in music quality evaluation. The shallow learning features and deep learning features are seamlessly combined into the SVM model for music quality modeling, based on which differential voting mechanisms are leveraged to realize the fusion of decision-making layers. Extensive experimental results have shown that the music quality recognition rate by this method can be significantly improved on our own compiled library and the Berlin database. Besides, it exhibits obvious advantages compared with the competitors.

Sentiment Classification of Tourism Reviews Based on Visual and Textual Multifeature Fusion

Most of the traditional tourism review sentiment analysis methods ignore the complementary role between review text and images and do not pay attention to the differences in the sentimental expression of different users and the characteristics of the text content of tourism reviews other than the plain text such as emojis. For this reason, a sentiment analysis method for tourism online reviews based on multifeature fusion of images and text is proposed. Firstly, a text sentiment classification model is constructed, and a variety of sentiment features are combined to form a multi-input matrix, and then, it is input into a multichannel CNN (convolutional neural network) to extract sentiment features in order to complete the text sentiment classification. In addition, an image sentiment classification model is constructed by merging the global image and the face image. On the basis of the CNN, the supervision module with weighted loss is added to extract the facial sentiment features, and the facial target sentiment is fused with the sentiment directly recognized by the whole image, and the sentiment polarity of the image in the posted tourism review is determined. Finally, a decision fusion method is designed to fuse the output of the text and image sentiment classification models. The experimental results show that the proposed image-text fusion sentiment classification model effectively enhances the ability of the model to capture sentimental semantics of tourism reviews through the combination of text content features and image sentiment features and achieves excellent results in multiple performance metrics, with better sentiment classification performance than other state-of-the-art models.

Moving Target Detection Using Fuzzy Bayesian Fusion in Multichannel SAR Framework

In this paper, a novel FBF-MTD is proposed for the detection of moving target by integrating the fuzzy concept in Bayesian fusion model. This method uses the decision fusion method that combines the matching filter, Fourier transform and the STFT. In the first step, acceleration, velocity, and RCS are simulated and the radar that returns from the target is calculated based on transmission power, distance of target, antenna gain, and RCS. Then, the FBF-MTD method combines the results of Fourier transform, short time Fourier transform, and matched filter, to produce the final decision. The performance of the proposed FBF-MTD method is analyzed with respect to the metrics, namely detection time, missed target rate, and MSE. The proposed FBF-MTD model obtained the detection time, missed target rate, and MSE values of 3.2495 sec, 0.0524, and 3344.04, respectively that show the superiority of the FBF-MTD model in MTD.

Cervical Lesion Classification Method Based on Cross-Validation Decision Fusion Method of Vision Transformer and DenseNet.

Objective In order to better adapt to clinical applications, this paper proposes a cross-validation decision-making fusion method of Vision Transformer and DenseNet161. Methods The dataset is the most critical acetic acid image for clinical diagnosis, and the SR areas are processed by a specific method. Then, the Vision Transformer and DenseNet161 models are trained by the fivefold cross-validation method, and the fivefold prediction results corresponding to the two models are fused by different weights. Finally, the five fused results are averaged to obtain the category with the highest probability. Results The results show that the fusion method in this paper reaches an accuracy rate of 68% for the four classifications of cervical lesions. Conclusions It is more suitable for clinical environments, effectively reducing the missed detection rate and ensuring the life and health of patients.

Research on Decision Fusion Method Based on Mixed Attributes

In order to solve the issues that different decision methods cannot be effectively fused with mixed attributes, this study proposes the idea of maximizing the total deviation between the evaluation values of multiple single evaluation methods, constructs a decision fusion model based on the deviation maximization method, obtains the combined evaluation results of each evaluation object via calculation, and ranks them accordingly, either superior or inferior. Secondly, the Spearman’s correlation coefficient is used to analyze the stability of the evaluation value of the portfolio. Finally, by combining the analysis in this study with a comparative analysis from previous papers, the results showed the scientific validity and effectiveness of the fusion method.

MVFusFra: A Multi-View Dynamic Fusion Framework for Multimodal Brain Tumor Segmentation.

Medical practitioners generally rely on multimodal brain images, for example based on the information from the axial, coronal, and sagittal views, to inform brain tumor diagnosis. Hence, to further utilize the 3D information embedded in such datasets, this paper proposes a multi-view dynamic fusion framework (hereafter, referred to as MVFusFra) to improve the performance of brain tumor segmentation. The proposed framework consists of three key building blocks. First, a multi-view deep neural network architecture, which represents multi learning networks for segmenting the brain tumor from different views and each deep neural network corresponds to multi-modal brain images from one single view. Second, the dynamic decision fusion method, which is mainly used to fuse segmentation results from multi-views into an integrated method. Then, two different fusion methods (i.e., voting and weighted averaging) are used to evaluate the fusing process. Third, the multi-view fusion loss (comprising segmentation loss, transition loss, and decision loss) is proposed to facilitate the training process of multi-view learning networks, so as to ensure consistency in appearance and space, for both fusing segmentation results and the training of the learning network. We evaluate the performance of MVFusFra on the BRATS 2015 and BRATS 2018 datasets. Findings from the evaluations suggest that fusion results from multi-views achieve better performance than segmentation results from the single view, and also implying effectiveness of the proposed multi-view fusion loss. A comparative summary also shows that MVFusFra achieves better segmentation performance, in terms of efficiency, in comparison to other competing approaches.

Evaluation of Decision Fusion Methods for Multimodal Biometrics in the Banking Application.

An evaluation of decision fusion methods based on Dempster-Shafer Theory (DST) and its modifications is presented in the article, studied over real biometric data from the engineered multimodal banking client verification system. First, the approaches for multimodal biometric data fusion for verification are explained. Then the proposed implementation of comparison scores fusion is presented, including details on the application of DST, required modifications, base probability, and mass conversions. Next, the biometric verification process is described, and the engineered biometric banking system principles are provided. Finally, the validation results of three fusion approaches on synthetic and real data are presented and discussed, considering the desired outcome manifested by minimized false non-match rates for various assumed thresholds and biometric verification techniques.