Journal of Intelligent & Fuzzy Systems - Volume 45, issue 4

Authors: Liu, Baokai | He, Fengjie | Du, Shiqiang | Li, Jiacheng | Liu, Wenjie

Article Type: Research Article

Abstract: Small object detection has important application value in the fields of autonomous driving and drone scene analysis. As one of the most advanced object detection algorithms, YOLOv3 suffers some challenges when detecting small objects, such as the problem of detection failure of small objects and occluded objects. To solve these problems, an improved YOLOv3 algorithm for small object detection is proposed. In the proposed method, the dilated convolutions mish (DCM) module is introduced into the backbone network of YOLOv3 to improve the feature expression ability by fusing the feature maps of different receptive fields. In the neck network of YOLOv3, …the convolutional block attention module (CBAM) and multi-scale fusion module are introduced to select the important information for small object detection in the shallow network, suppress the uncritical information, and use the fusion module to fuse the feature maps of different scales, so as to improve the detection accuracy of the algorithm. In addition, the Soft-NMS and Complete-IOU (ClOU) strategies are applied to candidate frame screening, which improves the accuracy of the algorithm for the detection of occluded objects. The experimental results on MS COCO2017, VOC2007, VOC2012 datasets and the ablation experiments on MS COCO2017 datasets demonstrate the effectiveness of the proposed method.The experimental results show that the proposed method achieves better accuracy in small object detection than the original YOLOv3 model. Show more

Keywords: Small object detection, Dilated convolutions mish, Fusion module, Soft-NMS

DOI: 10.3233/JIFS-224530

Citation: Journal of Intelligent & Fuzzy Systems, vol. 45, no. 4, pp. 5807-5819, 2023

Authors: Jiang, Minghua | Wang, Yulin | Yu, Feng | Peng, Tao | Hu, Xinrong

Article Type: Research Article

Abstract: Forest fires can pose a serious threat to the survival of living organisms, and wildfire detection technology can effectively reduce the occurrence of large forest fires and detect them faster. However, the unpredictable and diverse appearance of smoke and fire, as well as interference from objects that resemble smoke and fire, can lead to the overlooking of small objects and detection of false positives that resemble the objects in the detection results. In this work, we propose UAV-FDN, a forest fire detection network based on the perspective of an unmanned aerial vehicle (UAV). It performs real-time wildfire detection of various …forest fire scenarios from the perspective of UAVs. The main concepts of the framework are as follows: 1) The framework proposes an efficient attention module that combines channel and spatial dimension information to improve the accuracy and efficiency of model detection under complex backgrounds. 2) It also introduces an improved multi-scale fusion module that enhances the network’s ability to learn objects details and semantic features, thus reducing the chances of small objects being false negative during inspection and false positive issues. 3) Finally, the framework incorporates a multi-head structure and a new loss function, which aid in boosting the network’s updating speed and convergence, enabling better adaptation to different objects scales. Experimental results demonstrate that the UAV-FDN achieves high performance in terms of average precision (AP), precision, recall, and mean average precision (mAP). Show more

Keywords: Forest fire, wildfire detection, unmanned aerial vehicle, deep learning, attention mechanism, multi-scale feature fusion

DOI: 10.3233/JIFS-231550

Citation: Journal of Intelligent & Fuzzy Systems, vol. 45, no. 4, pp. 5821-5836, 2023

Authors: Guo, An | Sun, Kaiqiong | Wang, Meng

Article Type: Research Article

Abstract: While deep learning based object detection methods have achieved high accuracy in fruit detection, they rely on large labeled datasets to train the model and assume that the training and test samples come from the same domain. This paper proposes a cross-domain fruit detection method with image and feature alignments. It first converts the source domain image into the target domain through an attention-guided generative adversarial network to achieve the image-level alignment. Then, the knowledge distillation with mean teacher model is fused in the yolov5 network to achieve the feature alignment between the source and target domains. A contextual aggregation …module similar to a self-attention mechanism is added to the detection network to improve the cross-domain feature learning by learning global features. A source domain (orange) and two target domain (tomato and apple) datasets are used for the evaluation of the proposed method. The recognition accuracy on the tomato and apple datasets are 87.2% and 89.9%, respectively, with an improvement of 10.3% and 2.4%, respectively, compared to existing methods on the same datasets. Show more

Keywords: Domain adaptation, deep learning, knowledge distillation, fruit detection

DOI: 10.3233/JIFS-232104

Citation: Journal of Intelligent & Fuzzy Systems, vol. 45, no. 4, pp. 5837-5851, 2023

Authors: Liu, Junhui | Li, Guozhu | Gao, Chen

Article Type: Research Article

Abstract: In this study, we are concerned with the optimization of fuzzy clustering (Fuzzy C-Means) on the basis of a collection of distributed datasets without violating data confidentiality and security. The optimization of fuzzy clusters is realized using the differential evolution algorithm in a federated learning environment. Fuzzy clustering plays an important role in revealing the underlying structure of a given dataset. However, traditional iterative method is easy to get stuck at local optimum. With the growing concerning on data confidentiality and security, how to reveal the underlying structure of the data that are stored locally across different sites is becoming …an urgent problem. In order to overcome these two obstacles, we propose a federated differential evolution algorithm to realize fuzzy clustering. We augment the well-known differential evolution algorithm such that it can work in a federated learning environment to ensure local data privacy. The design practice of the federated differential evolution is elaborated on by highlighting its effectiveness in finding the optimal fuzzy clusters on the basis of distributed datasets. The performance of the proposed method is compared with traditional fuzzy clustering algorithm. Experimental studies completed on a series of real-world datasets coming from machine learning repository are reported to demonstrate the superiority of the proposed algorithm. Show more

Keywords: Differential evolution, horizontal federated learning, fuzzy clustering, global optimization

DOI: 10.3233/JIFS-232709

Citation: Journal of Intelligent & Fuzzy Systems, vol. 45, no. 4, pp. 5853-5860, 2023

Authors: Wang, Yajun

Article Type: Research Article

Abstract: In order to improve the detection accuracy of high-voltage dense channel satellite image, a satellite target detection algorithm based on deep learning is proposed. The convolution neural network is selected to extract the feature map of high-voltage dense channel satellite image, and the extracted feature map is input into the optimized deformation convolution neural network. The value of each sampling point and the corresponding position authority of block convolution kernel are weighted by using the regular region sampling feature map. The feature map output by the convolution operation of pooling layer is used to obtain the depth features of the …same dimension. The depth feature is input into the full connection layer to obtain the full connection feature of candidate target area, and the target detection in high-voltage dense channel satellite image is realized. The experimental results show that the target detection accuracy of the method is higher than 99% and the false alarm rate and false alarm rate are lower than 1.4%. Show more

Keywords: Deep learning, high voltage dense channel, satellite, target detection algorithm, convolution neural network, regular region

DOI: 10.3233/JIFS-223936

Citation: Journal of Intelligent & Fuzzy Systems, vol. 45, no. 4, pp. 5861-5869, 2023

Authors: Che, Gaofeng | Yu, Zhen

Article Type: Research Article

Abstract: In this work, the output-feedback fault-tolerant tacking control issue for underactuated autonomous underwater vehicle (AUV) with actuators faults is investigated. Firstly, an output-feedback error tacking system is constructed based on the theoretical model of underactuated AUV with actuators faults. Then, an adaptive dynamic programming (ADP) based fault-tolerant control controller is developed. In our proposed control scheme, a neural-network observer is designed to approximate the system states with actuators faults. An online policy iteration algorithm is designed with critic network and action network in order to improve the tracking accuracy. Based on Lyapunov stability theorem, the stability of the error tracking …system is guaranteed by the proposed controller. At last, the simulation results show that the underactuated AUV achieves better tracking performance. Show more

Keywords: Adaptive dynamic programming (ADP), fault-tolerant tracking control, actuators faults, neural network observer, autonomous underwater vehicle (AUV)

DOI: 10.3233/JIFS-223976

Citation: Journal of Intelligent & Fuzzy Systems, vol. 45, no. 4, pp. 5871-5883, 2023

Authors: Xu, Fei | Wang, Peng | Xu, Huimin

Article Type: Research Article

Abstract:  Deep convolutional neural networks (DCNNs) have shown remarkable performance in image classification tasks in recent years. In the network structure of DPRN, as the network depth increases, the number of convolutional kernels also increases linearly or nonlinearly. On the one hand, in the DPRN block, the size of the receptive field is only 3 × 3, which results in insufficient network ability to extract feature map information of different filter sizes. On the other hand, the number of convolution kernels in the second 1x1 convolution will be multiplied by a coefficient relative to the first convolution, which can cause overfitting to some …extent. In order to overcome these weaknesses, we introduce the inception-like structure on the basis of the DPRN network which is called by pyramid inceptional residual networks (PIRN). In addition, we also discuss the performance of PIRN network with squeeze and excitation (SE) mechanism and regularization term. Furthermore, some results in network performance are discussed when adding a stochastic depth networkto the PIRN model. Compared to DPRN, PIRN achieved better results on the CIFAR10, CIFAR100, and Mini-ImageNet datasets. In the case of using zero-padding, the multiplicative PIRN with SE mechanism achieves the best result of 95.01% on the CIFAR10 dataset. Meanwhile, on the CIFAR100 and Mini-ImageNet datasets, the additive PIRN network with a network depth of 92 achieves the best results of 76.06% and 65.86%, respectively. According to the experimental results, our method has achieved better accuray than that of DPRN with same network settings which demonstrate its effectiveness in generalization ability. Show more

Keywords: Convolution neural network, Deep pyramidal residual network, Squeeze and excitation mechanism, Pyramidal inceptional residual network, L2 regularization

DOI: 10.3233/JIFS-230569

Citation: Journal of Intelligent & Fuzzy Systems, vol. 45, no. 4, pp. 5885-5906, 2023

Authors: Zhang, Dong | Liu, Jinzhu | Liu, Duo | Li, Guanyu

Article Type: Research Article

Abstract: Knowledge graphs exhibit a typical hierarchical structure and find extensive applications in various artificial intelligence domains. However, large-scale knowledge graphs need to be completed, which limits the performance of knowledge graphs in downstream tasks. Knowledge graph embedding methods have emerged as a primary solution to enhance knowledge graph completeness. These methods aim to represent entities and relations as low-dimensional vectors, focusing on handling relation patterns and multi-relation types. Researchers need to pay more attention to the crucial feature of hierarchical relationships in real-world knowledge graphs. We propose a novel knowledge graph embedding model called H ierarchy-Aware P aired R elation …Vectors Knowledge Graph E mbedding (HPRE) to bridge this gap. By leveraging the power of 2D coordinates, HPRE adeptly model relation patterns, multi-relation types, and hierarchical features in the knowledge graph. Specifically, HPRE employs paired relation vectors to capture the distinct characteristics of head and tail entities, facilitating a better fit for relational patterns and multi-relation scenarios. Additionally, HPRE employs angular coordinates to differentiate entities at various levels of the hierarchy, effectively representing the hierarchical nature of the knowledge graph. The experimental results show that the HPRE model can effectively learn the hierarchical features of the knowledge graph and achieve state-of-the-art experimental results on multiple real-world datasets for the link prediction task. Show more

Keywords: Knowledge graph completion, link prediction, knowledge graph embedding, knowledge graph representation

DOI: 10.3233/JIFS-230982

Citation: Journal of Intelligent & Fuzzy Systems, vol. 45, no. 4, pp. 5907-5926, 2023

Authors: Wang, Hejin | He, Mingzhao | Zeng, Chengli | Qian, Lei | Wang, Jun | Pan, Wu

Article Type: Research Article

Abstract: Immersive virtual reality technology has been widely used in teaching and learning scenarios because of its unique visual and interactive experiences that bring learners a sense of immersive reality. However, how to better apply immersive virtual reality technology to learning environments to promote learning effectiveness is a direction that has been studied and explored by many scholars. Although a growing number of studies have concluded that immersive virtual reality technology can enhance learners’ attention in teaching and learning, few studies have directly linked both learning behaviors and attention to investigate the differences in behavioral performance across attention. In this study, …attention data monitored by EEG physiological brainwaves and a large number of videos recorded during learning were used to explore the differences in the sequence of high attention behaviors across performance levels in an immersive virtual reality environment using behavioral data mining techniques. The results found that there was a strong correlation between attention and performance in immersive virtual reality, that thinking and looking may be more conducive to learners’ concentration, and that high concentration behaviors in the high-performing group accompanied the test and appeared after the monitoring, while the action continued to be repeated after the high concentration behaviors in the low-performing group. Based on this, this study provides a reference method for the analysis of the learning process in this environment, and provides a theoretical basis and practical guidance for the improvement of participants’ attention and learning effectiveness. Show more

Keywords: Immersive virtual reality, EEG feedback, learning behaviour, data mining

DOI: 10.3233/JIFS-231383

Citation: Journal of Intelligent & Fuzzy Systems, vol. 45, no. 4, pp. 5927-5938, 2023

Authors: Chen, Fu

Article Type: Research Article

Abstract: How to guarantee the quality of college physical education (PE) teaching and reverse the declining trend of college students’ physique year by year has become a hot topic for the research of higher education and school PE workers. The quality assurance of higher education in China should give full play to the role of colleges in teaching quality assurance activities, constantly improve the level of school running and improve the efficiency of school running. Because colleges themselves are the main body of higher education and teaching activities, they have the most power, qualification and responsibility to explain the quality of …higher education. The classroom teaching quality (CTQ) evaluation of college badminton training is regarded as multi-attribute decision-making (MADM). The 2-tuple linguistic neutrosophic sets (2TLNSs) which the truth-membership, indeterminacy-membership and the falsity-membership are assessed by using the 2-tuple linguistic term sets is an appropriate form to express the indeterminate decision-making information in the classroom teaching quality (CTQ) evaluation of college badminton training. In this paper, the Hamy mean (HM) and the power average (PA) are connected with 2-tuple linguistic neutrosophic sets (2TLNSs) to propose the 2-tuple linguistic neutrosophic numbers weighted power HM (2TLNWPHM) operator. Then, use the 2TLNWPHM operator to handle MADM with 2TLNS. Finally, taking the CTQ evaluation of college badminton training as an example, the proposed method is explained. The main contributions of this study are summarized: the establishment of the 2TLNWPHM operator; (2) The 2TLNWPHM operator was developed to handle MADM with 2TLNS; (3) Through the empirical application of the CTQ evaluation of badminton training in universities, the proposed method is validated; (4) Some comparative studies have shown the rationality of the 2TLNWPHM operator. Show more

Keywords: Multi-attribute decision making (MADM), neutrosophic numbers, 2-tuple linguistic neutrosophic sets (2TLNSs), 2TLNWPHM operator, CTQ evaluation

DOI: 10.3233/JIFS-231731

Citation: Journal of Intelligent & Fuzzy Systems, vol. 45, no. 4, pp. 5939-5953, 2023