Zhong Dong

Dr. Zhong Dong


Research Focus & Metrics

Academic Career Length

8Years (2019-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
9 January 2020 in International Journal of Remote Sensing
Verified
Journal Article
Closed access
ZDZhong Dong
Zhong Dong
BLBaojun Lin
Baojun Lin
Deep convolutional neural networks (CNN) have been widely applied in various fields, especially in the field of object detection. Deep CNN-based models showed great advantages over many traditional methods, even so, there are still many specific problems in the application of certain scenarios. In very high resolution (VHR) remote-sensing image datasets, the uncertainty of the object direction angle causes big trouble to the learning of the detector. Although the pooling operation can slightly alleviate the deviation caused by small angle, the feature learning of the objects with larger angle rotation still relies mainly on the sufficiency of sample data or effective data augmentation, which means the insufficiency of the training instances may cause serious performance degradation of the detector. In this paper, we propose a multi-angle box-based rotation insensitive object detection structure (MRI-CNN), which is an extended exploration for typical region-based CNN methods. On the one hand, we defined a set of directionally rotated bounding boxes before learning, and restricted the classification scene in a small angular range by rotated RoI (Region of Interest) pooling. On the other hand, we proposed a more effective screening method of bounding boxes, enabling the detector to adapt to diverse ground truth annotation methods and learn more accurate object localization. We trained our detector with different datasets containing different amount of training data, and the test results showed that the method proposed in this paper performs better than some mainstream detection methods when limited training data are provided in VHR remote-sensing datasets.

30 January 2024 in Remote Sensing
Verified
Journal Article
Open access
PDF file
ZDZhong Dong
Zhong Dong
BLBaojun Lin
Baojun Lin
FXFang Xie
Fang Xie
In the realm of few-shot classification learning, the judicious application of data augmentation methods has a significantly positive impact on classification performance. In the context of few-shot classification tasks for remote sensing images, the augmentation of features and the efficient utilization of limited features are of paramount importance. To address the performance degradation caused by challenges such as high interclass overlap and large intraclass variance in remote sensing image features, we present a data augmentation-based classification optimization method for few-shot remote sensing image scene classification. First, we construct a distortion magnitude space using different types of features, and we perform distortion adjustments on the support set samples while introducing an optimal search for the distortion magnitude (ODS) method. Then, the augmented support set offers a wide array of feature distortions in terms of types and degrees, significantly enhancing the generalization of intrasample features. Subsequently, we devise a dual-path classification (DC) decision strategy, effectively leveraging the discriminative information provided by the postdistortion features to further reduce the likelihood of classification errors. Finally, we evaluate the proposed method using a widely used remote sensing dataset. Our experimental results demonstrate that our approach outperforms benchmark methods, achieving improved classification accuracy.

24 December 2019 in Remote Sensing Letters
Verified
Journal Article
Closed access
ZDZhong Dong
Zhong Dong
BLBaojun Lin
Baojun Lin
Object detection in very-high-resolution (VHR) remote sensing images is one of the important technical means in many fields. In recent years, conventional object detection methods have been completely replaced by Convolutional Neural Network (CNN)-based methods, which are more accurate and efficient. However, most current CNN-based methods applied in VHR image sets have certain defects: (1) Scale preference is common in the framework designs, and the representation ability of feature maps for large and small objects is quite different, so accuracy promotion can hardly be made comprehensively in the detection of different objects. (2) The scale difference of the objects leads to training difficulties. (3) Some high-precision methods require high hardware costs, and the overall frameworks lack practicality. To address such problems, we propose a new object detection method in this paper, namely Balanced Multi-Scale Fusion-based CNN (BMF-CNN). It is a redesigned two-stage detection framework according to the region-based object detection methods, which enabled the detection accuracy of both large and small objects to reach a high level. Through the evaluation in the open VHR remote sensing image sets, we found that BMF-CNN showed a better integrative performance than the current mainstream detection methods.

30 August 2023 in IEEE Access
Verified
Journal Article
Open access
ZDZhong Dong
Zhong Dong
XLXiaofei Li
Xiaofei Li
BLBaojun Lin
Baojun Lin
FXFang Xie
Fang Xie
Brain-inspired facial expression recognition is a promising but challenging research direction with significant potential for applications in human-computer interaction and intelligent image analysis. However, current mainstream FER models, which are single-pathway and global feature-based, still have a long way to go to match the performance of the human brain in FER. To narrow this gap, this paper proposes a novel multi-net fused FER framework that incorporates specific FER mechanisms from the human brain into deep learning. Different from most FER model designers, we transfer our main focus from obtaining high-quality global features through data augmentation and network optimization to referring the cognitive processes involved in different facial expression classes, classification preference, and other factors that contribute to the human brain’s FER mechanisms. The framework includes data preprocessing, activeness estimation, multi-net classification, and fusion recognition. We validate and test the proposed framework using general datasets, and the results show that it performs well in FER tasks. The works in this paper provide exploratory ideas and a basis for future research in this direction.

2 February 2024 in International Journal of Remote Sensing
Verified
Journal Article
Closed access
ZDZhong Dong
Zhong Dong
BLBaojun Lin
Baojun Lin
FXFang Xie
Fang Xie

14 December 2023 in IEEE Geoscience and Remote Sensing Letters
Verified
Journal Article
Closed access
ZDZhong Dong
Zhong Dong
BLBaojun Lin
Baojun Lin
FXFang Xie
Fang Xie
As a critical component of few-shot classification learning, feature augmentation methods have a direct bearing on classification performance. In the realm of few-shot classification tasks concerning remote sensing (RS) imagery, the instances of samples exhibiting relatively feeble features are commonplace. Employing conventional data augmentation approaches often falls short of achieving the desired classification outcomes. In response to such challenges, this letter introduces an optimization method for data augmentation in the context of few-shot RS image scene classification. We commence by analyzing the impact of distortion levels in data augmentation on classification performance. Then, we construct distortion-level spaces using different types of features and incorporate the distortion-level spaces into the model learning process using the support set, facilitating both model parameter optimization and dynamic optimization of distortion levels. The resulting optimized support data offer feature distortions of diverse types and magnitudes, thereby fortifying the feature robustness within the samples. We evaluate the proposed approach on widely used RS datasets, and the results demonstrate significant improvement in classification accuracy compared to the baseline. The introduction of this method offers a fresh perspective for data augmentation-based few-shot RS scene classification learning.