Bentian Hao

Dr. Bentian Hao

orcid

Research Focus & Metrics

Academic Career Length

3Years (2024-Present)

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

Representative Publications

View All Publications
1 January 2024 in Journal of Industrial Textiles
Verified
Journal Article
Open access
BHBentian Hao
Bentian Hao
WXWeidong Xu
Weidong Xu
XYXin Yang
Xin Yang
FXFeifei Xiao
Feifei Xiao
HLHao Li
Hao Li
WHWei Huang
Wei Huang
Manual evaluation of military target camouflage often fails to fully capture human visual characteristics. This paper introduces a method for intelligent camouflage feature extraction and effect evaluation using deep neural networks. Firstly, military target and background datasets are constructed, with annotation methods optimized through perceptual experiments. A GANomaly model for background anomaly detection and a transfer learning feature model are developed. These models evaluate data similarity to background and target features to categorize camouflage into three levels. PCA dimensionality reduction and RFE screening identify final features. Two logistic regression models are then combined to create a grading model. In experiments, the GANomaly model was trained and optimized, achieving 86.8% accuracy in distinguishing level 1 from non-level 1 camouflage using 28-dimensional features. Among three transfer models, the MobileNetV2 model identified 33-dimensional features with 85.1% accuracy in distinguishing level 3 from non-level 3 camouflage. Combined, the models reached an 84.3% grading accuracy on the test set, with Micro and Macro average AUC values of 0.939 and 0.945, respectively. This performance surpasses that of BP network models and linear models based on manual features, as well as metrics like SSIM, UIQI, MSE, and PSNR. The results indicate that the models effectively align with human annotations and can reliably assess military target camouflage levels.

Verified
Journal Article
Open access
PDF file
XYXin Yang
Xin Yang
SYSu Yan
Su Yan
BHBentian Hao
Bentian Hao
WXWei-dong XU
Wei-dong XU
HYHaibao Yu
Haibao Yu
Disruptive color on animals’ bodies can reduce the risk of being caught. This study explores the camouflaging effect of disruptive color when applied to military targets. Disruptive and non-disruptive color patterns were placed on the target surface to form simulation materials. Then, the simulation target was set in woodland-, grassland-, and desert-type background images. The detectability of the target in the background was obtained by collecting eye movement indicators after the observer observed the background targets. The influence of background type (local and global), camouflage pattern type, and target viewing angle on the disruptive-color camouflage pattern was investigated. This study aims to design eye movement observation experiments to statistically analyze the indicators of first discovery time, discovery frequency, and first-scan amplitude in the target area. The experimental results show that the first discovery time of mixed disruptive-color targets in a forest background was significantly higher than that of non-mixed disruptive-color targets (t = 2.54, p = 0.039), and the click frequency was reduced by 15% (p < 0.05), indicating that mixed disruptive color has better camouflage effectiveness in complex backgrounds. In addition, the camouflage effect of mixed disruptive colors on large-scale targets (viewing angle ≥ 30°) is significantly improved (F = 10.113, p = 0.01), providing theoretical support for close-range reconnaissance camouflage design.

16 February 2026 in Journal of King Saud University Computer and Information Sciences
Verified
Journal Article
Open access
PDF file
SLShenghui Li
Shenghui Li
BHBentian Hao
Bentian Hao
Modern battlefield perception systems, which rely on multi-modal target detection technologies such as optical and infrared sensing, pose a serious threat to the survivability of large mobile assets. Existing adversarial camouflage research has predominantly focused on single-band (optical or infrared) reconnaissance, lacking a comprehensive solution capable of simultaneously deceiving both optical and infrared detectors while maintaining visual concealment. To address this gap, this study proposes a Multi-modal Adversarial Camouflage (MAC) patch framework. Firstly, the Scene-Adaptive Digital Camouflage Generation (SADCG) algorithm is designed to generate a digital camouflage base for tanks that exhibits high fusion with the background. Secondly, a joint loss optimization function is devised to train a YOLOv5-based patch generation model for producing optical adversarial patches. Simultaneously, a shape parameterization method utilizing multi-anchor representation and Catmull-Rom splines, along with a three-layer constrained boundary restriction system, is developed to optimize infrared shape adversarial patches. These optical and infrared patches are then integrated onto the camouflage base using Indium Tin Oxide (ITO) transparent thin films, forming a unified multi-modal stealth system. Experimental results demonstrate that the MAC framework can effectively reduce the mean Average Precision (mAP) of both optical and infrared detectors (including Faster R-CNN and YOLOv4) by over 75%, while significantly suppressing the response of infrared detectors. Furthermore, subjective evaluations confirm its excellent visual and infrared concealment properties. This research provides a verifiable solution and a technical pathway for achieving multi-spectral stealth for equipment in the physical world.

12 November 2026 in Algorithms
Verified
Journal Article
Open access
PDF file
HPHao Peng
Hao Peng
BHBentian Hao
Bentian Hao
WXWei-dong XU
Wei-dong XU
PZPin Zhang
Pin Zhang
LQLiu Qi
Liu Qi
LZLiyan Zhu
Liyan Zhu
Forecasting heterogeneous vehicle–environment temperatures requires models that accommodate differences in amplitude, persistence and meteorological response while keeping component-model proliferation manageable. This study presents a channel-adaptive component forecasting framework and directed reuse of frozen dominant-component predictors. RIME selects complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) parameters using decomposition quality. A common component predictor combines static feature weighting, dilated causal convolution, historical-window bidirectional long short-term memory encoding, temporal attention and an autoregressive highway (DConvFTA-BiLSTM-AR). A field campaign in Xuzhou, China, provides 14,217 consecutive one-minute records from 64 vehicle and environmental temperature channels. Fourteen channels spanning contrasting materials and thermal regimes were benchmarked using a chronological 70/20/10 split, train-fitted scaling and decomposition restricted to the train–validation prefix. Under observed future meteorological inputs, macro MAE was 0.264 °C and macro MSE was 0.430 °C2. The MAE was 13.7% lower than that of CEEMDAN-LSTM, the strongest neural baseline by macro MAE, in the evaluated interval. Directed reuse tests identify compatible source–recipient pairs while retaining recipient-specific predictors for non-dominant components and the residual. For 13 exposed-surface channels, a fixed-emissivity Planck transformation gives mean self-emitted 8–14 μm radiant-exitance MAE and channel RMSE of 0.593 and 1.280 W/m2. These results establish component forecasting and partial predictor reuse as useful tools for the evaluated heterogeneous thermal array and provide temperature inputs for downstream infrared analysis.