Liu Pingbo

Dr. Liu Pingbo


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
25 October 2024 in Remote Sensing
Verified
Journal Article
Open access
PDF file
LPLiu Pingbo
Liu Pingbo
GZGui Zhang
Gui Zhang
Forest fires can occur suddenly and have significant environmental, economic, and social consequences. The timely and accurate evaluation and prediction of their progression, particularly the spread speed in difficult-to-access areas, are essential for emergency management departments to proactively implement prevention strategies and extinguish fires using scientific methods. This paper provides an analysis of models for predicting forest fire spread in China and globally. Incorporating remote sensing (RS) technology and forest fire science as the theoretical foundation, and utilizing the Wang Zhengfei forest fire spread model (1983), which is noted for its broad adaptability in China as the technical framework, this study constructs a forest fire spread model based on remote sensing interpretation. The model improves the existing model by adding elevation an factor and optimizes the method for acquiring certain parameters. By considering regional landforms (ridge lines, valley lines, and slopes) and vegetation coverage, this paper establishes three-dimensional visual interpretation markers for identifying hotspots; the orientation of the hotspots can be identified to simulate the spread of the fire uphill, downhill, in the direction of the wind, left-level slope, and right-level slope. Then, the data of Sentinel-2 and DEM were used to invert the fuel humidity and slope of pixels in the fire line areas. The statistical inversion data from pixels, which replaced fixed-point values in traditional models, were utilized for predicting forest fire spread speed. In this paper, the model was applied to the case of a forest fire in Mianning County, Sichuan Province, China, and verified using high-time-resolution Himawari-8 data, Gaofen-4 data, and historical data. The results demonstrate that the direction and maximum speed of fire spread for the fire lines in Baifen Mountai, Jiaguer Villageand, Muchanggou, Xujiabaozi, and Zhaizigou are uphill, 16.5 m/min; wind direction, 17.32 m/min; wind direction, 1.59 m/min; and wind direction, 5.67 m/min. The differences are mainly due to the locations of the fire lines, moisture content of combustibles, and maximum slopes being different. Across the entire fire line area, the average rate of increase in the area of open flames within one hour was 3.257 hm2/10 min (square hectares per 10 min), closely matching the average increase rate (3.297 hm2/10 min) monitored by the Himawari-8 satellite in 10 min intervals. In contrast, conventional fixed-point fire spread models predicted an average rate of increase of 3.5637 hm2/10 min, which shows a larger discrepancy compared to the Himawari-8 satellite monitoring results. Moreover, when compared to the fire spot monitoring results from the Gaofen-4 satellite taken 54 min after the initial location of the fire line, the predictions from the RS-enabled fire spread model, which integrates remote sensing interpretations, closely matched the actual observed fire boundaries. Although the predictions from the RS-enabled fire spread model and the traditional model both align with historical data in terms of the overall fire development trends, the RS-enabled model exhibits higher reliability and can provide more accurate information for forest fire emergency departments, enabling effective pre-emptive measures and scientific firefighting strategies.

12 January 2026 in Remote Sensing
Verified
Journal Article
Open access
PDF file
LPLiu Pingbo
Liu Pingbo
GZGui Zhang
Gui Zhang
JLJianzhong Li
Jianzhong Li
Forest resources are among the most important ecosystems on the earth. The semantic segmentation and accurate positioning of ground objects in forest remote sensing (RS) imagery are crucial to the emergency treatment of forest natural disasters, especially forest fires. Currently, most existing methods for image semantic segmentation are built upon convolutional neural networks (CNNs). Nevertheless, these techniques face difficulties in directly accessing global contextual information and accurately detecting geometric transformations within the image’s target regions. This limitation stems from the inherent locality of convolution operations, which are restricted to processing data structured in Euclidean space and confined to square-shaped regions. Inspired by the graph convolution network (GCN) with robust capabilities in processing irregular and complex targets, as well as Swin Transformers renowned for exceptional global context modeling, we present a hybrid semantic segmentation framework for forest RS imagery termed GSwin–Unet. This framework embeds the GCN model into Swin–Unet architecture to address the issue of low semantic segmentation accuracy of RS imagery in forest scenarios, which is caused by the complex texture features, diverse shapes, and unclear boundaries of land objects. GSwin–Unet features a parallel dual-encoder architecture of GCN and Swin Transformer. First, we integrate the Zero-DCE (Zero-Reference Deep Curve Estimation) algorithm into GSwin–Unet to enhance forest RS image feature representation. Second, a feature aggregation module (FAM) is proposed to bridge the dual encoders by fusing GCN-derived local aggregated features with Swin Transformer-extracted features. Our study demonstrates that, compared with the baseline models TransUnet, Swin–Unet, Unet, and DeepLab V3+, the GSwin–Unet achieves improvements of 7.07%, 5.12%, 8.94%, and 2.69% in the mean Intersection over Union (MIoU) and 3.19%, 1.72%, 4.3%, and 3.69% in the average F1 score (Ave.F1), respectively, on the RGB forest RS dataset. On the NIRGB forest RS dataset, the improvements in MIoU are 5.75%, 3.38%, 6.79%, and 2.44%, and the improvements in Ave.F1 are 4.02%, 2.38%, 4.72%, and 1.67%, respectively. Meanwhile, GSwin–Unet shows excellent adaptability on the selected GID dataset with high forest coverage, where the MIoU and Ave.F1 reach 72.92% and 84.3%, respectively.

20 August 2024 in Preprints.org
Verified
Preprint
Open access
PDF file
LPLiu Pingbo
Liu Pingbo
GZGui Zhang
Gui Zhang
Forest fires can occur suddenly and have significant environmental, economic, and social consequences. Timely and accurate evaluation and prediction of their progression, particularly the spread speed in difficult-to-access areas, are essential for emergency management departments to proactively implement prevention strategies and scientifically extinguish fires. This paper provides a comprehensive analysis of advanced technologies for predicting forest fire spread in China and globally. Incorporating remote sensing (RS) technology and forest fire science as the theoretical foundation, and utilizing the Wang Zhengfei forest fire spread model as the technical framework, this study constructs a forest fire spread model based on remote sensing interpretation. The model optimizes the method for acquiring parameters and enhances algorithm accuracy. By considering regional landforms (ridge lines, valley lines, slopes) and vegetation coverage, this paper establishes visual interpretation markers for identifying hotspots within the spread model. Utilizing statistical data from pixels within the fire line zone transforms the methodology for predicting forest fire spread speed, resulting in greater accuracy compared to conventional fixed-point predictions. Finally, the model was applied to the case of a forest fire in Mianning County, Sichuan Province, China and verified using high-time resolution mid-infrared image data and previous research findings. The results demonstrate that the predicted fire spread area from this RS-enabled model is consistent with both high-time resolution geostationary satellite data and previous research outcomes, indicating high reliability. This model can provide crucial fire spread information to relevant emergency departments, enabling effective preemptive measures and scientific firefighting strategies.

3 January 2025 in Remote Sensing
Verified
Journal Article
Open access
PDF file
LPLiu Pingbo
Liu Pingbo
GZGui Zhang
Gui Zhang
In the background of high frequency and intensity forest fires driven by future warming and a drying climate, early detection and effective control of fires are extremely important to reduce losses. Meteorological satellite imagery is commonly used for near-real-time forest fire monitoring, thanks to its high temporal resolution. To address the misjudgments and omissions caused by solely relying on changes in infrared band brightness values and a single image in forest fire early discrimination, this paper constructs the angle slope indexes ANIR, AMIR, AMNIR, ∆ANIR, and ∆AMIR based on the reflectance of the red band and near-infrared band, the brightness temperature of the mid-infrared and far-infrared band, the difference between the AMIR and ANIR, and the index difference between time-series images. These indexes integrate the strong inter-band correlations and the reflectance characteristics of visible and short-wave infrared bands to simultaneously monitor smoke and fuel biomass changes in forest fires. We also used the decomposed three-dimensional OTSU (maximum inter-class variance method) algorithm to calculate the segmentation threshold of the sub-regions constructed from the AMNIR data to address the different discrimination thresholds caused by different time and space backgrounds. In this paper, the Himawari-8 satellite imagery was used to detect forest fires based on the angle slope indices thresholds algorithm (ASITR), and the fusion of the decomposed three-dimensional OTSU and ASITR algorithm (FDOA). Results show that, compared with ASITR, the accuracy of FDOA decreased by 3.41% (0.88 vs. 0.85), the omission error decreased by 52.94% (0.17 vs. 0.08), and the overall evaluation increased by 3.53% (0.85 vs. 0.88). The ASITR has higher accuracy, and the fusion of decomposed three-dimensional OTSU and angle slope indexes can reduce forest fire omission error and improve the overall evaluation.

1 October 2024 in Preprints.org
Verified
Preprint
Open access
PDF file
LPLiu Pingbo
Liu Pingbo
GZGui Zhang
Gui Zhang
In the context of high frequency and intensity forest fires driven by future warming and drying climate, early detection and effective control of fires are extremely important to reduce losses. Satellite remote sensing, characterized by its broad observational scope and strong repeat observation capability, has become an important method for rapid monitoring of forest fires, replacing traditional means such as manual measurement and aerial photography. This paper addresses the misjudgments caused by solely relying on changes in infrared band brightness values and single-band separation in forest fire discrimination. It constructs the angle slope index ANIR based on the red band, near-infrared band and short-wave infrared band, Angle slope index AMIR based on the short-wave infrared band, mid-infrared band and far-infrared band, and angle slope index AMNIR based on the difference between the AMIR and ANIR, integrating the strong inter-band correlations and the reflectance characteristics of visible and short-wave infrared bands to simultaneously monitor smoke and fuel biomass changes in forest fires. To address the omissions caused by single-image fire discrimination algorithms, the paper constructs time-series Angle slope difference indices ∆ANIR and ∆AMIR. To address the different fire point discrimination thresholds caused by background difference in different time and space, the paper uses the decomposed three-dimensional OTSU(Maximum inter-class variance method) algorithm to calculate the segmentation threshold of the sub-regions constructed from the AMNIR data. The methods were validated using high temporal resolution meteorological satellite Himawari-8 and ground truth fire data from Hunan Province, China, from 2018-2019. The results show that the discrimination method based on the ASITR(Angle slope indices thresholds) method exhibits an accuracy of 0.88, a missed detection rate of 0.17, and an overall evaluation value of 0.85, which is 0.09 higher in accuracy and 0.03 higher in overall evaluation than the discrimination method based on infrared bands brightness values proposed by Feng et al. [1] (accuracy 0.79, missed detection rate 0.14, overall evaluation value 0.82); the method based on the FAMN_OTSU_ASITR (Fusion of Angle slope difference index (false) data, Decomposed three-dimensional OTSU adaptive threshold segmentation Algorithm and ASITR) exhibits an accuracy of 0.85, a missed detection rate of 0.08, and an overall evaluation value of 0.88, which is 0.09 lower in missed detection rate and 0.03 higher in overall evaluation as compared to the ASITR. The results demonstrate that the forest fire discrimination method based on Angle slope indices, Himawari-8 satellite data, and the decomposed OTSU adaptive threshold segmentation algorithm can improve the accuracy of forest fire discrimination and reduce errors in missed detections.

18 November 2025 in Preprints.org
Verified
Preprint
Open access
PDF file
LPLiu Pingbo
Liu Pingbo
GZGui Zhang
Gui Zhang
LJLi Jianzhong
Li Jianzhong
Forest resources are among the most important ecosystems on the earth. The semantic segmentation and accurate positioning of ground objects in forest remote sensing (RS) imagery are crucial to the emergency treatment of forest natural disasters, especially forest fires. Currently, most existing methods for image semantic segmentation are built upon convolutional neural network (CNN). Nevertheless, these techniques face difficulties in directly accessing global contextual information and accurately detecting geometric transformations within the image’s target regions. This limitation stems from the inherent locality of convolution operations, which are restricted to processing data structured in Euclidean space and confined to square-shaped regions.Inspired by the Graph Convolution Network (GCN) with robust capabilities in processing irregular and complex targets, as well as Swin Transformers renowned for exceptional global context modeling, we present an innovative semantic segmentation framework for forest remote sensing imagery termed GSwin-Unet. This framework embeds GCN model into Swin-Unet architecture, and for the first time apply the method of combining GCN and Transformer in the domain of forest RS imagery analysis. GSwin-Unet features an innovative parallel dual-encoder architecture of GCN and Swin transformer. First, we integrate the Zero-DCE (Zero-Reference Deep Curve Estimation) algorithm into GSwin-Unet to enhance forest RS image feature representation. Second, a feature aggregation module (FAM) is proposed to bridge the dual encoders by fusing GCN-derived local aggregated features with Swin transformer-extracted features. Our study demonstrates that the GSwin-Unet significantly improves performance on the Forest Remote Sensing Dataset and exhibits good adaptability on GID dataset.