Deep Learning Techniques for Forest Parameter Retrieval and Accurate Tree Modeling from Remote Sensing Data

A special issue of Forests (ISSN 1999-4907). This special issue belongs to the section "Forest Inventory, Modeling and Remote Sensing".

Deadline for manuscript submissions: closed (20 June 2023) | Viewed by 48126

Special Issue Editors


E-Mail Website
Guest Editor
Research Institute of Forest Resource Information Techniques, Chinese Academy of Forestry, Beijing 100091, China
Interests: forest digital twin; virtual reality; artificial intelligence for forestry
Special Issues, Collections and Topics in MDPI journals

E-Mail Website
Guest Editor
College of Information Science and Technology, Nanjing Forestry University, Nanjing 210037, China
Interests: Internet of Things in forestry; multispectral remote sensing; intelligence systems
Special Issues, Collections and Topics in MDPI journals

Special Issue Information

Dear Colleagues,

Deep learning and digital twin technologies have the potential to retrieve forest parameters and simulate the forest life cycle, which is beneficial for forest silvicultural management and tree phenotypic trait characterization. While the foundations of these technologies have been laid through proof-of-concept studies, we are now in a position to make transformative advances, especially in the forest studies.

In this issue, we welcome all studies which deploy deep learning technologies and digital twin techniques in forestry applications. We intend to cover some aspects including various remote sensing data analysis, deep learning method development, key issue remedy and forest scenario rendering, along with affording inspiration and heuristic concepts in the multidisciplinary field for promoting the implementation of the technologies in forestry.

Specific topics include, but are not limited to:

  • Demonstration of deep learning methodologies for processing forest remote sensing data
  • Software approaches to forest visualization and modeling
  • Comparison between deep learning methods and other algorithms in forest survey
  • Forest scenario reconstruction from LiDAR data or other remote sensing data
  • Virtual forest management based on the virtual reality technology
  • Computer graphics or machine vision algorithms to enhance the fidelity of the reproduced forest environment
  • Prediction of the variations in forest growth properties based on deep learning frameworks from remote sensing data
  • Application of multi-remote sensing data in combination with deep learning frameworks for forestry carbon sink measurement
  • Processing terminal forest data acquired from various peripherals using deep learning approaches

Prof. Dr. Ting Yun
Dr. Eben Broadbent
Prof. Dr. Huaiqing Zhang
Prof. Dr. Ling Jiang
Guest Editors

Manuscript Submission Information

Manuscripts should be submitted online at www.mdpi.com by registering and logging in to this website. Once you are registered, click here to go to the submission form. Manuscripts can be submitted until the deadline. All submissions that pass pre-check are peer-reviewed. Accepted papers will be published continuously in the journal (as soon as accepted) and will be listed together on the special issue website. Research articles, review articles as well as short communications are invited. For planned papers, a title and short abstract (about 100 words) can be sent to the Editorial Office for announcement on this website.

Submitted manuscripts should not have been published previously, nor be under consideration for publication elsewhere (except conference proceedings papers). All manuscripts are thoroughly refereed through a single-blind peer-review process. A guide for authors and other relevant information for submission of manuscripts is available on the Instructions for Authors page. Forests is an international peer-reviewed open access monthly journal published by MDPI.

Please visit the Instructions for Authors page before submitting a manuscript. The Article Processing Charge (APC) for publication in this open access journal is 2600 CHF (Swiss Francs). Submitted papers should be well formatted and use good English. Authors may use MDPI's English editing service prior to publication or during author revisions.

Keywords

  • remote sensing
  • forest phenotypic traits
  • tree modelling
  • forest scenario rendering
  • digital twin
  • deep learning
  • computer graphics
  • machine vision

Benefits of Publishing in a Special Issue

  • Ease of navigation: Grouping papers by topic helps scholars navigate broad scope journals more efficiently.
  • Greater discoverability: Special Issues support the reach and impact of scientific research. Articles in Special Issues are more discoverable and cited more frequently.
  • Expansion of research network: Special Issues facilitate connections among authors, fostering scientific collaborations.
  • External promotion: Articles in Special Issues are often promoted through the journal's social media, increasing their visibility.
  • e-Book format: Special Issues with more than 10 articles can be published as dedicated e-books, ensuring wide and rapid dissemination.

Further information on MDPI's Special Issue polices can be found here.

Published Papers (15 papers)

Order results
Result details
Select all
Export citation of selected articles as:

Research

23 pages, 10648 KiB  
Article
Defogging Learning Based on an Improved DeepLabV3+ Model for Accurate Foggy Forest Fire Segmentation
by Tao Liu, Wenjing Chen, Xufeng Lin, Yunjie Mu, Jiating Huang, Demin Gao and Jiang Xu
Forests 2023, 14(9), 1859; https://doi.org/10.3390/f14091859 - 13 Sep 2023
Cited by 1 | Viewed by 1350
Abstract
In recent years, the utilization of deep learning for forest fire detection has yielded favorable outcomes. Nevertheless, the accurate segmentation of forest fires in foggy surroundings with limited visibility remains a formidable obstacle. To overcome this challenge, a collaborative defogging learning framework, known [...] Read more.
In recent years, the utilization of deep learning for forest fire detection has yielded favorable outcomes. Nevertheless, the accurate segmentation of forest fires in foggy surroundings with limited visibility remains a formidable obstacle. To overcome this challenge, a collaborative defogging learning framework, known as Defog DeepLabV3+, predicated on an enhanced DeepLabV3+ model is presented. Improved learning and precise flame segmentation are accomplished by merging the defogging features produced by the defogging branch in the input image. Furthermore, dual fusion attention residual feature attention (DARA) is proposed to enhance the extraction of flame-related features. The FFLAD dataset was developed given the scarcity of specifically tailored datasets for flame recognition in foggy environments. The experimental findings attest to the efficacy of our model, with a Mean Precision Accuracy (mPA) of 94.26%, a mean recall (mRecall) of 94.04%, and a mean intersection over union (mIoU) of 89.51%. These results demonstrate improvements of 2.99%, 3.89%, and 5.22% respectively. The findings reveal that the suggested model exhibits exceptional accuracy in foggy conditions, surpassing other existing models across all evaluation metrics. Full article
Show Figures

Figure 1

17 pages, 5673 KiB  
Article
UAV Multispectral Imagery Predicts Dead Fuel Moisture Content
by Jian Xing, Chaoyong Wang, Ying Liu, Zibo Chao, Jiabo Guo, Haitao Wang and Xinfang Chang
Forests 2023, 14(9), 1724; https://doi.org/10.3390/f14091724 - 26 Aug 2023
Viewed by 1517
Abstract
Forest floor dead fuel moisture content (DFMC) is an important factor in the occurrence of forest fires, and predicting DFMC is important for accurate fire risk forecasting. Large areas of forest surface DFMC are difficult to predict via manual methods. In this paper, [...] Read more.
Forest floor dead fuel moisture content (DFMC) is an important factor in the occurrence of forest fires, and predicting DFMC is important for accurate fire risk forecasting. Large areas of forest surface DFMC are difficult to predict via manual methods. In this paper, we propose an unmanned aerial vehicle (UAV)-based forest surface DFMC prediction method, in which a UAV is equipped with a multispectral camera to collect multispectral images of dead combustible material on the forest surface over a large area, combined with a deep-learning algorithm to achieve the large-scale prediction of DFMC on the forest surface. From 9 March to 23 March 2023, 5945 multispectral images and 480 sets of dead combustible samples were collected from an urban forestry demonstration site in Harbin, China, using an M300 RTK UAV with an MS600Pro multispectral camera. The multispectral images were segmented by a K-means clustering algorithm to obtain multispectral images containing only dead combustibles on the ground surface. The segmented multispectral images were then trained with the actual moisture content measured by the weighing method through the ConvNeXt deep-learning model, with 3985 images as the training set, 504 images as the validation set, and 498 images as the test set. The results showed that the MAE and RMSE of the test set are 1.54% and 5.45%, respectively, and the accuracy is 92.26% with high precision, achieving the accurate prediction of DFMC over a large range. The proposed new method for predicting DFMC via UAV multispectral cameras is expected to solve the real-time large-range accurate prediction of the moisture content of dead combustible material on the forest surface during the spring fire-prevention period in northeast China, thus providing technical support for improving the accuracy of forest fire risk-level forecasting and forest fire spread trend prediction. Full article
Show Figures

Figure 1

21 pages, 17573 KiB  
Article
Extraction of Pine Wilt Disease Regions Using UAV RGB Imagery and Improved Mask R-CNN Models Fused with ConvNeXt
by Zhenyu Wu and Xiangtao Jiang
Forests 2023, 14(8), 1672; https://doi.org/10.3390/f14081672 - 18 Aug 2023
Cited by 8 | Viewed by 1596
Abstract
Pine wilt disease (PWD) is one of the most concerning diseases in forestry and poses a considerable threat to forests. Since the deep learning approach can interpret the raw images acquired by UAVs, it provides an effective means for forest health detection. However, [...] Read more.
Pine wilt disease (PWD) is one of the most concerning diseases in forestry and poses a considerable threat to forests. Since the deep learning approach can interpret the raw images acquired by UAVs, it provides an effective means for forest health detection. However, the fact that only PWD can be detected but not the degree of infection can be evaluated hinders forest management, so it is necessary to establish an effective method to accurately detect PWD and extract regions infected by PWD. Therefore, a Mask R-CNN-based PWD detection and extraction algorithm is proposed in this paper. Firstly, the extraction of image features is improved by using the advanced ConvNeXt network. Then, it is proposed to change the original multi-scale structure to PA-FPN and normalize it by using GN and WS methods, which effectively enhances the data exchange between the bottom and top layers under low Batch-size training. Finally, a branch is added to the Mask module to improve the ability to extract objects using fusion. In addition, a PWD region extraction module is proposed in this paper for evaluating the damage caused by PWD. The experimental results show that the improved method proposed in this paper can achieve 91.9% recognition precision, 90.2% mapping precision, and 89.3% recognition rate of the affected regions on the PWD dataset. It can effectively identify the distribution of diseased pine trees and calculate the damage proportion in a relatively accurate way to facilitate the management of forests. Full article
Show Figures

Figure 1

26 pages, 30869 KiB  
Article
Time Series Forest Fire Prediction Based on Improved Transformer
by Xinyu Miao, Jian Li, Yunjie Mu, Cheng He, Yunfei Ma, Jie Chen, Wentao Wei and Demin Gao
Forests 2023, 14(8), 1596; https://doi.org/10.3390/f14081596 - 7 Aug 2023
Cited by 4 | Viewed by 3155
Abstract
Forest fires, severe natural disasters causing substantial damage, necessitate accurate predictive modeling to guide preventative measures effectively. This study introduces an enhanced window-based Transformer time series forecasting model aimed at improving the precision of forest fire predictions. Leveraging time series data from 2020 [...] Read more.
Forest fires, severe natural disasters causing substantial damage, necessitate accurate predictive modeling to guide preventative measures effectively. This study introduces an enhanced window-based Transformer time series forecasting model aimed at improving the precision of forest fire predictions. Leveraging time series data from 2020 to 2021 in Chongli, a myriad of forest fire influencing factors were ascertained using remote sensing satellite and GIS technologies, with their interrelationships estimated through a multicollinearity test. Given the intricate nature of real-world forest fire prediction tasks, we propose a novel window-based Transformer architecture complemented by a dual time series input strategy premised on 13 influential factors. Subsequently, time series data were incorporated into the model to generate a forest fire risk prediction map in Chongli District. The model’s effectiveness was then evaluated using various metrics, including accuracy (ACC), root mean square error (RMSE), and mean absolute error (MAE), and compared with traditional deep learning methods. Our model demonstrated superior predictive performance (ACC = 91.56%, RMSE = 0.37, MAE = 0.05), harnessing spatial background information efficiently and effectively utilizing the periodicity of forest fire factors. Consequently, the study proves this method to be a novel and potent approach for time series fire prediction. Full article
Show Figures

Figure 1

20 pages, 5627 KiB  
Article
Sweetgum Leaf Spot Image Segmentation and Grading Detection Based on an Improved DeeplabV3+ Network
by Peng Wu, Maodong Cai, Xiaomei Yi, Guoying Wang, Lufeng Mo, Musenge Chola and Chilekwa Kapapa
Forests 2023, 14(8), 1547; https://doi.org/10.3390/f14081547 - 28 Jul 2023
Cited by 2 | Viewed by 1399
Abstract
Leaf spot disease and brown spot disease are common diseases affecting maple leaves. Accurate and efficient detection of these diseases is crucial for maintaining the photosynthetic efficiency and growth quality of maple leaves. However, existing segmentation methods for plant diseases often fail to [...] Read more.
Leaf spot disease and brown spot disease are common diseases affecting maple leaves. Accurate and efficient detection of these diseases is crucial for maintaining the photosynthetic efficiency and growth quality of maple leaves. However, existing segmentation methods for plant diseases often fail to accurately and rapidly detect disease areas on plant leaves. This paper presents a novel solution to accurately and efficiently detect common diseases in maple leaves. We propose a deep learning approach based on an enhanced version of DeepLabV3+ specifically designed for detecting common diseases in maple leaves. To construct the maple leaf spot dataset, we employed image annotation and data enhancement techniques. Our method incorporates the CBAM-FF module to fuse gradual features and deep features, enhancing the detection performance. Furthermore, we leverage the SANet attention mechanism to improve the feature extraction capabilities of the MobileNetV2 backbone network for spot features. The utilization of the focal loss function further enhances the detection accuracy of the affected areas. Experimental results demonstrate the effectiveness of our improved algorithm, achieving a mean intersection over union (MIoU) of 90.23% and a mean pixel accuracy (MPA) of 94.75%. Notably, our method outperforms traditional semantic segmentation methods commonly used for plant diseases, such as DeeplabV3+, Unet, Segnet, and others. The proposed approach significantly enhances the segmentation performance for detecting diseased spots on Liquidambar formosana leaves. Additionally, based on pixel statistics, the segmented lesion image is graded for accurate detection. Full article
Show Figures

Figure 1

16 pages, 5816 KiB  
Article
Individual Tree AGB Estimation of Malania oleifera Based on UAV-RGB Imagery and Mask R-CNN
by Maojia Gong, Weili Kou, Ning Lu, Yue Chen, Yongke Sun, Hongyan Lai, Bangqian Chen, Juan Wang and Chao Li
Forests 2023, 14(7), 1493; https://doi.org/10.3390/f14071493 - 21 Jul 2023
Cited by 1 | Viewed by 1886
Abstract
Forest aboveground biomass (AGB) is an important research topic in the field of forestry, with implications for carbon cycles and carbon sinks. Malania oleifera Chun et S. K. Lee (M. oleifera) is a valuable plant species that is listed on the [...] Read more.
Forest aboveground biomass (AGB) is an important research topic in the field of forestry, with implications for carbon cycles and carbon sinks. Malania oleifera Chun et S. K. Lee (M. oleifera) is a valuable plant species that is listed on the National Second-Class Protected Plant checklist and has received global attention for its conservation and resource utilization. To obtain accurate AGB of individual M. oleifera trees in a fast, low-finance-cost and low-labor-cost way, this study first attempted to estimate individual M. oleifera tree AGB by combining the centimeter-level resolution RGB imagery derived from unmanned aerial vehicles (UAVs) and the deep learning model of Mask R-CNN. Firstly, canopy area (CA) was obtained from the 3.5 cm high-resolution UAV-RGB imagery using the Mask R-CNN; secondly, to establish an allometric growth model between the diameter at breast height (DBH) and CA, the correlation analysis of both was conducted; thirdly, the AGB estimation method of individual M. oleifera trees was presented based on an empirical equation. The study showed that: (1) The deep learning model of Mask R-CNN achieved an average segmentation accuracy of 90% in the mixed forests to the extraction of the canopy of M. oleifera trees from UAV-RGB imagery. (2) The correlation between the extracted CA and field-measured DBH reached an R2 of 0.755 (n = 96). (3) The t-test method was used to verify the predicted and observed values of the CA-DBH model presented in this study, and the difference in deviation was not significant (p > 0.05). (4) AGB of individual M. oleifera was estimated for the first time. This study provides a reference method for the estimation of individual tree AGB of M. oleifera based on centimeter-level resolution UAV-RGB images and the Mask R-CNN deep learning. Full article
Show Figures

Figure 1

18 pages, 5292 KiB  
Article
Forest Fire Prediction Based on Long- and Short-Term Time-Series Network
by Xufeng Lin, Zhongyuan Li, Wenjing Chen, Xueying Sun and Demin Gao
Forests 2023, 14(4), 778; https://doi.org/10.3390/f14040778 - 10 Apr 2023
Cited by 19 | Viewed by 8718
Abstract
Modeling and prediction of forest fire occurrence play a key role in guiding forest fire prevention. From the perspective of the whole world, forest fires are a natural disaster with a great degree of hazard, and many countries have taken mountain fire prediction [...] Read more.
Modeling and prediction of forest fire occurrence play a key role in guiding forest fire prevention. From the perspective of the whole world, forest fires are a natural disaster with a great degree of hazard, and many countries have taken mountain fire prediction as an important measure for fire prevention and control, and have conducted corresponding research. In this study, a forest fire prediction model based on LSTNet is proposed to improve the accuracy of forest fire forecasts. The factors that influence forest fires are obtained through remote sensing satellites and GIS, and their correlation is estimated using Pearson correlation analysis and testing for multicollinearity. To account for the spatial aggregation of forest fires, the data set was constructed using oversampling methods and proportional stratified sampling, and the LSTNet forest fire prediction model was established based on eight influential factors. Finally, the predicted data were incorporated into the model and the predicted risk map of forest fires in Chongli, China was drawn. This paper uses metrics such as RMSE to compare with traditional machine learning methods, and the results show that the LSTNet model proposed in this paper has high accuracy (ACC 0.941). This study illustrates that the model can effectively use spatial background information and the periodicity of forest fire factors, and is a novel method for spatial prediction of forest fire susceptibility. Full article
Show Figures

Figure 1

18 pages, 10972 KiB  
Article
Framework of Virtual Plantation Forest Modeling and Data Analysis for Digital Twin
by Wanlu Li, Meng Yang, Benye Xi and Qingqing Huang
Forests 2023, 14(4), 683; https://doi.org/10.3390/f14040683 - 26 Mar 2023
Cited by 4 | Viewed by 2733
Abstract
Plantation forests, cultivated through artificial seeding and planting methods, are of great significance to human society. However, most experimental sites for these forests are located in remote areas. Therefore, in-depth studies on remote forest management and off-site experiments can better meet the experimental [...] Read more.
Plantation forests, cultivated through artificial seeding and planting methods, are of great significance to human society. However, most experimental sites for these forests are located in remote areas. Therefore, in-depth studies on remote forest management and off-site experiments can better meet the experimental and management needs of researchers. Based on an experimental plantation forest of Triploid Populus Tomentosa, this paper proposes a digital twin architecture for a virtual poplar plantation forest system. The framework includes the modeling of virtual plantation and data analysis. Regarding this system architecture, this paper theoretically analyzes the three main entities of the physical world, digital world, and researchers contained in it, as well as their interaction mechanisms. For virtual plantation modeling, a tree modeling method based on LiDAR point cloud data was adopted. The transitional particle flow method was proposed to combine with AdTree method for tree construction, followed by integration with other models and optimization. For plantation data analysis, a database based on forest monitoring data was established. Tree growth equations were derived by fitting the tree diameter at breast height data, which were then used to predict and simulate trends in diameter-related data that are difficult to measure. The experimental result shows that a preliminary digital twin-oriented poplar plantation system can be constructed based on the proposed framework. The system consists of 2160 trees and simulations of 10 types of monitored or predicted data, which provides a new practical basis for the application of digital twin technology in the forestry field. The optimized tree model consumes over 67% less memory, while the R2 of the tree growth equation with more than 100 data items could reach more than 87%, which greatly improves the performance and accuracy of the system. Thus, utilizing forestry information networking and digitization to support plantation forest experimentation and management contributes to advancing the digital transformation of forestry and the realization of a smart management model for forests. Full article
Show Figures

Figure 1

18 pages, 8636 KiB  
Article
Identification of Pine Wilt Disease Infected Wood Using UAV RGB Imagery and Improved YOLOv5 Models Integrated with Attention Mechanisms
by Peng Zhang, Zhichao Wang, Yuan Rao, Jun Zheng, Ning Zhang, Degao Wang, Jianqiao Zhu, Yifan Fang and Xiang Gao
Forests 2023, 14(3), 588; https://doi.org/10.3390/f14030588 - 16 Mar 2023
Cited by 5 | Viewed by 1950
Abstract
Pine wilt disease (PWD) is a great danger, due to two aspects: no effective cure and fast dissemination. One key to the prevention and treatment of pine wilt disease is the early detection of infected wood. Subsequently, appropriate treatment can be applied to [...] Read more.
Pine wilt disease (PWD) is a great danger, due to two aspects: no effective cure and fast dissemination. One key to the prevention and treatment of pine wilt disease is the early detection of infected wood. Subsequently, appropriate treatment can be applied to limit the further spread of pine wilt disease. In this work, a UAV (Unmanned Aerial Vehicle) with a RGB (Red, Green, Blue) camera was employed as it provided high-quality images of pine trees in a timely manner. Seven flights were performed above seven sample plots in northwestern Beijing, China. Then, raw images captured by the UAV were further pre-processed, classified, annotated, and formed the research datasets. In the formal analysis, improved YOLOv5 frameworks that integrated four attention mechanism modules, i.e., SE (Squeeze-and-Excitation), CA (Coordinate Attention), ECA (Efficient Channel Attention), and CBAM (Convolutional Block Attention Module), were developed. Each of them had been shown to improve the overall identification rate of infected trees at different ranges. The CA module was found to have the best performance, with an accuracy of 92.6%, a 3.3% improvement over the original YOLOv5s model. Meanwhile, the recognition speed was improved by 20 frames/second compared to the original YOLOv5s model. The comprehensive performance could well support the need for rapid detection of pine wilt disease. The overall framework proposed by this work shows a fast response to the spread of PWD. In addition, it requires a small amount of financial resources, which determines the duplication of this method for forestry operators. Full article
Show Figures

Figure 1

17 pages, 7142 KiB  
Article
Classification of Individual Tree Species Using UAV LiDAR Based on Transformer
by Peng Sun, Xuguang Yuan and Dan Li
Forests 2023, 14(3), 484; https://doi.org/10.3390/f14030484 - 28 Feb 2023
Cited by 8 | Viewed by 3296
Abstract
Tree species surveys are crucial in forest resource management and can provide references for forest protection policymakers. Traditional tree species surveys in the field are labor-intensive and time-consuming. In contrast, airborne LiDAR technology is highly capable of penetrating forest vegetation; it can be [...] Read more.
Tree species surveys are crucial in forest resource management and can provide references for forest protection policymakers. Traditional tree species surveys in the field are labor-intensive and time-consuming. In contrast, airborne LiDAR technology is highly capable of penetrating forest vegetation; it can be used to quickly obtain three-dimensional information regarding vegetation over large areas with a high level of precision, and it is widely used in the field of forestry. At this stage, most studies related to individual tree species classification focus on traditional machine learning, which often requires the combination of external information such as hyperspectral cameras and has difficulty in selecting features manually. In our research, we directly processed the point cloud from a UAV LiDAR system without the need to voxelize or grid the point cloud. Considering that relationships between disorder points can be effectively extracted using Transformer, we explored the potential of a 3D deep learning algorithm based on Transformer in the field of individual tree species classification. We used the UAV LiDAR data obtained in the experimental forest farm of Northeast Forestry University as the research object, and first, the data were preprocessed by being denoised and ground filtered. We used an improved random walk algorithm for individual tree segmentation and made our own data sets. Six different 3D deep learning neural networks and random forest algorithms were trained and tested to classify the point clouds of three tree species. The results show that the overall classification accuracy of PCT based on Transformer reached up to 88.3%, the kappa coefficient reached up to 0.82, and the optimal point density was 4096, which was slightly higher than that of the other deep learning algorithms we analyzed. In contrast, the overall accuracy of the random forest algorithm was only 63.3%. These results show that compared with the commonly used machine learning algorithms and a few algorithms based on multi-layer perceptron, Transformer-based networks provide higher accuracy, which means they can provide a theoretical basis and technical support for future research in the field of forest resource supervision based on UAV remote sensing. Full article
Show Figures

Figure 1

18 pages, 3749 KiB  
Article
YOLO-Tea: A Tea Disease Detection Model Improved by YOLOv5
by Zhenyang Xue, Renjie Xu, Di Bai and Haifeng Lin
Forests 2023, 14(2), 415; https://doi.org/10.3390/f14020415 - 17 Feb 2023
Cited by 72 | Viewed by 8802
Abstract
Diseases and insect pests of tea leaves cause huge economic losses to the tea industry every year, so the accurate identification of them is significant. Convolutional neural networks (CNNs) can automatically extract features from images of tea leaves suffering from insect and disease [...] Read more.
Diseases and insect pests of tea leaves cause huge economic losses to the tea industry every year, so the accurate identification of them is significant. Convolutional neural networks (CNNs) can automatically extract features from images of tea leaves suffering from insect and disease infestation. However, photographs of tea tree leaves taken in a natural environment have problems such as leaf shading, illumination, and small-sized objects. Affected by these problems, traditional CNNs cannot have a satisfactory recognition performance. To address this challenge, we propose YOLO-Tea, an improved model based on You Only Look Once version 5 (YOLOv5). Firstly, we integrated self-attention and convolution (ACmix), and convolutional block attention module (CBAM) to YOLOv5 to allow our proposed model to better focus on tea tree leaf diseases and insect pests. Secondly, to enhance the feature extraction capability of our model, we replaced the spatial pyramid pooling fast (SPPF) module in the original YOLOv5 with the receptive field block (RFB) module. Finally, we reduced the resource consumption of our model by incorporating a global context network (GCNet). This is essential especially when the model operates on resource-constrained edge devices. When compared to YOLOv5s, our proposed YOLO-Tea improved by 0.3%–15.0% over all test data. YOLO-Tea’s AP0.5, APTLB, and APGMB outperformed Faster R-CNN and SSD by 5.5%, 1.8%, 7.0% and 7.7%, 7.8%, 5.2%. YOLO-Tea has shown its promising potential to be applied in real-world tree disease detection systems. Full article
Show Figures

Figure 1

18 pages, 16386 KiB  
Article
Generalization of U-Net Semantic Segmentation for Forest Change Detection in South Korea Using Airborne Imagery
by JongCheol Pyo, Kuk-jin Han, Yoonrang Cho, Doyeon Kim and Daeyong Jin
Forests 2022, 13(12), 2170; https://doi.org/10.3390/f13122170 - 17 Dec 2022
Cited by 14 | Viewed by 2863
Abstract
Forest change detection is essential to prevent the secondary damage occurring by landslides causing profound results to the environment, ecosystem, and human society. The remote sensing technique is a solid candidate for identifying the spatial distribution of the forest. Even though the acquiring [...] Read more.
Forest change detection is essential to prevent the secondary damage occurring by landslides causing profound results to the environment, ecosystem, and human society. The remote sensing technique is a solid candidate for identifying the spatial distribution of the forest. Even though the acquiring and processing of remote sensing images are costly and time- and labor-consuming, the development of open source data platforms relieved these burdens by providing free imagery. The open source images also accelerate the generation of algorithms with large datasets. Thus, this study evaluated the generalizability of forest change detection by using open source airborne images and the U-Net model. U-Net model is convolutional deep learning architecture to effectively extract the image features for semantic segmentation tasks. The airborne and tree annotation images of the capital area in South Korea were processed for building U-Net input, while the pre-trained U-Net structure was adopted and fine-tuned for model training. The U-Net model provided robust results of the segmentation that classified forest and non-forest regions, having pixel accuracies, F1 score, and intersection of union (IoU) of 0.99, 0.97, and 0.95, respectively. The optimal epoch and excluded ambiguous label contributed to maintaining virtuous segmentation of the forest region. In addition, this model could correct the false label images because of showing exact classification results when the training labels were incorrect. After that, by using the open map service, the well-trained U-Net model classified forest change regions of Chungcheong from 2009 to 2016, Gangwon from 2010 to 2019, Jeolla from 2008 to 2013, Gyeongsang from 2017 to 2019, and Jeju Island from 2008 to 2013. That is, the U-Net was capable of forest change detection in various regions of South Korea at different times, despite the training on the model with only the images of the capital area. Overall, this study demonstrated the generalizability of a deep learning model for accurate forest change detection. Full article
Show Figures

Figure 1

18 pages, 7182 KiB  
Article
Simulation on Different Patterns of Mobile Laser Scanning with Extended Application on Solar Beam Illumination for Forest Plot
by Kang Jiang, Liang Chen, Xiangjun Wang, Feng An, Huaiqing Zhang and Ting Yun
Forests 2022, 13(12), 2139; https://doi.org/10.3390/f13122139 - 13 Dec 2022
Cited by 10 | Viewed by 1963
Abstract
Light detection and ranging (LiDAR) technology has become a mainstream tool for forest surveys, significantly contributing to the improved accuracy of forest inventories. However, the accuracy of the scanned data and tree properties derived using LiDAR technology may differ depending on the occlusion [...] Read more.
Light detection and ranging (LiDAR) technology has become a mainstream tool for forest surveys, significantly contributing to the improved accuracy of forest inventories. However, the accuracy of the scanned data and tree properties derived using LiDAR technology may differ depending on the occlusion effect, scanning configurations, various scanning patterns, and vegetative characteristics of forest plots. Hence, this paper presents a computer simulation program to build a digital forest plot composed of many tree models constructed based on in situ measurement information and two mobile scanning patterns, i.e., airborne laser scanning (ALS) and ground-based mobile laser scanning (MLS). Through the adjustment of scanning parameters and the velocity of vehicle loading LiDAR sensors, the points scanned using two scanning patterns were compared with the original sampling points, derived from the constructed digital forest plots. The results show that only 2% of sampling points were collected by LiDAR sensors with the fastest vehicle speed (10 m/s) and coarsest scanning angular resolution (horizontal angular resolution 0.16° and vertical angular resolution 1.33°), and approximately 50% of sampling points were collected by LiDAR sensors with slow vehicle velocity (1.25 m/s) and a finer scanning angular resolution (horizontal angular resolution 0.08° and vertical angular resolution 0.33°). Meanwhile, the potential extended application of the proposed computer simulation program as a light model of forest plots was discussed to underpin the creation of the forest digital twin. Three main conclusions are drawn: (1) the collected points from airborne laser scanning (ALS) are higher than those collected from ground-based mobile laser scanning (MLS); (2) reducing the vehicle velocity is more efficient at improving the high density of the point cloud data than by increasing the scanning angular resolution; (3) the lateral extension of crown area increasing the light beams’ receptor area and the clumped leaf dispersion augmenting the light penetration with vertical elongation are the two paramount factors influencing the light transmittance of tree crowns. Full article
Show Figures

Figure 1

18 pages, 12374 KiB  
Article
Tropical Forest Disturbance Monitoring Based on Multi-Source Time Series Satellite Images and the LandTrendr Algorithm
by Xiong Yin, Weili Kou, Ting Yun, Xiaowei Gu, Hongyan Lai, Yue Chen, Zhixiang Wu and Bangqian Chen
Forests 2022, 13(12), 2038; https://doi.org/10.3390/f13122038 - 30 Nov 2022
Cited by 3 | Viewed by 2621
Abstract
Monitoring disturbances in tropical forests is important for assessing disturbance-related greenhouse gas emissions and the ability of forests to sequester carbon, and for formulating strategies for sustainable forest management. Thanks to a long-term observation history, large spatial coverage, and support from powerful cloud [...] Read more.
Monitoring disturbances in tropical forests is important for assessing disturbance-related greenhouse gas emissions and the ability of forests to sequester carbon, and for formulating strategies for sustainable forest management. Thanks to a long-term observation history, large spatial coverage, and support from powerful cloud platforms such as Google Earth Engine (GEE), remote sensing is increasingly used to detect forest disturbances. In this study, three types of forest disturbances (abrupt, gradual, and multiple) were identified since the late 1980s on Hainan Island, the largest tropical island in China, using an improved LandTrendr algorithm and a dense time series of Landsat and Sentinel-2 satellite images on the GEE cloud platform. Results show that: (1) the algorithm identified forest disturbances with high accuracy, with the R2 for abrupt and gradual disturbance detection reaching 0.92 and 0.83, respectively; (2) the total area in which forest disturbances occurred on Hainan Island over the past 30 years accounted for 10.84% (2.33 × 105 hm2 in total area, at 0.35% per year) of the total forest area in 2020 and peaked around 2005; (3) the areas of abrupt, gradual, and multiple disturbances were 1.21 × 105 hm2, 9.96 × 104 hm2, and 1.25 × 104 hm2, accounting for 51.93%, 42.75%, and 5.32% of the total disturbed area, respectively; and (4) most forest disturbance occurred in low-lying (<600 m elevation accounts for 97.42%) and gentle (<25° slope accounts for 94.42%) regions, and were mainly caused by the rapid expansion of rubber, eucalyptus, and tropical fruit plantations and natural disasters such as typhoons and droughts. The resulting algorithm and data products provide effective support for assessments of such things as tropical forest productivity and carbon storage on Hainan Island. Full article
Show Figures

Figure 1

23 pages, 6043 KiB  
Article
Semi-Supervised Learning for Forest Fire Segmentation Using UAV Imagery
by Junling Wang, Xijian Fan, Xubing Yang, Tardi Tjahjadi and Yupeng Wang
Forests 2022, 13(10), 1573; https://doi.org/10.3390/f13101573 - 26 Sep 2022
Cited by 10 | Viewed by 2174
Abstract
Unmanned aerial vehicles (UAVs) are an efficient tool for monitoring forest fire due to its advantages, e.g., cost-saving, lightweight, flexible, etc. Semantic segmentation can provide a model aircraft to rapidly and accurately determine the location of a forest fire. However, training a semantic [...] Read more.
Unmanned aerial vehicles (UAVs) are an efficient tool for monitoring forest fire due to its advantages, e.g., cost-saving, lightweight, flexible, etc. Semantic segmentation can provide a model aircraft to rapidly and accurately determine the location of a forest fire. However, training a semantic segmentation model requires a large number of labeled images, which is labor-intensive and time-consuming to generate. To address the lack of labeled images, we propose, in this paper, a semi-supervised learning-based segmentation network, SemiFSNet. By taking into account the unique characteristics of UAV-acquired imagery of forest fire, the proposed method first uses occlusion-aware data augmentation for labeled data to increase the robustness of the trained model. In SemiFSNet, a dynamic encoder network replaces the ordinary convolution with dynamic convolution, thus enabling the learned feature to better represent the fire feature with varying size and shape. To mitigate the impact of complex scene background, we also propose a feature refinement module by integrating an attention mechanism to highlight the salient feature information, thus improving the performance of the segmentation network. Additionally, consistency regularization is introduced to exploit the rich information that unlabeled data contain, thus aiding the semi-supervised learning. To validate the effectiveness of the proposed method, extensive experiments were conducted on the Flame dataset and Corsican dataset. The experimental results show that the proposed model outperforms state-of-the-art methods and is competitive to its fully supervised learning counterpart. Full article
Show Figures

Figure 1

Back to TopTop