Camellia oleifera is an important woody oil crop in southern China, and accurate extraction of crown and height attributes is essential for plantation management and yield assessment. However, joint extraction of these attributes and task-specific adaptation of deep learning models remain insufficiently explored. In this study, we proposed a UAV-based two-stage framework based on an improved U-Net++ architecture for crown segmentation and tree height estimation in C. oleifera plantations. Specifically, U-Net++ with an EfficientNet-B0 encoder was used in the segmentation stage to improve multiscale crown boundary delineation, while a U-Net++ regression model with a ResNet50 encoder and an scSE attention module was developed in the height estimation stage to enhance canopy-related feature learning. Different combinations of RGB imagery, multispectral bands, and elevation data were evaluated to assess the contribution of multisource data fusion. The results showed that the Multiband-CHM combination achieved the best crown segmentation performance, with an F1 score of 0.8719 and an IoU of 0.7807, whereas the Multiband-DSM combination yielded the best tree height prediction accuracy, with an R2 of 0.7996 and an rRMSE of 27.06%. Furthermore, the framework demonstrated stable performance across regions with different planting densities and under large-area plantation conditions. These findings validate the framework for effective crown and height extraction, supporting precision management and yield assessment of large-scale C. oleifera plantations.

Detecting forest changes through remote sensing imagery is essential for effectively managing and monitoring forest resources. Algorithms for forest change detection leveraging deep learning have shown remarkable performance across diverse scenarios. However, these algorithms face challenges such as imbalanced forest change samples and high false detection rates. To address these issues, this article proposes a global grouped coordinate attention (GGCA) mechanism, which improves the model's detection accuracy by emphasizing key features while suppressing irrelevant changes during the feature reconstruction process. Based on this, a lightweight forest change detection model, TinyForestCD, is designed. Furthermore, a hybrid loss function integrating CE loss and Dice loss is employed alongside an online hard example mining (OHEM) strategy, aiming to accurately extract key features from highly imbalanced forest change data. Extensive experiments conducted on a custom-built forest change detection dataset demonstrate that the TinyForestCD model outperforms other comparative models across various performance metrics, achieving a performance metric of 80.65% in F1-score and 67.57% in IoU. Moreover, in the practical deployment in Hunan Province, the average F1-score and IoU for accuracy assessment of six sample plots are 81.06% and 68.25%, respectively, which are consistent with the test set results. In addition, in the deployment results of the Shaoguan area, the model also showed the best performance compared with other models, validating the model's robustness and applicability in large-scale forest change detection. The research indicates that the TinyForestCD model has the potential for large-scale practical deployment, significantly improving the efficiency and accuracy of forest change surveys.

Topographic radiation distortion, induced by terrain relief, is a significant challenge limiting the accuracy of quantitative analysis in mountainous remote sensing applica-tions. Although semi-empirical topographic correction models, such as the widely used sun-canopy-sensor with C-correction (SCS+C), are prevalent, they often exhibit significant overcorrection tendencies and spectral distortions in steep terrain, presenting unresolved challenges. This study proposes DeepSCS+C, a physics-regularized network for adaptive topographic correction in mountainous forests. By characterizing a spatially varying C-field from multiscale terrain features, this framework utilizes SCS+C outputs as weakly supervised proxy labels to guide optimization while effectively suppressing intrinsic over-correction noise. Subsequently, using Sentinel-2 imagery of mountainous forested areas, the performance of the DeepSCS+C (DSC) model was systematically validated, and multidimensional evaluation criteria confirmed its robust performance. Experiments demonstrated that the DSC framework mitigates the terrain shadow effect with an efficacy comparable to traditional models and leverages the nonlinear mapping capabilities of deep learning to restore the intrinsic spectral characteristics of surface features accurately. The model's effectiveness is evidenced by several key metrics: the correlation between reflectance and illumination conditions (ICs) was reduced to near zero ( R-2 < 0.1); the spectral variance in forested areas decreased by 47.99% in the near-infrared (NIR) band compared to the original imagery; and the radiance difference between sunlit and shaded slopes converged to a near zero level (<1.0 Wm(-2)sr(-1 )mu m(-1)) across all bands. The findings of this research present a promising and advanced solution for high-precision topographic correction.

Digital elevation models (DEMs) are recognized as fundamental to numerous geospatial applications, yet the spatial resolution of widely available DEMs is often insufficient for fine-scale analysis. Deep learning-based super-resolution (SR) has been developed as a powerful technique for enhancing DEM resolution. However, challenges remain in modeling complex, long-range topographic dependencies and in effectively fusing guidance information from auxiliary data, such as high-resolution (HR) optical imagery, which can result in the loss of fine details and the generation of geomorphologically inconsistent surfaces. In this study, an attention-transformer-guided terrain SR (ATTSR) is proposed for high-fidelity, image-guided DEM SR. The ATTSR framework is designed with a dual-branch Swin Transformer backbone to capture global spatial context, and a novel topographic-aware attention (TAA) module is introduced for the adaptive fusion of image and DEM features. Furthermore, a terrain gradient constraint is incorporated into a collaborative loss function to explicitly preserve topographic structure. The performance of ATTSR is evaluated on three diverse datasets for SR tasks from 30 to 10 m and from 90 to 30 m. Compared to state-of-the-art methods, the elevation root-mean-square error (RMSE) is reduced by 7%-36%, with corresponding improvements observed in slope and aspect fidelity. These results demonstrate that ATTSR provides a state-of-the-art solution for image-guided DEM SR, advancing the generation of accurate, large-scale HR terrain products.

The widespread cultivation of Camellia oleifera in subtropical regions has played a significant role in promoting economic development in southern China. In recent years, the cultivation area has continuously expanded, highlighting the importance of extracting Camellia oleifera plantations automatically and monitoring ecological changes in these regions. This study proposes a novel method based on deep learning, utilizing Sentinel-2 images for efficient assessing ecological quality of Camellia oleifera plantations. First, we undertook a comprehensive examination and assessment of different semantic segmentation models, utilizing a dataset of Camellia oleifera plantations that we compiled. The most effective segmentation model was then employed to automatically extract Camellia oleifera plantations in Shaoyang City. Finally, to gain a deeper understanding of the ecological status of Camellia oleifera plantations, the Remote Sensing Ecological Index (RSEI) was introduced to analyse ecological changes from 2018 to 2022. The research results indicate that when the U-Net++ model is applied to extract Camellia oleifera plantations from Sentinel-2 images, its performance surpasses other semantic segmentation methods. This method achieves notable metrics: accuracy of 0.80, recall of 0.79, F1-Score of 0.81, and MIoU of 0.88, demonstrating its effectiveness in identifying Camellia oleifera plantations. The research findings of Camellia oleifera plantations in Shaoyang City show a notable increase in the mean RSEI, with a rise from 0.166 in 2018 to 0.345 in 2022. Through the application of deep learning for automated Camellia oleifera plantation extraction and RSEI for ecological assessment, the study discovers a continuous expansion of Camellia oleifera plantations and a yearly increase in ecological quality.

Forest change detection is crucial in environmental protection and ecological monitoring, but this task faces many challenges. First, forests are significantly affected by seasonal factors, and seasonal defoliation of trees can lead to misclassification, making it difficult to accurately identify actual forest changes. Crop harvests are also prone to misclassification due to their spectral similarity to forest changes, further complicating the detection task. The high degree of irregularity in forest change patches increases the difficulty of accurately detecting forest change boundary. Traditional convolutional neural network (CNN) backbone networks, such as VGG and ResNet, faced with the above challenges, are prone to misclassification in forest change detection tasks due to their small receptive fields relatively. To alleviate this problem, we propose the ConvNeXt-Uphead, a forest change detection model which can reduce the misclassification by extending the receptive field to improved ability to capture more information around changing areas. The experimental result demonstrate that the ConvNeXt-Uphead model excels in accurately identifying forest changes. Visualization results further show that its boundary detection is more precise and this way can better capture subtle changes.

High-resolution land cover mapping is crucial in various disciplines but is often hindered by the lack of accurately matched labels. Our study introduces an innovative deep learning methodology for effective land cover mapping, independent of matched labels. The approach comprises three main components: (1) An advanced fully convolutional neural network, augmented with super-resolution features, to refine labels; (2) The application of an instance-batch normalization network (IBN), leveraging these enhanced labels from the source domain, to generate 2-m resolution land cover maps for test sites in the target domain; (3) Noise assessment tests to evaluate the impact of varying noise levels on the model’s mapping accuracy using external labels. The model achieved an overall accuracy of 83.40% in the target domain using endogenous super-resolution labels. In contrast, employing exogenous, high-precision labels from the National Land Cover Database in the source domain led to a notable accuracy increase of 2.55%, reaching 85.48%. This improvement highlights the model’s enhanced generalizability and performance during domain shifts, attributed significantly to the IBN layer. Our findings reveal that, despite the absence of native high-precision labels, the utilization of high-quality external labels can substantially benefit the development of precise land cover mapping, underscoring their potential in scenarios with unmatched labels.

A rapid and accurate yield estimation is of great significance to the management and sustainable development of Camellia Oleifera forests. Consequently, the simultaneous and accurate detection of key phenology features of Camellia Oleifera (buds, flowers, fruits) is crucial for precise yield estimation. This not only enables robotic harvesting but also allows for the prediction of peak flowering and fruit ripening periods to determine the optimal harvesting time. However, in recent studies, only Camellia Oleifera fruits have been marked and detected. Therefore, to enable rapid yield estimation, it is necessary to simultaneously detect the key phenology stages (buds, flowers, fruits) of Camellia Oleifera. In this study, we annotated, trained, and predicted Camellia Oleifera buds, flowers, and fruits collected via field cameras to observe their daily quantitative changes. Quantity change curves were generated to estimate crucial phenology stages. Phenology feature detection and transfer learning were performed using the YOLO v5 model, widely used YOLO v3 model, and CenterNet model with center point prediction, all trained on the same dataset. The best model for phenology feature detection was selected based on a comparison of average precision, with the YOLO v5 model achieving a higher mean Average Precision (mAP) value of 91.31 % compared to the CenterNet (85.43 %) and YOLO v3 (81.00 %) models. In YOLO v5, the AP values for bud, flower, and fruit detection were 82.80 %, 98.13 %, and 92.99 %, respectively, surpassing the CenterNet model by 3.97 %, 2.44 %, and 11.23 %, and the YOLO v3 model by 6.39 %, 17.13 %, and 11.67 %. The image size was adapted from 4000 x 3000 pixels to 512 x 512 pixels for model optimization. Additionally, data from the Seedling Center of Liuyang City collected at different years and times were utilized to showcase the generalizability and scalability of the selected models, resulting in mAP values of 86.14 %, 80.17 %, and 69.20 % for the three above-mentioned models respectively. The plotted phenology change curves unveiled that Camellia Oleifera undergoes four stages: fruit enlargement period, bud enlargement period, flowering period, and flower wilting period. The conclusion can be drawn that using field cameras and YOLOv5 can simultaneously achieve real-time detection of key phenology features (buds, flowers, and fruits) of Camellia Oleifera, in order to further record crucial phenology patterns (such as flowering peaks and fruit ripening periods). This study offers theoretical references and scientific evidence for monitoring changes in key phenology features of Camellia Oleifera.

Dynamic detection of forest change is the fundamental method of monitoring forest resources and an essential means of preserving the accuracy and timeliness of forest land resource data. This study focuses on a deep learning-based method for dynamic forest change detection using Sentinel-2 satellite data, especially within mountainous areas. First, the performance of various deep learning models (U-Net++, U-Net, LinkNet, DeepLabV3+, and STANet) and various loss functions (CrossEntropyLoss(CELoss), DiceLoss, FocalLoss, and their combinations) are compared on a self-made dataset. Next, the best model and loss function is used to predict the annual forest change in Hunan Province from 2017 to 2021, and the detection results are evaluated in 12 sample areas. Finally, forest changes are detected in Sentinel-2 images for each quarter of 2017–2021. In addition, a dynamic detection map of forest change in Hunan Province from 2017 to 2021 is drawn. The results reveal that the U-Net++ model and the CELoss performed the best on the self-made dataset, with a Precision of 0.795, a Recall of 0.748, and an F1-score of 0.771. The results of annual and quarterly forest change detection were consistent with the changes in the Sentinel-2 images with accurate boundaries. This result demonstrates the high practicality and generalizability of the method used in this paper. This paper achieves a rapid and accurate extraction of multi-temporal Sentinel-2 image forest change areas based on the U-Net++ model, which can be used as a benchmark for future large territorial areas monitoring and management of forest resources.

The mapping of forest pests and diseases using remote sensing presents a critical challenge in forest management and conservation. Forestry professionals have shown particular concern about the rapid decline of pine trees caused by pine wilt disease, conducting annual inspections to monitor its impact. However, achieving a balance between large-scale monitoring areas and the identification of individual trees poses a challenging issue. To tackle this issue, we employed a random sampling plan and utilized an unmanned aerial vehicle (UAV) to capture high-resolution images on two separate occasions. This methodology facilitated the comparison of temporal variations in tree health. We applied three object detection algorithms (YOLOv5, Faster-RCNN, CenterNet) to identify diseased trees in the images and introduced an enhanced vision of YOLOv5 for better performance. The improved YOLOv5 demonstrated higher precision of average precision (AP) (0.8806, 0.9004) compared to the other algorithms across two datasets. The improved YOLOv5 generated predictions that identified the topo-graphic regions most severely affected by pine wilt nematode infestation. Geographically, the highest disease incidence was observed at altitudes between 200 and 299 m, slopes ranging from 10 to 15 degrees, and locations facing a sunny slope direction. The primary source of infection is pre-existing diseased trees that are not properly controlled. Additionally, the density of infected trees contributes to the spread, with the most severe outbreaks occurring in densely clustered over large areas. By aggregating the cumulative sampling data, we derived an estimation of the total tree mortality in the two towns. This study introduces an effective monitoring approach for pine wilt disease at the district level by integrating a random sampling protocol, unmanned aerial vehicle (UAV) strategies, and a deep learning algorithm. Our approach can assist forest managers in timely and cost-effective detection and control of pine wilt disease.

Scattered graves are a significant source of forest fire risk and are important data for forest, emergency, and civil affairs departments to track. However, there is a lack of research on the rapid detection and automatic inventory of scattered graves. This study proposes a Mask RCNN recognition network based on the Detectron2 toolbox to detect and inventory scattered graves in Changsha County of Hunan Province, using high-resolution aerial imagery and field survey data. The distribution of scattered graves was mapped and analyzed using various spatial distribution indices and factors such as the nearest neighbor index, geographic concentration index, unbalanced index, and kernel density index. The results showed that the improved Mask RCNN network was effective at identifying scattered graves with average F1 scores of 90.33%, 92.61%, and 90.41% for different land cover, distribution densities, and research areas. Scattered graves were found to have a clustered-random distribution with a density of 25.13 graves/km~2, and were primarily influenced by topography and road factors. This study presents the first data on the distribution of fire hazard points in scattered graves and has potential applications for the automatic detection and timely warning of scattered graves nationwide.

内陆滩涂是湿地的重要组成部分,具有保护生物多样性、维护区域生态平衡等重要生态功能.内陆滩涂和水体时刻处于动态变化之中,需要及时、准确地掌握内陆滩涂的现状,才能实现其可持续管理.传统的滩涂提取研究依赖于训练样本、人工设定阈值等方法,而收集足够且准确的训练样本费时费力.本研究基于GEE(Google Earth Engine)云平台,以Sentinel-2为数据源,结合最大光谱指数合成法(MSIC)和最大类间方差法(Otsu),提取了2018—2021年东洞庭湖年际最大面积滩涂,得到的总分类精度为96.3％,Kappa系数为0.95.同时,对东洞庭湖滩涂面积和水位高度展开定量分析,得到的总分类精度为97.3％,Kappa系数为0.96.本研究证实了MSIC-Otsu算法可以高度完成内陆滩涂的提取工作,实现根据水位高度即可预测滩涂面积的功能.

Fine fuel load (FFL) is a crucial variable influencing the occurrence of wildfire. Accurate knowledge of the distribution of FFL in mountainous forests is essential for ongoing wildfire risk management and the stability of mountain ecosystems. Traditional methods of estimating forest fuel load typically involve ground surveys combined with remote sensing, which can be costly and inefficient. Therefore, low-cost, large-scale FFL estimation remains challenging. In this study, Sentinel-2A satellite imagery from the Changsha forest region was used as the data source. Firstly, different feature variables were constructed based on false-color (B843), true-color (B432), four-band (B8432) combinations, and the Normalized Difference Water Index (NDWI). Subsequently, a machine learning approach based on random convolution was employed to estimate FFL. This study also included accuracy assessments of the estimation results and the creation of FFL maps for the study area. The results showed that the FFL estimation based on the B8432 band combination achieved the highest accuracy, with RMSE and R2 values of 5.847 t·hm−2 and 0.656, respectively. FFL estimation results based on false-color imagery followed, with true-color imagery and NDWI index-based estimation results exhibiting lower accuracy. This study offers critical FFL insights using random convolution techniques applied to Sentinel-2A imagery, enhancing the ability to monitor and manage forest fuel conditions effectively, thereby facilitating more informed regional wildfire risk management strategies.

Using typical topographic correction models to compare the removal effectiveness of topographic effect on different vegetation indices under complex mountain conditions can provide a scientific basis for accurate assessment of vegetation. Taking Sentinel-2 images of Yanling County, Hunan Province as data source, C model, SCS+C model, and Teillet-regression model were used for topographic correction of the band-ratio vegetation indices（NDVI and SR） and non-band-ratio vegetation indices（MNDVI and RSR）, aiming to analyze the removal effectiveness of topographic effect on the four vegetation indices from five aspects: visual effects, correlation, slope, aspect, and vegetation coverage. The results showed that:（1） The band-ratio vegetation indices effectively suppressed the noise caused by terrain. When the slope was less than 15°, the topographic effects of vegetation indices were suppressed to different degrees. When the slope was greater than 15°, topographic correction was the most effective on MNDVI, followed by RSR, while NDVI and SR were prone to over-correction.（2） All the three topographic correction models could reduce the topographic effects of the vegetation indices in rugged terrain areas, especially when the slope was greater than 15°. The C model was the most effective, followed by the SCS+C model, and the Teillet model was the least effective.（3） MNDVI after topographic correction effectively restrained the influence of complex mountain topographic effects and improved the estimation accuracy of vegetation coverage. When slope was less than 25°, the Teillet model was the most effective. When slope was greater than 25°, the C model was the most effective.

Camellia oleifera is a vital economic crop of southern China. Accurate mapping and monitoring of Camellia oleifera plantations are essential for promoting sustainable operations within the Camellia oleifera industry. However, traditional remote sensing interpretation methods are no longer feasible for the large-scale extraction of plantation areas. This study proposes a novel deep learning-based method that utilizes GF-2 remote sensing imagery to achieve precise mapping and efficient monitoring of Camellia oleifera plantations. First, we conducted a comparative analysis of the performance of various semantic segmentation models using a self-compiled dataset of Camellia oleifera plantations. Subsequently, we proceeded to validate the prediction results obtained from the most effective deep-learning network model for Camellia oleifera plantations in Hengyang City. Finally, we incorporated DEM data to analyze the spatial distribution patterns. The findings indicate that the U-Net++ network model outperforms other semantic segmentation methods when applied to our self-generated dataset of Camellia oleifera plantations. It achieves a recall rate of 0.89, a precision rate of 0.92, and an mIOU of 0.83, demonstrating the effectiveness of the proposed method in identifying and monitoring Camellia oleifera plantations. By combining the predicted results with the data from DEM, we discovered that these plantations are typically situated at elevations ranging from 50 to 200 m, with slopes below 25°, and facing south or southeast. Moreover, a significant positive spatial correlation and clustering phenomenon are observed among the townships in Hengyang City. The method proposed in this study facilitates rapid and precise identification and monitoring of Camellia oleifera plantations, offering significant theoretical support and a scientific foundation for the management and ecological conservation of Camellia oleifera plantations.

[目的]开发基于图像的香花油茶品种识别技术,快速、准确识别香花油茶无性系品种.[方法]选择自然光照条件下生长的20个香花油茶无性系品种的油茶果作为研究对象,利用红米K30 Pro、华为P40、华为nova7、iPhone 12、魅族 16s多种品牌型号的智能手机对自然状态下油茶果的脐面、侧面进行图像采集,去除低质量图像,通过数据增强方法增加图像数量,采用深度学习网络GoogLeNet-V3 对 20 个品种的香花油茶果图像进行识别.同时对使用多个设备采集图像的品种,进行不同设备型号及数量、相同设备数量但不同比例等情况可能造成的影响进行探讨.[结果]构建了图像数量为16 832 张的香花油茶果品种数据集.使用GoogLeNet-V3模型能满足基于油茶果图像的香花油茶品种识别要求,其中识别准确率、召回率、宏观F1 值和微观F1 值分别为 89.13%、89.31%、89.22%和 94.29%.对多设备的研究结果表明数据量的增加能有效提高模型精度,且在保证单设备采集数据量的条件下,使用多种设备采集数据构建的模型具有更高的鲁棒性.以GoogLeNet-V3 模型为基础构建的移动端识别APP,具有PC端同等精度,可用于香花油茶果品种的识别.[结论]使用深度学习网络GoogLeNet-V3模型能够实现香花油茶果的品种识别.

Change detection is a crucial task in remote sensing that finds broad application in land resource planning, forest resource monitoring, natural disaster monitoring, and evaluation. In this paper, we propose a change detection model for cross-domain recognition, which we call CrossCDNet. Our model significantly improves the modeling ability of the change detection on one dataset and demonstrates good generalization on another dataset without any additional operations. To achieve this, we employ a Siamese neural network for change detection and design an IBNM (Instance Normalization and Batch Normalization Module) that utilizes instance normalization and batch normalization in order to serve as the encoder backbone in the Siamese neural network. The IBNM extracts feature maps for each layer, and the Siamese neural network fuses the feature maps of the two branches using a unique operation. Finally, a simple MLP decoder is used for end-to-end change detection. We train our model on the LEVIR-CD dataset and achieve competitive performance on the test set. In cross-domain dataset testing, CrossCDNet outperforms all the other compared models. Specifically, our model achieves an F1-score of 91.69% on the LEVIR-CD dataset and an F1-score of 77.09% on the WHU-CD dataset, where the training set was LEVIR-CD.

Lightweight change detection models are essential for industrial applications and edge devices. Reducing the model size while maintaining high accuracy is a key challenge in developing lightweight change detection models. However, many existing methods oversimplify the model architecture, leading to a loss of information and reduced performance. Therefore, developing a lightweight model that can effectively preserve the input information is a challenging problem. To address this challenge, we propose LightCDNet, a novel lightweight change detection model that effectively preserves the input information. LightCDNet consists of an early fusion backbone network and a pyramid decoder for end-to-end change detection. The core component of LightCDNet is the Deep Supervised Fusion Module (DSFM), which guides the early fusion of primary features to improve performance. We evaluated LightCDNet on the LEVIR-CD dataset and found that it achieved comparable or better performance than state-of-the-art models while being 10–117 times smaller in size.

The cliff ecosystem is one of the least human-disturbed ecosystems in nature, and its inaccessible and often extreme habitats are home to many ancient and unique plant species. Because of the harshness of cliff habitats, their high elevation, steepness of slopes, and inaccessibility to humans, surveying cliffs is incredibly challenging. Comprehensive and systematic information on cliff vegetation cover is not unavailable but obtaining such information on these cliffs is fundamentally important and of high priority for environmentalists. Traditional coverage survey methods-such as large-area normalized difference vegetation index (NDVI) statistics and small-area quadratic sampling surveys-are not suitable for cliffs that are close to vertical. This paper presents a semi-automatic systematic investigation and a three-dimensional reconstruction of karst cliffs for vegetation cover evaluation. High-resolution imagery with structure from motion (SFM) was captured by a smart unmanned aerial vehicle (UAV). Using approximately 13,000 records retrieved from high-resolution images of 16 cliffs in the karst region Guilin, China, 16 models of cliffs were reconstructed. The results show that this optimized UAV photogrammetry method greatly improves modeling efficiency and the vegetation cover from the bottom to the top of cliffs is high-low-high, and very few cliffs have high-low cover at the top. This study highlights the unique vegetation cover of karst cliffs, which warrants further research on the use of SFM to retrieve cliff vegetation cover at large and global scales.

The use of remote sensing images to detect forest changes is of great significance for forest resource management. With the development and implementation of deep learning algorithms in change detection, a large number of models have been designed to detect changes in multi-phase remote sensing images. Although synthetic aperture radar (SAR) data have strong potential for application in forest change detection tasks, most existing deep learning-based models have been designed for optical imagery. Therefore, to effectively combine optical and SAR data in forest change detection, this paper proposes a double Siamese branch-based change detection network called DSNUNet. DSNUNet uses two sets of feature branches to extract features from dual-phase optical and SAR images and employs shared weights to combine features into groups. In the proposed DSNUNet, different feature extraction branch widths were used to compensate for a difference in the amount of information between optical and SAR images. The proposed DSNUNet was validated by experiments on the manually annotated forest change detection dataset. According to the obtained results, the proposed method outperformed other change detection methods, achieving an F1-score of 76.40%. In addition, different combinations of width between feature extraction branches were analyzed in this study. The results revealed an optimal performance of the model at initial channel numbers of the optical imaging branch and SAR image branch of 32 and 8, respectively. The prediction results demonstrated the effectiveness of the proposed method in accurately predicting forest changes and suppressing cloud interferences to some extent.