We mapped the forest cover of Khadimnagar National Park (KNP) of Sylhet Forest Division and estimated forest change over a period of 22 years (1988-2010) using Landsat TM images and other GIS data. Supervised clas...We mapped the forest cover of Khadimnagar National Park (KNP) of Sylhet Forest Division and estimated forest change over a period of 22 years (1988-2010) using Landsat TM images and other GIS data. Supervised classification and Normalized Difference Vegetation Index (NDVI) image classification approaches were applied to the images to produce three cover classes, viz. dense forest, medium dense forest, and bare land. The change map was produced by differencing classified imageries of 1988 and 2010 as before image and after image, respectively, in ERDAS IMAGINE. Error matrix and kappa statistics were used to assess the accuracy of the produced maps. Overall map accuracies resulting from supervised classification of 1988 and 2010 imageries were 84.6% (Kappa 0.75) and 87.5% (Kappa 0.80), respec- tively. Forest cover statistics resulting from supervised classification showed that dense forest and bare land declined from 526 ha (67%) to 417 ha (59%) and 105 ha (13%) to 8 ha (1%), respectively, whereas medium dense forest increased from 155 ha (20%) to 317 ha (40%). Forest cover change statistics derived from NDVI classification showed that dense forest declined from 525 ha (67%) to 421 ha (54%) while medium dense forest increased from 253 ha (32%) to 356 ha (45%). Both supervised and NDVI classification approaches showed similar trends of forest change, i.e. decrease of dense forest and increase of medium dense forest, which indicates dense forest has been converted to medium dense forest. Area of bare land was unchanged. Illicit felling, encroachment, and settlement near forests caused the dense forest decline while short and long rotation plantations raised in various years caused the increase in area of medium dense forest. Protective measures should be undertaken to check further degradation of forest at KNP.展开更多
Background: Monitoring the changing pattern of vegetation across diverse landscapes through remote sensing is instrumental in understanding the interactions of human activities and the ecological environment. Land us...Background: Monitoring the changing pattern of vegetation across diverse landscapes through remote sensing is instrumental in understanding the interactions of human activities and the ecological environment. Land use pattern i the state of Himachal Pradesh in the Indian Western Himalayas has been undergoing rapid modifications due to changing cropping patterns, rising anthropogenic pressure on forests and government policies. We studied land use change in Solan Forest Division of Himachal Pradesh to assess species wise area changes in the forests of the region. Methods: The supervised classification (Maximum likelihood) on two dates of IRS (LISS III) satellite data was performed to assess land use change over the period 1998-2010. Results: Seven land use categories were identified namely, chir pine (Pinus roxburghii) forest, broadleaved forest, bamboo (Dendrocolamus strictus) forest, ban oak (Quercus leucotrichophora) forest, khair (Acacia catechu) forest, culturable blank and cultivation. The area under chir pine, cultivation and khair forests increased by 191 ha (4.55 %), 129 ha (13.81%) and 77 ha (23.40 %), whereas the area under ban oak, broadleaved, culturable blank and bamboo decreased by 181 ha (16.58 %), 152 ha (6.30 %), 71 ha (2.72 %) and 7 ha (0.47 %), respectively. Conclusions: The study revealed a decrease in the area under forest and culturable blank categories and a simultaneous increase in the area under cultivation primarily due to the large scale introduction of horticultural cash crops in the state. The composition of forests also exhibited some major changes, with an increase in the area of commercially important monoculture plantation species such as pine and khair, and a decline in the area of oak, broadleaved and bamboo which are facing a high anthropogenic pressure in meeting the livelihood demands of forest dependent communities. In time deforestation, forest degradation and ecological imbalances due to the changing forest species composition may inflict irreversible damages upon unstable and fragile mountain zones such as the Indian Himalayas. The associated common property externalities involved at local, regional and global scales, necessitate the monitoring of land use dynamics across forested landscapes in developing future strategies and policies concerning agricultural diversification, natural forest conservation and monoculture tree plantations.展开更多
Many supervised classification algorithms have been proposed, however, they are rarely evaluated for specific application. This research examines the performance of machine learning classifiers support vector machine ...Many supervised classification algorithms have been proposed, however, they are rarely evaluated for specific application. This research examines the performance of machine learning classifiers support vector machine (SVM), neural network (NN), Random Forest (RF) against maximum classifier (MLC) (traditional supervised classifier) in forest resources and land cover categorization, based on combination of Advanced Land Observing Satellite (ALOS) Phased Array type L-band Synthetic Aperture Radar (PALSAR) and Landsat Thematic Mapper (TM) data, in Northern Tanzania. Various data categories based on Landsat TM surface reflectance, ALOS PALSAR backscattering and their derivatives were generated for various classification scenarios. Then a separate and joint processing of Landsat and ALOS PALSAR data were executed using SVM, NN, RF and ML classifiers. The overall classification accuracy (OA), kappa coefficient (KC) and F1 score index values were computed. The result proves the robustness of SVM and RF in classification of forest resource and land cover using mere Landsat data and integration of Landsat and PALSAR (average OA = 92% and F1 = 0.7 to 1). A two sample t-statistics was utilized to evaluate the performance of the classifiers using different data categories. SVM and RF indicate there is no significance difference at 5% significance level. SVM and RF show a significant difference when compared to NN and ML. Generally, the study suggests that parametric classifiers indicate better performance compared to parametric classifier.展开更多
【目的】森林火灾识别是避免森林火灾大面积蔓延的一项重要研究。随着深度学习的快速发展,基于卷积神经网络的模型因其在图像识别领域的优异表现,被广泛应用到森林火灾识别任务当中。然而,基于卷积神经网络的方法通常在标签数据不充分时...【目的】森林火灾识别是避免森林火灾大面积蔓延的一项重要研究。随着深度学习的快速发展,基于卷积神经网络的模型因其在图像识别领域的优异表现,被广泛应用到森林火灾识别任务当中。然而,基于卷积神经网络的方法通常在标签数据不充分时,难以取得令人满意的森林火灾识别结果。【方法】本研究提出了一种基于视觉变换网络的自监督森林火灾识别模型(Self supervised forest fire identification model based on visual transformation network),来提高模型在标签稀缺情况下的森林火灾识别精度。具体来说,该模型采用视觉变换网络作为主干网络,通过视觉变换网络中的多头自注意力机制来捕获森林火灾图像的全局信息特征。并且引入自监督学习中的图像重建任务来辅助模型训练,从而减少模型对标签数据的依赖。模型通过对掩盖图像的特征恢复和重建学习相关语义信息。同时,本研究还提出了一种基于傅里叶低频混合变换的数据增强方法来提高模型的泛化性和鲁棒性。【结果】通过开展详细的试验来验证模型的有效性,结果表明,与其他常见的网络模型相比,FFDM模型在森林火灾识别任务中取得了最佳的识别效果,其识别准确率为89.51%,比VGG16网络高13.7%,比ResNet50网络高8.2%,比InceptionV3网络高7.2%。【结论】通过自监督学习辅助模型训练的方法,FFDM模型即使在标签稀缺下依然可以取得不错的森林火灾识别效果。展开更多
文摘We mapped the forest cover of Khadimnagar National Park (KNP) of Sylhet Forest Division and estimated forest change over a period of 22 years (1988-2010) using Landsat TM images and other GIS data. Supervised classification and Normalized Difference Vegetation Index (NDVI) image classification approaches were applied to the images to produce three cover classes, viz. dense forest, medium dense forest, and bare land. The change map was produced by differencing classified imageries of 1988 and 2010 as before image and after image, respectively, in ERDAS IMAGINE. Error matrix and kappa statistics were used to assess the accuracy of the produced maps. Overall map accuracies resulting from supervised classification of 1988 and 2010 imageries were 84.6% (Kappa 0.75) and 87.5% (Kappa 0.80), respec- tively. Forest cover statistics resulting from supervised classification showed that dense forest and bare land declined from 526 ha (67%) to 417 ha (59%) and 105 ha (13%) to 8 ha (1%), respectively, whereas medium dense forest increased from 155 ha (20%) to 317 ha (40%). Forest cover change statistics derived from NDVI classification showed that dense forest declined from 525 ha (67%) to 421 ha (54%) while medium dense forest increased from 253 ha (32%) to 356 ha (45%). Both supervised and NDVI classification approaches showed similar trends of forest change, i.e. decrease of dense forest and increase of medium dense forest, which indicates dense forest has been converted to medium dense forest. Area of bare land was unchanged. Illicit felling, encroachment, and settlement near forests caused the dense forest decline while short and long rotation plantations raised in various years caused the increase in area of medium dense forest. Protective measures should be undertaken to check further degradation of forest at KNP.
文摘Background: Monitoring the changing pattern of vegetation across diverse landscapes through remote sensing is instrumental in understanding the interactions of human activities and the ecological environment. Land use pattern i the state of Himachal Pradesh in the Indian Western Himalayas has been undergoing rapid modifications due to changing cropping patterns, rising anthropogenic pressure on forests and government policies. We studied land use change in Solan Forest Division of Himachal Pradesh to assess species wise area changes in the forests of the region. Methods: The supervised classification (Maximum likelihood) on two dates of IRS (LISS III) satellite data was performed to assess land use change over the period 1998-2010. Results: Seven land use categories were identified namely, chir pine (Pinus roxburghii) forest, broadleaved forest, bamboo (Dendrocolamus strictus) forest, ban oak (Quercus leucotrichophora) forest, khair (Acacia catechu) forest, culturable blank and cultivation. The area under chir pine, cultivation and khair forests increased by 191 ha (4.55 %), 129 ha (13.81%) and 77 ha (23.40 %), whereas the area under ban oak, broadleaved, culturable blank and bamboo decreased by 181 ha (16.58 %), 152 ha (6.30 %), 71 ha (2.72 %) and 7 ha (0.47 %), respectively. Conclusions: The study revealed a decrease in the area under forest and culturable blank categories and a simultaneous increase in the area under cultivation primarily due to the large scale introduction of horticultural cash crops in the state. The composition of forests also exhibited some major changes, with an increase in the area of commercially important monoculture plantation species such as pine and khair, and a decline in the area of oak, broadleaved and bamboo which are facing a high anthropogenic pressure in meeting the livelihood demands of forest dependent communities. In time deforestation, forest degradation and ecological imbalances due to the changing forest species composition may inflict irreversible damages upon unstable and fragile mountain zones such as the Indian Himalayas. The associated common property externalities involved at local, regional and global scales, necessitate the monitoring of land use dynamics across forested landscapes in developing future strategies and policies concerning agricultural diversification, natural forest conservation and monoculture tree plantations.
文摘Many supervised classification algorithms have been proposed, however, they are rarely evaluated for specific application. This research examines the performance of machine learning classifiers support vector machine (SVM), neural network (NN), Random Forest (RF) against maximum classifier (MLC) (traditional supervised classifier) in forest resources and land cover categorization, based on combination of Advanced Land Observing Satellite (ALOS) Phased Array type L-band Synthetic Aperture Radar (PALSAR) and Landsat Thematic Mapper (TM) data, in Northern Tanzania. Various data categories based on Landsat TM surface reflectance, ALOS PALSAR backscattering and their derivatives were generated for various classification scenarios. Then a separate and joint processing of Landsat and ALOS PALSAR data were executed using SVM, NN, RF and ML classifiers. The overall classification accuracy (OA), kappa coefficient (KC) and F1 score index values were computed. The result proves the robustness of SVM and RF in classification of forest resource and land cover using mere Landsat data and integration of Landsat and PALSAR (average OA = 92% and F1 = 0.7 to 1). A two sample t-statistics was utilized to evaluate the performance of the classifiers using different data categories. SVM and RF indicate there is no significance difference at 5% significance level. SVM and RF show a significant difference when compared to NN and ML. Generally, the study suggests that parametric classifiers indicate better performance compared to parametric classifier.
文摘【目的】森林火灾识别是避免森林火灾大面积蔓延的一项重要研究。随着深度学习的快速发展,基于卷积神经网络的模型因其在图像识别领域的优异表现,被广泛应用到森林火灾识别任务当中。然而,基于卷积神经网络的方法通常在标签数据不充分时,难以取得令人满意的森林火灾识别结果。【方法】本研究提出了一种基于视觉变换网络的自监督森林火灾识别模型(Self supervised forest fire identification model based on visual transformation network),来提高模型在标签稀缺情况下的森林火灾识别精度。具体来说,该模型采用视觉变换网络作为主干网络,通过视觉变换网络中的多头自注意力机制来捕获森林火灾图像的全局信息特征。并且引入自监督学习中的图像重建任务来辅助模型训练,从而减少模型对标签数据的依赖。模型通过对掩盖图像的特征恢复和重建学习相关语义信息。同时,本研究还提出了一种基于傅里叶低频混合变换的数据增强方法来提高模型的泛化性和鲁棒性。【结果】通过开展详细的试验来验证模型的有效性,结果表明,与其他常见的网络模型相比,FFDM模型在森林火灾识别任务中取得了最佳的识别效果,其识别准确率为89.51%,比VGG16网络高13.7%,比ResNet50网络高8.2%,比InceptionV3网络高7.2%。【结论】通过自监督学习辅助模型训练的方法,FFDM模型即使在标签稀缺下依然可以取得不错的森林火灾识别效果。