Weakly Supervised Network with Scribble-Supervised and Edge-Mask for Road Extraction from High-Resolution Remote Sensing Images

Significant advancements have been achieved in road surface extraction based on high-resolution remote sensingimage processing. Most current methods rely on fully supervised learning, which necessitates enormous humaneffort to label the image. Within this field, other research endeavors utilize weakly supervised methods. Theseapproaches aim to reduce the expenses associated with annotation by leveraging sparsely annotated data, such asscribbles. This paper presents a novel technique called a weakly supervised network using scribble-supervised andedge-mask (WSSE-net). This network is a three-branch network architecture, whereby each branch is equippedwith a distinct decoder module dedicated to road extraction tasks. One of the branches is dedicated to generatingedge masks using edge detection algorithms and optimizing road edge details. The other two branches supervise themodel’s training by employing scribble labels and spreading scribble information throughout the image. To addressthe historical flaw that created pseudo-labels that are not updated with network training, we use mixup to blendprediction results dynamically and continually update new pseudo-labels to steer network training. Our solutiondemonstrates efficient operation by simultaneously considering both edge-mask aid and dynamic pseudo-labelsupport. The studies are conducted on three separate road datasets, which consist primarily of high-resolutionremote-sensing satellite photos and drone images. The experimental findings suggest that our methodologyperforms better than advanced scribble-supervised approaches and specific traditional fully supervised methods.

Land Use Land Cover Analysis for Godavari Basin in Maharashtra Using Geographical Information System and Remote Sensing

The dynamic transformation of land use and land cover has emerged as a crucial aspect in the effective management of natural resources and the continual monitoring of environmental shifts. This study focused on the land use and land cover (LULC) changes within the catchment area of the Godavari River, assessing the repercussions of land and water resource exploitation. Utilizing LANDSAT satellite images from 2009, 2014, and 2019, this research employed supervised classification through the Quantum Geographic Information System (QGIS) software’s SCP plugin. Maximum likelihood classification algorithm was used for the assessment of supervised land use classification. Seven distinct LULC classes—forest, irrigated cropland, agricultural land (fallow), barren land, shrub land, water, and urban land—are delineated for classification purposes. The study revealed substantial changes in the Godavari basin’s land use patterns over the ten-year period from 2009 to 2019. Spatial and temporal dynamics of land use/cover changes (2009-2019) were quantified using three Satellite/Landsat images, a supervised classification algorithm and the post classification change detection technique in GIS. The total study area of the Godavari basin in Maharashtra encompasses 5138175.48 hectares. Notably, the built-up area increased from 0.14% in 2009 to 1.94% in 2019. The proportion of irrigated cropland, which was 62.32% in 2009, declined to 41.52% in 2019. Shrub land witnessed a noteworthy increase from 0.05% to 2.05% over the last decade. The key findings underscored significant declines in barren land, agricultural land, and irrigated cropland, juxtaposed with an expansion in forest land, shrub land, and urban land. The classification methodology achieved an overall accuracy of 80%, with a Kappa Statistic of 71.9% for the satellite images. The overall classification accuracy along with the Kappa value for 2009, 2014 and 2019 supervised land use land cover classification was good enough to detect the changing scenarios of Godavari River basin under study. These findings provide valuable insights for discerning land utilization across various categories, facilitating the adoption of appropriate strategies for sustainable land use in the region.

Evaluation of Eco-Environmental Frangibility Based on Remote Sensing and Geographic Information System

The eco-environmental frangibility is studied by choosing the factors of land use class change and vegetation cover rate, and the equation of eco-environmental frangibility and its evaluation system are established based on remote sensing （RS） and geographic information system technology （GIS）. Four different years of TM images are selected to calculate land use change grads and vegetation cover rate, and the relationship between the two factors and eco-environment frangibility index are build, taking Fuzhou as an example. The character of times change and space distribution of eco-environment frangibility are described. The result indicates the area of eco-environment frangibility increased 2.6% in Fuzhou during twelve years, and expands from the region between infield and forest land to forest land in space distribution.

Integrated Evaluation Model for Eco-Environmental Quality in Mountainous Region Based on Remote Sensing and GIS

Based on Remote Sensing （RS）, Geographic Information System （GIS）, and combining Principal Component Analysis, this paper designed a numerical integrated evaluation model for mountain eco-environment on the base of grid scale. Using this model, we evaluated the mountain eco-environmental quality in a case study area-the upper reaches of Minjiang River, and achieved a good result, which accorded well with the real condition. The study indicates that, the integrated evaluation model is suitable for multi-layer spatial factor computation, effectively lowing man＇s subjective influence in the evaluation process; treating the whole river basin as a system, the model shows full respect to the circulation of material and energy, synthetically embodies the determining impact of such natural condition as water-heat and landform, as well as human interference in natural eco-system; the evaluation result not only clearly presents mountainous vertical distribution features of input factors, but also provides a scientific and reliable thought for quantitatively evaluating mountain eco-environment.

INTEGRATED VEGETATION CLASSIFICATION AND MAPPINGUSING REMOTE SENSING AND GIS TECHNIQUES

NOAA-AVHRR data have been more and more used by scientists because of its short temporal resolution,large scope, inexpensive cost and broad wave bands. On macro and middle scale of vegetation remote sensing, NOAAAVHRR possesses an advantage when compared with other satellites. However, because NOAA-AVHRR also problem of low resolution, data distortion and geometrical distortion, in the area of application of NOAA-AVHRR in largescale vegetation - mapping, the accuracy of vegetation classification should be improved. This paper discuss the feasibilityof integrating the geographic data in GIS(Geographical Information System) and remotely sensed data in GIS. Under theenvironment of GIS, temperature, precipitation and elevation, which serve as main factors affecting vegetation growth,were processed by a mathematical model and qualified into geographic image under a certain grid system. The geographicimage were overlaid to the NOAA-AVHRR data which had been compressed and processed. In order to evaluate the usefulness of geographic data for vegetation classification, the area under study was digitally classified by two groups of interpreter: the proposed methodology using maximum likelihood classification assisted by the geographic database and a conventional maximum likelihood classification only. Both result were compared using Kappa statistics. The indices to both theproposed and the conventional digital classification methodology were 0. 668(yew good) and 0. 563(good), respetively.The geographic database rendered an improvement over the conventional digital classification. Furthermore, in this study,some problems related to multi-sources data integration are also discussed.

Extending self-organizing maps for supervised classification of remotely sensed data

An extended self-organizing map for supervised classification is proposed in this paper. Unlike other traditional SOMs, the model has an input layer, a Kohonen layer, and an output layer. The number of neurons in the input layer depends on the dimensionality of input patterns. The number of neurons in the output layer equals the number of the desired classes. The number of neurons in the Kohonen layer may be a few to several thousands, which depends on the complexity of classification problems and the classification precision. Each training sample is expressed by a pair of vectors ： an input vector and a class codebook vector. When a training sample is input into the model, Kohonen＇s competitive learning rule is applied to selecting the winning neuron from the Kohouen layer and the weight coefficients connecting all the neurons in the input layer with both the winning neuron and its neighbors in the Kohonen layer are modified to be closer to the input vector, and those connecting all the neurons around the winning neuron within a certain diameter in the Kohonen layer with all the neurons in the output layer are adjusted to be closer to the class codebook vector. If the number of training sam- ples is sufficiently large and the learning epochs iterate enough times, the model will be able to serve as a supervised classifier. The model has been tentatively applied to the supervised classification of multispectral remotely sensed data. The author compared the performances of the extended SOM and BPN in remotely sensed data classification. The investigation manifests that the extended SOM is feasible for supervised classification.

Studying the Impact of Pollution from Wadi Gaza on the Mediterranean Sea Using GIS and Remote Sensing Techniques

Wadi Gaza is considered as one of the most important coastal wetlands located on the Eastern Mediterranean Basin. It is witnessing rapid degradation due to anthropogenic activities including but not limited to discharge of municipal sewage, dumping of solid wastes, rampant use of pesticides and illegal poaching. They form a river of untreated wastewater, more than 5 km long, before its discharge into the Mediterranean Sea. This study aims to perform an analytical study of Wadi Gaza and study its effects on the pollution of the seawater opposite to it using GIS and remote sensing techniques. The flow accumulation, the watershed and the stream orders inside and outside the Gaza Strip are determined based on a DEM which involves a radar terrestrial scanning of Palestine carried out by NASA’s Endeavor Space Shuttle. The area of the watershed inside Gaza is estimated to be equal to 58.792 km2. The Study also shows that the total amount of contaminated water that flows into the sea can be estimated to reach 146.5 mm3/year. The total area of coastal sea contamination approximately reaches 38.8 km2 and is oriented to the north direction along the coastal shore and its influence extends to Gaza seaport, 10 km apart from the Wadi.

Land Use/Land Cover Change Detection in Pokhara Metropolitan, Nepal Using Remote Sensing

Land use and land cover are essential for maintaining and managing the natural resources on the earth surface. A complex set of economic, demographic, social, cultural, technological, and environmental processes usually result in the change in the land use/land cover change (LULC). Pokhara Metropolitan is influenced mainly by the combination of various driving forces: geographical location, high rate of population growth, economic opportunity, globalization, tourism activities, and political activities. In addition to this, geographically steep slope, rugged terrain, and fragile geomorphic conditions and the frequency of earthquakes, floods, and landslides make the Pokhara Metropolitan region a disaster-prone area. The increment of the population along with infrastructure development of a given territory leads towards the urbanization. It has been rapidly changing due to urbanization, industrialization and internal migration since the 1970s. The landscapes and ground patterns are frequently changing on time and prone to disaster. Here a study has been carried to study on LULC for the last 18 years (2000-2018). The supervised classification on Landsat Imagery was performed and verified the classification through computing the error matrix. Besides, the water bodies and vegetation area were extracted through the Normalized Difference Water Index (NDWI) and Normalized Difference Vegetation Index (NDWI) respectively. This research shows that during the last 18 years the agricultural areas diminishing by 15.66% while urban area is increasing by 13.2%. This research is beneficial for preparing the plan and policy in the sustainable development of Pokhara Metropolitan.

MDSNet:a multiscale decoupled supervision network for semantic segmentation of remote sensing images

Recent deep-learning successes have led to a new wave of semantic segmentation in remote sensing(RS)applications.However,most approaches rarely distinguish the role of the body and edge of RS ground objects;thus,our understanding of these semantic parts has been frustrated by the lack of detailed geometry and appearance.Here we present a multiscale decoupled supervision network for RS semantic segmentation.Our proposed framework extends a densely supervised encoder-decoder network with a feature decoupling module that can decouple semantic features with different scales into distinct body and edge components.We further conduct multiscale supervision of the original and decoupled body and edge features to enhance inner consistency and spatial boundaries in remote sensing image(RSl)ground objects,enabling new segmentation designs and semantic components that can learn to perform multiscale geometry,and appearance.Our results outperform the previous algorithm and are robust to different datasets.These results demonstrate that decoupled supervision is an effective solution to semantic segmentation tasks of RS images.

Riverine Flood Damage Assessment of Cultivated Lands along Chenab River Using GIS and Remotely Sensed Data: A Case Study of District Hafizabad, Punjab, Pakistan

Flood is one of a kind of disasters which harms human and animal life around the globe. Pakistan has been observing massive floods for many years because of daily and seasonal variation in the temperature levels. Wheat, rice, sugarcane and cotton are major crops cultivated in Punjab region of Pakistan in which rice and sugarcane are mostly effected by floods. In this research paper, damage assessment of cultivated land in district Hafizabad along Chenab River has been calculated. Supervised Classification and Soil Adjusted Vegetation Index (SAVI) methods are applied. Pre-flood 2014, post-flood 2014, and pre-flood 2015 Landsat 8 images have been used to calculate the extent of damages to cultivated lands. Water, sand, silt, bare soil and vegetation are classified to identify damage. Results show that vegetation cover has plummeted to 50% after the arrival of flood 2014 in the Chenab. Similarly, 6.7047% of sand and 15.7339% of bare soil deposits have surfaced which have not yet been removed from fertile lands in 2015. 18.4376% standing crop damage has been analyzed under this study. 14.0245% silt deposits have been calculated as post-flood effects. 46.4260% land has been cultivated in 2015 which is 15.5024% lower than 2014 cultivated land. Furthermore, field verification survey has given promising results and has a great correlation with satellite based recovery results.

Eco-environment evolvement analysis of Ertan reservoir catchment based on remote sensing

Hydropower project may bring with it social-economic profits as well as side effects.The built dam and reservoir often cause some problems to the surrounding areas,among which the ecological and environmental effects caused by hydropower projects are always concerned by the public.In this article,we take the Ertan reservoir catchment as the research area and try to quantitatively analyze the variation of vegetation cover and soil erosion by remote sensing technique,and to comprehensively assess the evolvement and development trend of reservoir catchment.Soil erosion,land use/cover are used as ecological and environmental indicators which reflect the changes before,after and in the period of the construction of Ertan hydropower station.Supported by the multi-source remote sensing data(from satellite Landsat and CBERS) and DEM data,the land use/cover is interpreted through RS images which are classified both by unsupervised and supervised method,and the driving factors of the ecological changes are also analyzed.At the same time,the changes of soil loss are also monitored and analyzed during flood seasons of Ertan reservoir area before and after reservoir impoundment(1995,2000 and 2005) using the revised universal soil loss equation(RUSLE) .The results show that during the recent 13 years the arable land area has decreased obviously,and construction area and water surface have increased slightly.The increase of vegetation cover has some relations with the implementation of local ecological projects,i.e.,de-farming to forestry and de-farming to pasture projects.At the same time,changes may also be caused by the climate adjustment in the reservoir area.In the ten years from 1995 to 2005,the high soil loss classes were transforming to lowly level classes continuously.All of these show that the soil loss of Ertan reservoir area is getting better.

Spatiotemporal variations of eco-environment in the Guangxi Beibu Gulf Economic Zone based on remote sensing ecological index and granular computing

Accurate and rapid evaluation of the regional eco-environment is critical to policy formulation.The remote sensing ecological index(RSEI)model of the Guangxi Beibu Gulf Economic Zone(GBGEZ)during 2001-2020 was established and evaluated using four indices:dryness,wetness,greenness,and heat.This paper proposes an information granulation method for remote sensing based on the RSEI index value that uses granular computing.We found that:(1)From 2001 to 2020,the eco-environmental quality(EEQ)of GBGEZ tended to improve,and the spatial difference tended to expand.The regional spatial distribution of the eco-environment is primarily in the second-level and third-level areas,and the EEQ in the east and west is better than that in the middle.The contribution of greenness,wetness,and dryness to the improvement of EEQ in the study region increased year by year.(2)From 2001to 2020,the order of the contribution of the EEQ index in the GBGEZ was dryness,wetness,greenness,and heat.(3)The social and economic activities in the study region had a certain inhibitory effect on the improvement of the EEQ.

面向多源异质遥感影像地物分类的自监督预训练方法

近年来,深度学习改变了遥感图像处理的方法。由于标注高质量样本费时费力,标签样本数量不足的现实问题会严重影响深层神经网络模型的性能。为解决这一突出矛盾,本文提出了用于多源异质遥感影像地物分类的自监督预训练和微调分类方案,旨在缓解模型对于标签样本的严重依赖。具体来讲,生成式自监督学习模型由非对称的编码器-解码器结构组成,其中深度编码器从多源遥感数据中学习高阶关键特征,任务特定的解码器用于重建原始遥感影像。为提升特性表示能力,交叉注意力机制模型用于融合异源特征中的信息,进而从多源异质遥感影像中学习更多的互补信息。在微调分类阶段,预训练好的编码器作为无监督特征提取器,基于Transformer结构的轻量级分类器将学习到的特征与光谱信息结合并用于地物分类。这种自监督预训练方案能够从多源异质遥感影像中学习到刻画原始数据的高级关键特征,并且此过程不需要任何人工标注信息,从而缓解了对标签样本的依赖。与现有的分类范式相比,本文提出的自监督预训练和微调方案在多源遥感影像地物分类中能够取得更优的分类结果。

基于Sparse R-CNN的遥感目标检测研究

遥感图像目标检测任务在天气预报、环境监测及军事应用等领域均有应用,但其小目标众多、类间相似度大、尺度多样等问题导致提取特征困难。基于深度学习的方法在目标检测领域已经流行起来,Sparse R-CNN是一种结构简单且效果较好的模型,但将其直接应用到遥感图像上结果较差,针对遥感图像特点引入了自监督学习框架跟选择性查询收集提高了遥感图像目标检测的效果,在mAP指标上提高3.8个百分点。

基于高分辨率遥感影像的居民地分类方法研究

随着我国城镇化进程的不断深入,大量乡村居民地已经转化为城镇居民地,为了度量二者实时的转化情况,该文提出一种基于监督分类机制的居民地分类方法,该方法首先采用边缘特征及高斯函数量化影像上的局部特征,然后构建5种城镇及乡村居民地分类规则,其次创建训练样本对各类规则进行学习,最后通过城镇及乡村测试样本验证该文方法的精度。实验表明,该文方法可以对高分辨率遥感影像城镇及乡村居民地进行初级分类,为“城镇化”进程提供一个新的衡量指标。

基于半监督对抗学习的遥感图像水体提取

语义分割技术被广泛应用于遥感图像水体提取任务中,然而语义分割的结果极大依赖于数据集的规模,针对遥感图像中水体数据集较少、获得精准标注数据成本高的问题,提出一种用于水体提取的半监督对抗语义分割方法。作为生成器的分割网络中的卷积操作具有受限的感受野,缺乏对长距离上下文关系的建模能力,Transformer能够建模图像的全局信息。该方法在分割网络中采用Swin Transformer建模深层特征的全局上下文信息,挖掘像素之间的语义关系,提高网络的特征提取能力。采用双卷积块提取图像的局部特征,保留高分辨率细节信息。特征增强模块(FEM)用于抑制图像的背景噪声干扰,进一步提高水体提取的精度。分割网络和判别器网络共同训练,以提高在使用少量有标签数据条件下模型提取水体的性能。在GID数据集上进行大量实验,结果表明,该方法在不同比例有标签数据条件下均提高了水体提取的精度,在仅1/8有标签数据的条件下,该方法取得的F1-Score和交并比(Io U)分别为90.02%和81.86%,优于U-Net、MWEN等语义分割网络。

多尺度引导滤波对多光谱遥感影像分类精度的影响

由于地物的复杂多样性,准确识别其分类精度对遥感数据处理具有重要意义。为提高多光谱遥感数据的分类精度,本文以Landsat 8数据为基础,提出不同尺度引导滤波特征融合NDVI与NDBI的方法,进行多光谱遥感图像的分类。首先,提取多光谱数据第一主成分作为引导图像,原图像为输入图像,依次提取滤波半径为2、4、6、8的引导滤波特征集;然后,将不同滤波半径的引导滤波特征集与图像NDVI与NDBI特征进行融合,采用支持向量机的方法进行监督分类,以此探究不同尺度的引导滤波对多光谱遥感影像分类精度的影响。试验结果表明:①引导滤波在去除噪声的同时能够较好地保留图像的边缘特征;②引导滤波可以提高多光谱遥感影像的分类精度,不同大小引导滤波半径图像在分类方面与原图像相比,较其分类精度均有不同程度的提升,最高总体精度达99.7763%,Kappa系数为0.9971;③不同尺度的引导滤波会得到不同的分类结果,当滤波半径R=2时,图像的分类精度最高。

基于数字孪生的水土流失防治责任范围变化预测模型

为建立实用、高效的水土流失防治责任范围变化预测模型,提高合规性判断的准确性,在总结水土流失防治责任范围可撤销性、时空交错性、临时性和条件适用性等4个主要动态特性的基础上,提出一种基于数字孪生的水土流失防治责任范围变化预测模型。数学模型为描述防治责任范围状态的虚拟模型,防治责任范围为物理实体,模型基于防治责任范围变化主要因素的孪生数据,对防治责任范围有效性状态进行评估和预测。结果表明:模型对防治责任范围变化进行预测的正确率提升约38%,具有良好的预测防治责任范围变化的性能,有助于进一步提升提高遥感监管成果的质量,在生产建设项目遥感监管应用领域有一定的借鉴意义。

多级对比学习下的弱监督高分遥感影像城市固废堆场提取

城市固体废物是城市化进程中的重要污染源,对城市生态环境和公共健康造成了巨大危害。高分影像固废堆场智能解译是实现自动排查,提升监测效率的核心和关键技术。基于深度学习的固废堆场自动提取方法严重依赖于获取成本高、制作难度大的高质量像素级标注。为此,本文提出使用更易获取的影像级标注,利用影像自监督学习实现像素级固废堆场提取。围绕固废堆场的影像特征,本文方法在尺度对比约束下综合像素、影像两个层次的对比学习方法,对固废堆场的类别激活图细化和完善,并基于此生成高质量的固废堆场伪像素级标注,用于训练固废堆场提取模型。试验结果表明,本文方法在固废堆场提取的F 1值和IoU分数方面分别达到了71.58%和55.74%,显著优于所有对比方法。这说明利用多级对比学习的弱监督方法能够获得更加完整且准确的类别激活图,从而取得更高的固废堆场提取精度。

基于布局化-语义联合表征遥感图文检索方法

遥感图文检索可以从类别繁多、内容复杂的遥感数据中检索到有价值的信息,对环境评估、城市规划以及灾害预测具有重要意义。但是,遥感图文跨模态检索存在一个关键问题,即遥感图像的空间布局信息被忽略。其主要体现在2个方面:①遥感目标的远距离建模困难;②遥感相邻次要目标被淹没。基于以上问题,提出了一种基于布局化-语义联合表征的跨模态遥感图像文本检索(SL-SJR),主要包括主导语义监督的布局化视觉特征提取(DSSL)模块、布局化视觉-全局语义交叉指导(LV-GSCG)模块和多视角匹配(MVM)模块。DSSL模块实现主导语义类别特征监督下图像的布局化建模。LV-GSCG模块计算布局化视觉特征与文本中提取的全局语义特征的相似度来实现不同模态特征的交互。MVM模块建立跨模态特征指导的多视角度量匹配机制以消除跨模态数据之间的语义鸿沟。在4个基线遥感图像文本数据集上的实验验证,结果表明所提方法在大多数跨模态遥感图像文本检索任务中可以达到最先进的性能。