Multi-scale Segmentation Research Articles

Improving Unsupervised Object-Based Change Detection via Hierarchical Multi-Scale Binary Partition Tree Segmentation: A Case Study in the Yellow River Source Region

Change detection in remote sensing enables identifying alterations in surface characteristics over time, underpinning diverse applications. However, conventional pixel-based algorithms encounter constraints in terms of accuracy when applied to medium- and high-resolution remote sensing images. Although object-oriented methods offer a step forward, they frequently grapple with missing small objects or handling complex features effectively. To bridge these gaps, this paper proposes an unsupervised object-oriented change detection approach empowered by hierarchical multi-scale segmentation for generating binary ecosystem change maps. This approach meticulously segments images into optimal sizes and leverages multidimensional features to adapt the Iteratively Reweighted Multivariate Alteration Detection (IRMAD) algorithm for GaoFen WFV data. We rigorously evaluated its performance in the Yellow River Source Region, a critical ecosystem conservation zone. The results unveil three key strengths: (1) the approach achieved excellent object-level change detection results, making it particularly suited for identifying changes in subtle features; (2) while simply increasing object features did not lead to a linear accuracy gain, optimized feature space construction effectively mitigated dimensionality issues; and (3) the scalability of our approach is underscored by its success in mapping the entire Yellow River Source Region, achieving an overall accuracy of 90.09% and F-score of 0.8844. Furthermore, our analysis reveals that from 2015 to 2022, changed ecosystems comprised approximately 1.42% of the total area, providing valuable insights into regional ecosystem dynamics.

Automatic Extraction for Land Parcels Based on Multi-Scale Segmentation

Different land parcels possess unique microclimates, soils, and biological conditions, which in turn significantly influence the land parcels themselves, impacting biodiversity, hydrological relationships, land degradation, geological disasters, and other ecological environments. Therefore, researching an efficient and accurate method capable of extracting land parcels with the least internal heterogeneity at the macro, meso, and micro scales is extremely important. Multi-scale segmentation, based on scale and resolution analysis techniques, is a bottom-up merging technology that minimizes internal heterogeneity within regions and maximizes heterogeneity between different units. This approach is extensively applied in multi-scale spectral feature extraction and classification and is further combined with deep learning techniques to enhance the accuracy of image classification. This study, using Xinghai County in Qinghai Province as an example, employs multi-scale segmentation and hydrological analysis methods to extract land parcels at different spatial scales. The results show (1) that the land parcels extracted using the hydrological analysis method are catchment units centered around rivers, including slopes on both sides of the river. In contrast, multi-scale segmentation extracts regions comprising land parcels with similar properties, enabling the segregation of slopes and channels into independent units. (2) At a classification threshold of 19, multi-scale segmentation divides the study area into five different types of land parcels, reflecting the heterogeneity of terrain undulations and their hydrological connections. When the classification threshold is set to 31, the study area is divided into 15 types of land parcels, primarily highlighting micro-topographic features. (3) Multi-scale segmentation can merge and categorize areas with the least heterogeneity in land parcels, facilitating subsequent statistical analysis. Therefore, mesoscale land parcels extracted through multi-scale segmentation are invaluable for analyzing regional Earth surface processes such as soil erosion, sediment distribution and transportation. Microscale land parcels are significantly important for identifying high-risk areas in relation to geological disasters like landslides and collapses.

HmsU-Net: A hybrid multi-scale U-net based on a CNN and transformer for medical image segmentation

Accurate medical image segmentation is of great significance for subsequent diagnosis and analysis. The acquisition of multi-scale information plays an important role in segmenting regions of interest of different sizes. With the emergence of Transformers, numerous networks adopted hybrid structures incorporating Transformers and CNNs to learn multi-scale information. However, the majority of research has focused on the design and composition of CNN and Transformer structures, neglecting the inconsistencies in feature learning between Transformer and CNN. This oversight has resulted in the hybrid network's performance not being fully realized. In this work, we proposed a novel hybrid multi-scale segmentation network named HmsU-Net, which effectively fused multi-scale features. Specifically, HmsU-Net employed a parallel design incorporating both CNN and Transformer architectures. To address the inconsistency in feature learning between CNN and Transformer within the same stage, we proposed the multi-scale feature fusion module. For feature fusion across different stages, we introduced the cross-attention module. Comprehensive experiments conducted on various datasets demonstrate that our approach surpasses current state-of-the-art methods.

Forest stand delineation using airborne LiDAR and hyperspectral data

Forest stands, crucial for inventory, planning, and management, traditionally rely on time-consuming visual analysis by forest managers. To enhance efficiency, there is a growing need for automated methods that take into account essential forest attributes. In response, we propose a novel approach utilizing airborne Light Detection and Ranging (LiDAR) and hyperspectral data for automated forest stand delineation. Our approach initiates with over-segmentation of the Canopy Height Model (CHM), followed by attribute calculation for each segment using both CHM and hyperspectral data. Two rules are applied to merge homogeneous segments and eliminate others based on calculated attributes. The effectiveness of our method was validated using three types of reference forest stands with two indices: the explained variance (R2) and Intersection over Union (IoU). Results from our study demonstrated notable accuracy, with a R2 of 97.35% and 97.86% for mean tree height and mean diameter at breast height (DBH), respectively. The R2 for mean canopy height is 81.80%, outperforming manual delineation by 7.31% and multi-scale segmentation results by 2.13%. Furthermore, our approach achieved high IoU values, which indicates a strong spatial agreement with manually delineated forest stands and leading to fewer manual adjustments when applied directly to forest management. In conclusion, our forest stand delineation method enhances both internal consistency and spatial accuracy. This method contributes to improving practical performance and forest management efficiency.

UAVPal: A New Dataset for Semantic Segmentation in Complex Urban Landscape With Efficient Multiscale Segmentation

UAVPal: A New Dataset for Semantic Segmentation in Complex Urban Landscape With Efficient Multiscale Segmentation

One-Shot Learning for Optical Coherence Tomography Angiography Vessel Segmentation Based on Multi-Scale U2-Net

Vessel segmentation in optical coherence tomography angiography (OCTA) is crucial for the detection and diagnosis of various eye diseases. However, it is hard to distinguish intricate vessel morphology and quantify the density of blood vessels due to the large variety of vessel sizes, significant background noise, and small datasets. To this end, a retinal angiography multi-scale segmentation network, integrated with the inception and squeeze-and-excitation modules, is proposed to address the above challenges under the one-shot learning paradigm. Specifically, the inception module extends the receptive field and extracts multi-scale features effectively to handle diverse vessel sizes. Meanwhile, the squeeze-and-excitation module modifies channel weights adaptively to improve the vessel feature extraction ability in complex noise backgrounds. Furthermore, the one-shot learning paradigm is adapted to alleviate the problem of the limited number of images in existing retinal OCTA vascular datasets. Compared with the classic U2-Net, the proposed model gains improvements in the Dice coefficient, accuracy, precision, recall, and intersection over union by 3.74%, 4.72%, 8.62%, 4.87%, and 4.32% respectively. The experimental results demonstrate that the proposed one-shot learning method is an effective solution for retinal angiography image segmentation.

Remote sensing methods for striped marine oil spill detection in narrow ship channels

Maritime transport is the backbone of international trade and the global economy. With the increasing frequency of maritime transport, oil spills are on the rise and so is the risk of oil spills in shipping lanes. Therefore, timely oil spill monitoring at the ship channel is of far-reaching significance for marine disaster prevention and control. This study selects oil spill in the ship channel of the northern Yellow Sea as an example and proposes an object-oriented fuzzy logic classification method combining selected multi-feature parameters based on Sentinel-2 data. Firstly, sensitive bands for oil spills are selected to perform multi-scale segmentation of the oil spill object. Then, 30 samples of oil spill objects were uniformly selected in the oil spill area and the average fuzzy membership values were calculated using 11 membership functions by combining the preferred spectral, geometric, texture, and custom features. Finally, five fuzzy logic operators are used to achieve the final classification, while the oil spill misclassification information is processed and compared with the pixel-based SVM and RF algorithms for accuracy. The experimental results show that the oil spill detection based on the “AND” operator is the best compared to other fuzzy operators, with an overall accuracy of 85.19% and a KAPPA coefficient of 0.753. Relative to the pixel-based Support Vector Machine (SVM) and Random Forest (RF) algorithms, the overall accuracy was improved by 15.66% and 11.25%, respectively, and the KAPPA coefficients were improved by 0.072 and 0.161, respectively. The method is feasible for oil spill detection at ship channels and provides technical support for future marine oil spill detection work.

An commercial area extraction approach using time series nighttime light remote sensing data——Take Wuhan city as a case

The degree of nighttime illumination in a region is closely related to its economic development, therefore, urban commercial areas can be located based on the intensity of the nighttime light. Aiming at the problem of discontinuous and incomparable brightness of nighttime remote sensing images in different periods, a method of urban commercial area extraction based on temporal nighttime remote sensing data is proposed. Firstly, multi- nighttime remote sensing image correction was carried out based on the invariant target region method, and stable time series NPP/VIIRS data set was generated. Secondly, the impervious water surface was extracted by three index synthesis method to obtain the candidate urban commercial area using Landsat-8 image. Finally, multi-scale segmentation and unsupervised classification are carried out on the temporal nighttime remote sensing data set of the candidate area to realize the extraction of urban commercial area. Taking Wuhan city as an example, the method in this paper is used to extract and verify the urban commercial area. The effectiveness and accuracy of the method are verified according to the commercial planning map of Wuhan city and previous studies.

Integrating multisource information to delineate oasis farmland salinity management zones in southern Xinjiang, China

Soil salinization and the shortage of water resources have become important factors affecting the sustainable development of agriculture in the southern Xinjiang. Agricultural flooding in winter or spring is the main means to reduce the damage of secondary soil salinization in southern Xinjiang. However, the spatial heterogeneity of soil salinization is not considered in previous irrigation strategy. Excessive irrigation always leads to a serious waste of agricultural water resources. Irrigation zones delineation based on the degree of soil salinization is beneficial for the management of agricultural salinized land and the improvement of water use efficiency in arid areas. This study used 1599 sets of apparent electrical conductivity (ECa) data obtained before winter irrigation and after crop harvest to characterize soil salinity directly. And the spatial heterogeneity of soil salinity in the study area was analyzed directly using ECa values and statistical as well as geostatistical methods. Three main variables obtained from apparent electrical conductivity measurements (EC0.375, EC0.75, EC1.5), vegetation information (NDVI, GNDVI, SAVI), which indirectly reflected the degree of salinity, and environmental factors (Clay, Sand, Silt, DEM), which affecting the distribution of salinity, were used as auxiliary variables. Four zoning models including soil information, soil + vegetation information, soil information + environmental factors and soil + vegetation information + environmental factors were constructed. A multi-scale segmentation algorithm was used to delineate for four zone patterns and evaluate the results of management zones based on the different information. The results showed that there was a strong spatial heterogeneity of soil salinity in the study area, with the coefficients of variation of ECa all higher than 74 %. But the mean of coefficient of variation intra-zone in each zone after delineation was less than 24 %, which was significantly lower than the value before delineation. The zoning result based on only soil information was more fragmented than other zoning results, and the homogeneity intra-zone and heterogeneity among zones were both lower. This result was not desirable to the development and effective implementation of water resource allocation. Management zones delineation with additional vegetation information or environmental factors showed a better zoning result. Delineation using multisource data of soil, vegetation information and environmental factors not only had the highest homogeneity intra-zone and heterogeneity among zones, but also facilitated the implementation of variable-rate irrigation. Consequently, salinity management zoning, which integrates soil, vegetation information and environmental factors, will facilitate the implementation of efficient irrigation with variable irrigating rate.

Multi-scale spatio-temporal network for skeleton-based gait recognition

Gait has unique physiological characteristics and supports long-distance recognition, so gait recognition is ideal for areas such as home security and identity detection. Methods using graph convolutional networks usually extract features in the spatial and temporal dimensions by stacking GCNs and TCNs, but different joints are interconnected at different moments, so splitting the spatial and temporal dimensions can cause the loss of gait information. Focus on this problem, we propose a gait recognition network, Multi-scale Spatio-Temporal Gait (MST-Gait), which can learn multi-scale gait information simultaneously from spatial and temporal dimensions. We design a multi-scale spatio-temporal groups Transformer (MSTGT) to model the correlation of intra-frame and inter-frame joints simultaneously. And a multi-scale segmentation strategy is designed to capture the periodic and local features of the gait. To fully exploit the temporal information of gait motion, we design a fusion temporal convolution (FTC) to aggregate temporal information at different scales and motion information. Experiments on the popular CASIA-B gait dataset and OUMVLP-Pose dataset show that our method outperforms most existing skeleton-based methods, verifying the effectiveness of the proposed modules.

DSA: Deformable Segmentation Attention for Multi-Scale Fisheye Image Segmentation

With a larger field of view (FOV) than ordinary images, fisheye images are becoming mainstream in the field of autonomous driving. However, the severe distortion problem of fisheye images also limits its application. The performance of neural networks designed for narrow FOV images degrades drastically for fisheye images, and the use of large composite models can improve the performance, but it brings huge time overhead and hardware costs. Therefore, we decided to balance real time and accuracy by designing the deformable segmentation attention(DSA) module, a generalpurpose architecture based on a deformable attention mechanism and a spatial pyramid architecture. The deformable mechanism serves to accurately extract feature information from fisheye images, together with attention to learn the global context and the spatial pyramid structure to balance multiscale feature information, thus improving the perception of fisheye images by traditional networks without increasing the amount of excessive computation. Lightweight networks such as SegNeXt equipped with the DSA module enable effective and rapid multi-scale segmentation of fisheye images in complex scenes. Our architecture achieves outstanding results on the WoodScape dataset, while our ablation experiments demonstrate the effectiveness of various parts of the architecture.

Using Time Series Sentinel Images for Object-Oriented Crop Extraction of Planting Structure in the Google Earth Engine

Timely and accurate extraction of crop planting structure information is of great importance for food security and sustainable agricultural development. However, long time series data with high spatial resolution have a much larger data volume, which seriously limits the quality and efficiency of the application of remote sensing to agriculture in complex crop rotation areas. To address this problem, this paper takes Lujiang County, a typical complex crop rotation region in the middle and lower reaches of the Yangtze River in China as an example, and proposes utilizing the Google Earth Engine (GEE) platform to extract the Normalized Difference Vegetation Index (NDVI), Normalized Difference Yellowness Index (NDYI) and Vertical-Horizontal Polarization (VH) time series sets of the whole planting year, and combining the Simple Non-Iterative Clustering (SNIC) multi-scale segmentation with the Support Vector Machine (SVM) and Random Forest (RF) algorithms to realize the fast and high-quality planting information of the main crop rotation patterns in the complex rotation region. The results show that by combining time series and object-oriented methods, SVM leads to better improvement than RF, with its overall accuracy and Kappa coefficient increasing by 4.44% and 0.0612, respectively, but RF is more suitable for extracting the planting structure in complex crop rotation areas. The RF algorithm combined with time series object-oriented extraction (OB + T + RF) achieved the highest accuracy, with an overall accuracy and Kappa coefficient of 98.93% and 0.9854, respectively. When compared to the pixel-oriented approach combined with the Support Vector Machine algorithm based on multi-temporal data (PB + M + SVM), the proposed method effectively reduces the presence of salt-and-pepper noise in the results, resulting in an improvement of 6.14% in overall accuracy and 0.0846 in Kappa coefficient. The research results can provide a new idea and a reliable reference method for obtaining crop planting structure information efficiently and accurately in complex crop rotation areas.

A Deep Learning Network for Individual Tree Segmentation in UAV Images with a Coupled CSPNet and Attention Mechanism

Accurate individual tree detection by unmanned aerial vehicles (UAVs) is a critical technique for smart forest management and serves as the foundation for evaluating ecological functions. Existing object detection and segmentation methods, on the other hand, have reduced accuracy when detecting and segmenting individual trees in complicated urban forest landscapes, as well as poor mask segmentation quality. This study proposes a novel Mask-CSP-attention-coupled network (MCAN) based on the Mask R-CNN algorithm. MCAN uses the Cross Stage Partial Net (CSPNet) framework with the Sigmoid Linear Unit (SiLU) activation function in the backbone network to form a new Cross Stage Partial Residual Net (CSPResNet) and employs a convolutional block attention module (CBAM) mechanism to the feature pyramid network (FPN) for feature fusion and multiscale segmentation to further improve the feature extraction ability of the model, enhance its detail information detection ability, and improve its individual tree detection accuracy. In this study, aerial photography of the study area was conducted by UAVs, and the acquired images were used to produce a dataset for training and validation. The method was compared with the Mask Region-based Convolutional Neural Network (Mask R-CNN), Faster Region-based Convolutional Neural Network (Faster R-CNN), and You Only Look Once v5 (YOLOv5) on the test set. In addition, four scenes—namely, a dense forest distribution, building forest intersection, street trees, and active plaza vegetation—were set up, and the improved segmentation network was used to perform individual tree segmentation on these scenes to test the large-scale segmentation ability of the model. MCAN’s average precision (AP) value for individual tree identification is 92.40%, which is 3.7%, 3.84%, and 12.53% better than that of Mask R-CNN, Faster R-CNN, and YOLOv5, respectively. In comparison to Mask R-CNN, the segmentation AP value is 97.70%, an increase of 8.9%. The segmentation network’s precision for the four scenes in multi-scene segmentation ranges from 95.55% to 92.33%, showing that the proposed network performs high-precision segmentation in many contexts.

Systematic assessment method for post-earthquake damage of regional buildings using adaptive-network-based fuzzy inference system

Rapid damage assessment of regional buildings is critical in post-earthquake rescue and victim resettlement. In this study, a novel rapid damage assessment method for post-earthquake buildings based on multi-types of damage features was proposed. First, the roof-related damage features are extracted from UAV tilt photography with the method of multiscale segmentation and object-oriented classification. Then, the normalized digital surface model was introduced to determine the height-related damage features. Meanwhile, a lightweight CNN model was employed to realize the preliminary evaluation of building facade conditions. Finally, all obtained damage features were taken as inputs to an adaptive-network-based fuzzy inference system for damage assessment. The proposed method was characterized by the capability to comprehensively cover the damage features of regional buildings involved in the evaluation, and thus be able to obtain more accurate judgments than current methods based on a single feature. At the same time, the effectiveness and robustness of the proposed method were evaluated and demonstrated by the field test on the world's largest earthquake site in Beichuan, China, and the prediction of damage level on the testing set achieved an accuracy of 87.1%.

Research on the identification of land types and tree species in the Engebei ecological demonstration area based on GF-1 remote sensing

Identifying forest types is crucial in satellite remote sensing monitoring. Research focusing on the identification of forest tree species using high-resolution remote sensing images is on the rise. Traditional pixel-based classification methods have not been successful in fully utilizing the rich spatial and texture feature information present in the image. Further, these methods are plagued by “classification noise”, which impedes the precise extraction of information regarding forest tree species. The object-oriented classification method, on the other hand, offers an extraction methodology rooted in image segmentation object features. It efficiently leverages the spectral, texture, and spatial geometry information of high-resolution remote sensing images. In this study, an initial application of the object-oriented, multi-level information extraction was conducted on ground objects within the Engebei ecological demonstration area. This was performed utilizing Gaofen (GF)-1 remote sensing images. Following this, a multi-scale segmentation algorithm was utilized to establish different segmentation scales relative to the features of the different objects. The optimal segmentation scale, shape factor, and compactness factor were determined to be 222/0.5/0.5, 167/0.5/0.6, and 81/0.4/0.6, respectively, for the three levels. Subsequently, training samples were used to select optimal features during the classification process. The nearest neighbor classification and random forest classification were performed respectively. The results indicate an overall classification accuracy of 88.58% and 87.26% for the nearest neighbor classification of remote sensing images in 2020 and 2021 respectively. Meanwhile, the overall classification accuracy for the random forest classification was 92.95% and 92.02% respectively. The Kappa coefficients for the nearest neighbor classification were 0.86 and 0.85, whereas those for the random forest classification were 0.92 and 0.90. These outcomes underscore the utility of the object-oriented classification method in enhancing the accuracy of forest type identification.

Dilated residual FPN-based segmentation for mouse retinal images

Background and objectiveDiabetes can induce diabetic retinopathy (DR), and the blindness caused by this disease is irreversible. The early analysis of mouse retinal images, including the layer and cell segmentation properties of these images, can help to effectively diagnose this disease. MethodIn the study, we design a dilated residual method based on a feature pyramid network (FPN), in which the FPN is adopted as the base network for solving the multiscale segmentation problem concerning mouse retinal images. In the bottom-up encoding pathway, we construct our backbone feature extraction network via the combination of dilated convolution and a residual block, further increasing the range of the receptive field to obtain more context information. At the same time, we integrate a squeeze-and-excitation (SE) attention module into the backbone network to obtain more small object details. In the top-down decoding pathway, we replace the traditional nearest-neighbor upsampling method with the transposed convolution method and add a segmentation head to obtain semantic segmentation results. ResultsThe effectiveness of our network model is verified in two segmentation tasks: ganglion cell segmentation and mouse retinal cell and layer segmentation. The outcomes demonstrate that, compared to other supervised segmentation methods based on deep learning, our model attains the utmost precision in both binary segmentation and multiclass semantic segmentation tasks. ConclusionThe dilated residual FPN is a robust method for mouse retinal image segmentation and it can effectively assist DR diagnosis.

Semi-supervised segmentation of multi-scale soil pores based on a novel receptive field structure

The study of soil pores using computed tomography (CT) technology and deep learning has been proven to be effective. However, conventional neural networks struggle to accurately segment both large connected pores and small scatter pores in soil CT images at different scales. To address this limitation, this paper proposes a semi-supervised multi-scale segmentation method (SMS). The SMS method comprises two parts: 1) Upstream multi-scale pore segmentation branches that generate multi-feature pore segmentation maps based on a novel receptive field structure, and 2) Downstream semi-supervised classification task that selects the optimal pore feature map for output. Experimental results demonstrate that SMS effectively enhances segmentation performance for multi-scale soil pores, achieving the highest accuracy, precision, recall, and F1-score values of 99.55 %, 83.76 %, 96.12 %, and 89.35 %, respectively. These results outperform four common deep learning methods and three traditional image processing software. This study represents a significant advancement in high-precision automated segmentation of soil pores, providing valuable image processing support for investigating soil structure characteristics, soil conservation, and soil ecosystem services.

Mountain Segmentation Based on Global Optimization with the Cloth Simulation Constraint

Mountains are an important research object for surveying, mapping, cartography, space science, and ecological remote sensing. Automatic mountain segmentation is one of the most critical techniques in large-scale mountain analyses. However, several factors limit the segmentation accuracy, such as the complexity of mountains, the noise of geospatial data, and the confusion in distinguishing non-mountainous objects with similar features. In order to improve the accuracy of mountain segmentation against these limiting factors, we impose the cloth constraint over the digital elevation model (DEM) with the underlying assumption that the mountain has a sizeable relative elevation and slope. We propose a robust mountain segmentation method with the cloth simulation constraint. The core algorithm extracts the relative elevation of the region using a cloth simulation filtering algorithm by transforming the mountain segmentation problem into an optimization problem based on the global energy function consisting of the relative elevation and slope. Experiments on a wide range of Earth and lunar elevation datasets with some of the aforementioned limitations show that the proposed method can extract complex mountain baselines, avoid the misclassification of lunar craters, and significantly improve the robustness and accuracy of mountain segmentation. Compared to three state-of-the-art methods (the Lunar Mountain Detection Method, the Landform Mask Method in SNAP™ from European Space Agency (located in Paris, France), and the Multiscale Segmentation Method in eCognition™ from Definiens Imaging (located in Munich, Germany), the F1 and IoU improved by 14.70% and 20.46% on average and 29.07% and 38.94% at most, respectively, which validates that the proposed method has a better all-around performance.

Bi-Temporal change detection of high-resolution images by referencing time series medium-resolution images

ABSTRACT Seasonal changes usually exist and cause false alarms in the bi-temporal change detection from high-resolution remote sensing images. It is difficult to remove these false alarms only using bi-temporal images for traditional change detection methods. A change detection method is proposed to remove seasonal false alarms in bi-temporal change detection by introducing time series information of medium-resolution remote sensing images. First, the mid-resolution time series results are mapped to the ground objects obtained by multiscale segmentation of high-resolution remote sensing images. Second, set the thresholds for the proportion of each category of pixels in the object to obtain high-resolution time series results. Finally, the high-resolution change detection results are optimized by the improved high-resolution time series results. Experimental results show that this method can optimize the results of high-resolution change detection, and the accuracy of this method was improved by at least 0.23 than that of traditional change detection by reducing seasonal errors. The proposed method was an effective change detection approach for high-resolution images to reduce detection errors due to seasonal differences.

Consensus Techniques for Unsupervised Binary Change Detection Using Multi-Scale Segmentation Detectors for Land Cover Vegetation Images

Change detection in very-high-spatial-resolution (VHR) remote sensing images is a very challenging area with applicability in many problems ranging from damage assessment to land management and environmental monitoring. In this study, we investigated the change detection problem associated with analysing the vegetation corresponding to crops and natural ecosystems over VHR multispectral and hyperspectral images obtained by sensors onboard drones or satellites. The challenge of applying change detection methods to these images is the similar spectral signatures of the vegetation elements in the image. To solve this issue, a consensus multi-scale binary change detection technique based on the extraction of object-based features was developed. With the objective of capturing changes at different granularity levels taking advantage of the high spatial resolution of the VHR images and, as the segmentation operation is not well defined, we propose to use several detectors based on different segmentation algorithms, each applied at different scales. As the changes in vegetation also present high variability depending on capture conditions such as illumination, the use of the CVA-SAM applied at the segment level instead of at the pixel level is also proposed. The results revealed the effectiveness of the proposed approach for identifying changes over land cover vegetation images with different types of changes and different spatial and spectral resolutions.