Figure 1 - available via license: Creative Commons Attribution 3.0 Unported
Content may be subject to copyright.
Source publication
This paper presents a new unsupervised classification method which aims to effectively and efficiently map remote sensing data. The
Mean-Shift (MS) algorithm, a non parametric density-based clustering technique, is at the core of our method. This powerful clustering
algorithm has been successfully used for both the classification and the segmentati...
Similar publications
Automatic security inspection has received increasing interests in recent years. Due to the fixed top-down perspective of X-ray scanning of often tightly packed luggages, such images typically suffer from penetration-induced occlusions, severe object overlapping and violent changes in appearance. For this particular application, few research effort...
Automatic security inspection has received increasing interests in recent years. Due to the fixed top-down perspective of X-ray scanning of often tightly packed luggages, such images typically suffer from penetration-induced occlusions, severe object overlapping and violent changes in appearance. For this particular application, few research effort...
Citations
... This was performed by applying a multi-threaded Mean-Shift algorithm on the 2008 aerial imagery. This algorithm delineates clusters within a dataset by shifting each data point toward a local mode in the kernel density estimation of the feature space (Boukir et al., 2012). Boukir et al. (2012) note that an important parameter of the Mean-Shift algorithm is the radiometric range as it refers to a unique spectral radius algorithm's kernel. ...
... This algorithm delineates clusters within a dataset by shifting each data point toward a local mode in the kernel density estimation of the feature space (Boukir et al., 2012). Boukir et al. (2012) note that an important parameter of the Mean-Shift algorithm is the radiometric range as it refers to a unique spectral radius algorithm's kernel. Further, they add that the size of the smallest image object (or the minimum mapping unit) can influence the image segmentation result. ...
... This was performed by applying a multi-threaded Mean-Shift algorithm on the 2008 aerial imagery. This algorithm delineates clusters within a dataset by shifting each data point toward a local mode in the kernel density estimation of the feature space (Boukir et al., 2012). Boukir et al. (2012) note that an important parameter of the Mean-Shift algorithm is the radiometric range as it refers to a unique spectral radius algorithm's kernel. ...
... This algorithm delineates clusters within a dataset by shifting each data point toward a local mode in the kernel density estimation of the feature space (Boukir et al., 2012). Boukir et al. (2012) note that an important parameter of the Mean-Shift algorithm is the radiometric range as it refers to a unique spectral radius algorithm's kernel. Further, they add that the size of the smallest image object (or the minimum mapping unit) can influence the image segmentation result. ...
... Specifically, if this range distance is below the range radius, the pixels are grouped into the same cluster. The MS algorithm does not require prior knowledge of the number and shape of the clusters (Boukir, Jones, and Reinke 2012), so the best segmentation parameters (i.e. Spatial Radius (hs) equal to 4 pixels, Range Radius (hr) of 500 and 15 pixels as Minimum size (ms)) were selected by visual evaluation using a trial-and-error approach of the alignment between the shape of the polygons generated by segmentation and the boundaries identified in the image (Mathieu, Aryal, and Chong 2007). ...
Poplars are one of the most widespread fast-growing tree species used for forest plantations. Owing to their distinct features (fast growth and short rotation) and the dependency on the timber price market, poplar plantations are characterized by large inter-annual fluctuations in their extent and distribution. Therefore, monitoring poplar plantations requires a frequent update of information – not feasible by National Forest Inventories due to their periodicity – achievable by remote sensing systems applications. In particular, the new Sentinel-2 mission, with a revisiting period of 5 days, represents a potentially efficient tool for meeting this need.
In this paper, we present a deep learning approach for mapping poplar plantations using Sentinel-2 time series. A reference dataset of poplar plantations was available for a large study area of more than 46,000 km² in Northern Italy and served as training and testing data. Two classification methods were compared: (1) a fully connected neural network (also called multilayer perceptron), and (2) a traditional logistic regression. The performance of the two approaches was estimated through bootstrapping procedure with a confidence interval of 99%. Results indicated for deep learning an omission error rate of 2.77%±2.76%, showing improvements compared to logistic regression, omission error rate = 8.91%±4.79%.
... By 2015, a superior segmentation algorithm, Segment Mean Shift (SMS), was available within ArcMap 10.3 and this segmentation routine was applied during the 1st phase of workflow deployment over the Firebag AOI. Segment Mean Shift provides excellent results in clustering and object delineation (Boukir et al., 2012), but only accepts three bands. To provide maximum image information in three bands, Tasseled Cap (TC) transformations (Kauth and Thomas, 1976) were performed on the QB2, WV2, and WV3 image surface-reflectance data. ...
To achieve reclamation certification, oil-and-gas operations in Alberta, Canada are required to monitor the revegetation of idle well pads that no longer support operations. Currently, monitoring is completed by oblique, helicopter-collected photography and on-the-ground field surveys. Both monitoring strategies present safety and logistical challenges. To mitigate these challenges, a remote-sensing project was completed to develop and deploy a reproducible workflow using high-spatial-resolution satellite imagery to monitor revegetation progress on idle well pads. Seven well pads in the Aspen region of Alberta, Canada were selected for workflow development, using imagery from 2007, 2009, and 2011. Land-cover classes were derived from the satellite imagery using a training dataset, a series of vegetation indices derived from the satellite imagery, and regression tree classification programs, and were used to evaluate changes in vegetation cover over time. A refined version of this general workflow was then deployed across 39 well pads in the Firebag region of Alberta, Canada, using imagery from 2010 to 2016. In 2016, fieldwork was conducted across a subset of 16 well pads in the Firebag region, which facilitated a formal accuracy assessment of the land-cover classifications. This project demonstrated that high-spatial-resolution satellite imagery could be used to develop accurate land-cover classifications on these relatively small landscape features and that temporal land-cover classifications could be used to track revegetation through time. Overall, these results show the feasibility of remote-sensing-based workflows in monitoring revegetation on idle well pads. Copyright © 2017. The American Association of Petroleum Geologists/Division of Environmental Geosciences. All rights reserved.
... Additional trial-and-error tests (not discussed here) were conducted to assess the optimal settings for hs, hr, and ms. For the test site Munich South, following Boukir et al. [45], we decided to set hs to half the standard deviation (SD) of all image digital numbers and for radiometric radius hr a quarter of the SD was used. Minimum object size (ms) was set to 10 pixels (250 m 2 ), representing a small group of trees (see Table 7). ...
With climate change, extreme storms are expected to occur more frequently. These storms can cause severe forest damage, provoking direct and indirect economic losses for forestry. To minimize economic losses, the windthrow areas need to be detected fast to prevent subsequent biotic damage, for example, related to beetle infestations. Remote sensing is an efficient tool with high potential to cost-efficiently map large storm affected regions. Storm Niklas hit South Germany in March 2015 and caused widespread forest cover loss. We present a two-step change detection approach applying commercial very high-resolution optical Earth Observation data to spot forest damage. First, an object-based bi-temporal change analysis is carried out to identify windthrow areas larger than 0.5 ha. For this purpose, a supervised Random Forest classifier is used, including a semi-automatic feature selection procedure; for image segmentation, the large-scale mean shift algorithm was chosen. Input features include spectral characteristics, texture, vegetation indices, layer combinations and spectral transformations. A hybrid-change detection approach at pixel-level subsequently identifies small groups of fallen trees, combining the most important features of the previous processing step with Spectral Angle Mapper and Multivariate Alteration Detection. The methodology was evaluated on two test sites in Bavaria with RapidEye data at 5 m pixel resolution. The results regarding windthrow areas larger than 0.5 ha were validated with reference data from field visits and acquired through orthophoto interpretation. For the two test sites, the novel object-based change detection approach identified over 90% of the windthrow areas (≥0.5 ha). The red edge channel was the most important for windthrow identification. Accuracy levels of the change detection at tree level could not be calculated, as it was not possible to collect field data for single trees, nor was it possible to perform an orthophoto validation. Nevertheless, the plausibility and applicability of the pixel-based approach is demonstrated on a second test site.
... The window between these two points is moved to the mean value. The process is iterative and ends when the final clustering is given (Boukir & Reinke, 2012). Finally, the Spectral Clustering requires the number of clusters and it uses the similarity matrix between data samples to compute the best eigenvalues for posterior clustering (Liao, 2005). ...
... Les données de référence (limitées aussi, voir section 7.3.2) utilisées pour valider l'approche proposée ne sont pas comptabilisées, d'autant plus que d'autres moyens que la matrice de confusion sont utilisés pourévaluer les approches non supervisées (Boukir et al., 2012). ...
... En outre, l'algorithme du mean shift,à la base de notre approche de détection de changementsChehata et al., 2011), aété peu mis en oeuvre pour des travaux de cartographie forestière. Des travaux relatifs sont certes apparus récemment(Boukir et al., 2012;Taud et al., 2012) mais ils concernent des approches pixel et non région comme notre méthode qui implique le mean shiftà deux niveaux : segmentation puis cartographie. Les performances de cet algorithme puissant sur des images de forêt, aussi bien dans un contexte de segmentation que de classification multispectrale, seront analysées dans la section 7. ...
... It is a powerful unsupervised data analysis technique which does not require prior knowledge of the number of clusters, and does not constrain the shape of the clusters. Comaniciu et al. [5] and Boukir et al. [6] have done studies on Mean shift and have put forward the advantages of using Mean Shift as a segmentation process in the remote sensing images. ...
... Boukir et al. [6] have used a fast mean shift algorithm called path assigned mean shift, which is said to be 5 times faster than original mean shift. The method uses mean shift segmentation first to segment the image then as a second refinement used k-means algorithm to converge to a refined solution. ...
