Pub Date : 2018-08-01DOI: 10.1109/PRRS.2018.8486408
A. Nikonorov, M. Petrov, Sergei Bibikov, V. Kutikova, P. Yakimov, A. Morozov, R. Skidanov, N. Kazanskiy
The paper addresses the problem of imaging quality enhancement for the Offner hyperspectrometer using a convolutional neural network. We use a deep convolutional neural network with residual training and PReLU activation, inspired by the super-resolution task for RGB images. In the case of hyperspectral imaging, it is often a problem to find a large enough ground truth dataset for training a neural network from scratch. Transfer learning using the network pretrained for RGB images with some pre- and postprocessing is one of the possible workarounds. In this paper, we propose to simulate the necessary ground truth data using non-imaging spectrometer. The obtained dataset with partially simulated ground truth is then used to train the convolutional neural network directly for hyperspectral image quality enhancement. The proposed training approach also allows to incorporate distortions specific for hyperspectral images into the enhancement procedure. It allows to successfully remove the striping distortions inherent to the Offner scheme of image acquisition. The experimental results of the proposed approach show a significant quality gain.
{"title":"Deep Learning-Based Enhancement of Hyperspectral Images Using Simulated Ground Truth","authors":"A. Nikonorov, M. Petrov, Sergei Bibikov, V. Kutikova, P. Yakimov, A. Morozov, R. Skidanov, N. Kazanskiy","doi":"10.1109/PRRS.2018.8486408","DOIUrl":"https://doi.org/10.1109/PRRS.2018.8486408","url":null,"abstract":"The paper addresses the problem of imaging quality enhancement for the Offner hyperspectrometer using a convolutional neural network. We use a deep convolutional neural network with residual training and PReLU activation, inspired by the super-resolution task for RGB images. In the case of hyperspectral imaging, it is often a problem to find a large enough ground truth dataset for training a neural network from scratch. Transfer learning using the network pretrained for RGB images with some pre- and postprocessing is one of the possible workarounds. In this paper, we propose to simulate the necessary ground truth data using non-imaging spectrometer. The obtained dataset with partially simulated ground truth is then used to train the convolutional neural network directly for hyperspectral image quality enhancement. The proposed training approach also allows to incorporate distortions specific for hyperspectral images into the enhancement procedure. It allows to successfully remove the striping distortions inherent to the Offner scheme of image acquisition. The experimental results of the proposed approach show a significant quality gain.","PeriodicalId":197319,"journal":{"name":"2018 10th IAPR Workshop on Pattern Recognition in Remote Sensing (PRRS)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2018-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127644935","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2018-08-01DOI: 10.1109/PRRS44410.2018.9396733
Farah Jahan, Jun Zhou, M. Awrangjeb, Yongsheng Gao
Exploiting the multi-source data is an effective but challenging problem for land cover classification. Popular remote sensor data, e.g., hyperspectral (HS) and light detection and ranging (LiDAR), contain complementary information for land cover if they are co-registered. In this paper, we aim to integrate information extracted from these data sources for land cover classification. At first, we propose a novel feature extraction method by calculating the inverse coefficient of variation (ICV) using the Gaussian probability of neighbourhood between every pair of bands in HS data. This is calculated for each band with respect to every other band to form an ICV cube. We reduce the number of planes in the cube by applying principal component analysis (PCA) on it and spatial features are then extracted for significant principal components. The spectral information from HS data, their ICV responses, and spatial information from ICV responses have complementary information; that is why we fuse them together by layer stacking to generate discriminant features. Secondly, we also derive height and spatial features from LiDAR Digital Elevation Model (DSM), which are later concatenated with the HS derived features. Finally, these features are classified using linear discriminant analysis (LDA) classifier. The classification results prove the effectiveness of the derived features from both data sources.
{"title":"Integration of Heterogeneous Features from Co-registered Hyperspectral and LiDAR Data for Land Cover Classification","authors":"Farah Jahan, Jun Zhou, M. Awrangjeb, Yongsheng Gao","doi":"10.1109/PRRS44410.2018.9396733","DOIUrl":"https://doi.org/10.1109/PRRS44410.2018.9396733","url":null,"abstract":"Exploiting the multi-source data is an effective but challenging problem for land cover classification. Popular remote sensor data, e.g., hyperspectral (HS) and light detection and ranging (LiDAR), contain complementary information for land cover if they are co-registered. In this paper, we aim to integrate information extracted from these data sources for land cover classification. At first, we propose a novel feature extraction method by calculating the inverse coefficient of variation (ICV) using the Gaussian probability of neighbourhood between every pair of bands in HS data. This is calculated for each band with respect to every other band to form an ICV cube. We reduce the number of planes in the cube by applying principal component analysis (PCA) on it and spatial features are then extracted for significant principal components. The spectral information from HS data, their ICV responses, and spatial information from ICV responses have complementary information; that is why we fuse them together by layer stacking to generate discriminant features. Secondly, we also derive height and spatial features from LiDAR Digital Elevation Model (DSM), which are later concatenated with the HS derived features. Finally, these features are classified using linear discriminant analysis (LDA) classifier. The classification results prove the effectiveness of the derived features from both data sources.","PeriodicalId":197319,"journal":{"name":"2018 10th IAPR Workshop on Pattern Recognition in Remote Sensing (PRRS)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2018-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126092497","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2018-08-01DOI: 10.1109/PRRS.2018.8486388
Xuchun Li, Huimin Xu, Fushan Zhang, A. Suo
In this paper, domestic GF-1 satellite remote sensing imagery is used to analyze the internal spatial pattern of the port and the characteristics of its constituent elements, and an object-oriented remote sensing monitoring method and process for port space pattern is established. The dock shoreline index and the dock coastline utilization index are explored and constructed. Dock index, storage yard index and dock basin index were used to evaluate the intensive use of port space patterns, and an empirical study was conducted in the Yingkou Bayuquan port area. The results showed that the dock shoreline index of Yingkou Bayuquan Port area was 0.51, the dock coastline utilization index was 15.08 million tons/km, the dock index was 12.23 hm2/km, the dock basin index was 242.76 hm2/km, and the storage yard index was 108.46 hm2/km. The utilization index of the dock coastline is basically 150 million tons/km, and the basic ratio of the dock and dock area, storage yard area and dock basin area is 1.00:12.00:108.00:250.00. Yingkou Bayuquan Port still has a potential of 88.21 million tons of throughput per year.
{"title":"Monitoring and Assessment of the Space Pattern of Ports Based on GF-1 Satellite Remote-Sensing Images","authors":"Xuchun Li, Huimin Xu, Fushan Zhang, A. Suo","doi":"10.1109/PRRS.2018.8486388","DOIUrl":"https://doi.org/10.1109/PRRS.2018.8486388","url":null,"abstract":"In this paper, domestic GF-1 satellite remote sensing imagery is used to analyze the internal spatial pattern of the port and the characteristics of its constituent elements, and an object-oriented remote sensing monitoring method and process for port space pattern is established. The dock shoreline index and the dock coastline utilization index are explored and constructed. Dock index, storage yard index and dock basin index were used to evaluate the intensive use of port space patterns, and an empirical study was conducted in the Yingkou Bayuquan port area. The results showed that the dock shoreline index of Yingkou Bayuquan Port area was 0.51, the dock coastline utilization index was 15.08 million tons/km, the dock index was 12.23 hm2/km, the dock basin index was 242.76 hm2/km, and the storage yard index was 108.46 hm2/km. The utilization index of the dock coastline is basically 150 million tons/km, and the basic ratio of the dock and dock area, storage yard area and dock basin area is 1.00:12.00:108.00:250.00. Yingkou Bayuquan Port still has a potential of 88.21 million tons of throughput per year.","PeriodicalId":197319,"journal":{"name":"2018 10th IAPR Workshop on Pattern Recognition in Remote Sensing (PRRS)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2018-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126116483","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2018-08-01DOI: 10.1109/PRRS.2018.8486185
Wei Yujun, Xiangyun Hu, Gong Jinqi
Road extraction from aerial and satellite image is one of complex and challenging tasks in remote sensing field. The task is required for a wide range of application, such as autonomous driving, urban planning and automatic mapping for GIS data collection. Most approaches cast the road extraction as image segmentation and use thinning algorithm to get road centerline. However, these methods can easily produce spurs around the true centerline which affects the accuracy of road centerline extraction and lacks the topology of road network. In this paper, we propose a novel method to directly extract accurate road centerline from aerial images and construct the topology of the road network. First, an end-to-end regression network based on convolutional neural network is designed to learn and predict a road centerline confidence map which is a 2D representation of the probability of each pixel to be on the road centerline. Our network combines multi-scale and multi-level feature information to produce refined confidence map. Then a canny-like non-maximum suppression is followed to attain accurate road centerline. Finally, we use spoke wheel to find the road direction of the initialized road center point and take advantage of road tracking to construct the topology of road network. The results on the Massachusetts Road dataset shows an significant improvement on the accuracy of location of extracted road centerline.
{"title":"End-to-End Road Centerline Extraction via Learning a Confidence Map","authors":"Wei Yujun, Xiangyun Hu, Gong Jinqi","doi":"10.1109/PRRS.2018.8486185","DOIUrl":"https://doi.org/10.1109/PRRS.2018.8486185","url":null,"abstract":"Road extraction from aerial and satellite image is one of complex and challenging tasks in remote sensing field. The task is required for a wide range of application, such as autonomous driving, urban planning and automatic mapping for GIS data collection. Most approaches cast the road extraction as image segmentation and use thinning algorithm to get road centerline. However, these methods can easily produce spurs around the true centerline which affects the accuracy of road centerline extraction and lacks the topology of road network. In this paper, we propose a novel method to directly extract accurate road centerline from aerial images and construct the topology of the road network. First, an end-to-end regression network based on convolutional neural network is designed to learn and predict a road centerline confidence map which is a 2D representation of the probability of each pixel to be on the road centerline. Our network combines multi-scale and multi-level feature information to produce refined confidence map. Then a canny-like non-maximum suppression is followed to attain accurate road centerline. Finally, we use spoke wheel to find the road direction of the initialized road center point and take advantage of road tracking to construct the topology of road network. The results on the Massachusetts Road dataset shows an significant improvement on the accuracy of location of extracted road centerline.","PeriodicalId":197319,"journal":{"name":"2018 10th IAPR Workshop on Pattern Recognition in Remote Sensing (PRRS)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2018-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114918935","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2018-08-01DOI: 10.1109/PRRS.2018.8486228
Qinghua Li, S. Ural, J. Shan
This paper applies pattern recognition methods to airborne lidar waveform decomposition. The parametric and nonparametric approaches are compared in the experiments. The popular Gaussian mixture model (GMM) and expectation-maximization (EM) decomposition algorithm are selected as the parametric approach. Nonparametric mixture model (NMM) and fuzzy mean-shift (FMS) are used as the nonparametric approach. We first run our experiment on simulated waveforms. The experiment setup is in favor of the parametric approach because GMM is used to generate the waveforms. We show that both parametric and nonparametric approaches return satisfying results on the simulated mixture of Gaussian components. In the second experiment, real data acquired with an airborne lidar are used. We find that NMM fits the data better than GMM because the Gaussian assumption is not well satisfied in the real dataset. Considering that the emitted signals of a laser scanner may even not satisfy the Gaussian assumption, we conclude that nonparametric approaches should generally be utilized for practical applications.
{"title":"A Comparative Study on Airborne Lidar Waveform Decomposition Methods","authors":"Qinghua Li, S. Ural, J. Shan","doi":"10.1109/PRRS.2018.8486228","DOIUrl":"https://doi.org/10.1109/PRRS.2018.8486228","url":null,"abstract":"This paper applies pattern recognition methods to airborne lidar waveform decomposition. The parametric and nonparametric approaches are compared in the experiments. The popular Gaussian mixture model (GMM) and expectation-maximization (EM) decomposition algorithm are selected as the parametric approach. Nonparametric mixture model (NMM) and fuzzy mean-shift (FMS) are used as the nonparametric approach. We first run our experiment on simulated waveforms. The experiment setup is in favor of the parametric approach because GMM is used to generate the waveforms. We show that both parametric and nonparametric approaches return satisfying results on the simulated mixture of Gaussian components. In the second experiment, real data acquired with an airborne lidar are used. We find that NMM fits the data better than GMM because the Gaussian assumption is not well satisfied in the real dataset. Considering that the emitted signals of a laser scanner may even not satisfy the Gaussian assumption, we conclude that nonparametric approaches should generally be utilized for practical applications.","PeriodicalId":197319,"journal":{"name":"2018 10th IAPR Workshop on Pattern Recognition in Remote Sensing (PRRS)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2018-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133734839","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2018-08-01DOI: 10.1109/PRRS.2018.8486168
Chuang Luo, Li Ma
We perform unsupervised domain adaptation for classification of remote sensing images by exploiting unsupervised manifold alignment approach. Manifold alignment method utilized corresponding points between domains to align data manifolds of source and target domains, where the corresponding points can be constructed by labeled information. Supposing labeled samples are not available in target domain, we proposed neighbor consistency (NC) constraint to select some target points that have reliable predictions. These points and labeled source data are then used to construct corresponding relationships, resulting in unsupervised manifold alignment. The neighbor consistency based unsupervised manifold alignment is denoted as NCUMA in this paper. Both multispectral and hyperspectral remote sensing data have been used to demonstrate the effectiveness of the NCUMA approach.
{"title":"Neighbor Consistency Baced Unsupervised Manifold Alignment for Classification of Remote Sensing Image","authors":"Chuang Luo, Li Ma","doi":"10.1109/PRRS.2018.8486168","DOIUrl":"https://doi.org/10.1109/PRRS.2018.8486168","url":null,"abstract":"We perform unsupervised domain adaptation for classification of remote sensing images by exploiting unsupervised manifold alignment approach. Manifold alignment method utilized corresponding points between domains to align data manifolds of source and target domains, where the corresponding points can be constructed by labeled information. Supposing labeled samples are not available in target domain, we proposed neighbor consistency (NC) constraint to select some target points that have reliable predictions. These points and labeled source data are then used to construct corresponding relationships, resulting in unsupervised manifold alignment. The neighbor consistency based unsupervised manifold alignment is denoted as NCUMA in this paper. Both multispectral and hyperspectral remote sensing data have been used to demonstrate the effectiveness of the NCUMA approach.","PeriodicalId":197319,"journal":{"name":"2018 10th IAPR Workshop on Pattern Recognition in Remote Sensing (PRRS)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2018-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116998237","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2018-08-01DOI: 10.1109/PRRS.2018.8486166
Tengping Jiang, Yongjun Wang, Shuaibing Tao, Yunli Li, Shan Liu
To alleviate the difficulties in obtaining training data sets of 3D point clouds, an active learning (AL) framework is proposed to iteratively select a small portion of unlabeled points to query their labels, and creates a minimum manually-annotated training set. To handle the biased sampling problem caused by category imbalance and local similarities, a neighbor-consistency prior is used to conduct an unbiased sampling for selecting the value samples into the training set. Additionally, to reduce the number of categories used in labeling, a higher-order MRF containing a regional label cost term, is exploited to refine the labeling results.
{"title":"Integrating Active Learning and Contextually Guide for Semantic Labeling of LiDAR Point Cloud","authors":"Tengping Jiang, Yongjun Wang, Shuaibing Tao, Yunli Li, Shan Liu","doi":"10.1109/PRRS.2018.8486166","DOIUrl":"https://doi.org/10.1109/PRRS.2018.8486166","url":null,"abstract":"To alleviate the difficulties in obtaining training data sets of 3D point clouds, an active learning (AL) framework is proposed to iteratively select a small portion of unlabeled points to query their labels, and creates a minimum manually-annotated training set. To handle the biased sampling problem caused by category imbalance and local similarities, a neighbor-consistency prior is used to conduct an unbiased sampling for selecting the value samples into the training set. Additionally, to reduce the number of categories used in labeling, a higher-order MRF containing a regional label cost term, is exploited to refine the labeling results.","PeriodicalId":197319,"journal":{"name":"2018 10th IAPR Workshop on Pattern Recognition in Remote Sensing (PRRS)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2018-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115432416","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2018-08-01DOI: 10.1109/PRRS.2018.8486220
Yusheng Xu, Zhenghao Sun, L. Hoegner, Uwe Stilla, W. Yao
In this paper, an instance segmentation method for tree extraction from MLS data sets in urban scenes is developed. The proposed method utilizes a supervoxel structure to organize the point clouds, and then extracts the detrended geometric features from the local context of supervoxels. Combined with the detrended features of the local context, the Random Forest (RF) classifier will be adopted to obtain the initial semantic labeling results of trees from point clouds. Afterwards, a local context-based regularization is iteratively performed to achieve global optimum on a global graphical model, in order to spatially smoothing the semantic labeling results. Finally, a graph-based segmentation is conducted to separate individual trees according to the semantic labeling results. The use of supervoxel structure can preserve the geometric boundaries of objects in the scene, and compared with point-based solutions, the supervoxel-based method can largely decrease the number of basic elements during the processing. Besides, the introduction of supervoxel contexts can extract the local information of an object making the feature extraction more robust and representative. Detrended geometric features can get over the redundant and in-salient information in the local context, so that discriminative features are obtained. Benefiting from the regularization process, the spatial smoothing is obtained based on initial labeling results from classic classifications such as RF classification. As a result, misclassification errors are removed to a large degree and semantic labeling results are thus smoothed. Based on the constructed global graphical model during the spatially smoothing process, a graph-based segmentation is applied to partition the graphical model for the clustering the instances of trees. The experiments on two test datasets have shown promising results, with an accuracy of the semantic labeling of trees reaching around 0.9. The segmentation of trees using graph-based algorithm also show acceptable results, with trees having simple structures and sparse distributions correctly separated, but for those cramped trees with complex structures, the points are over- or under-segmented.
{"title":"Instance Segmentation of Trees in Urban Areas from MLS Point Clouds Using Supervoxel Contexts and Graph-Based Optimization","authors":"Yusheng Xu, Zhenghao Sun, L. Hoegner, Uwe Stilla, W. Yao","doi":"10.1109/PRRS.2018.8486220","DOIUrl":"https://doi.org/10.1109/PRRS.2018.8486220","url":null,"abstract":"In this paper, an instance segmentation method for tree extraction from MLS data sets in urban scenes is developed. The proposed method utilizes a supervoxel structure to organize the point clouds, and then extracts the detrended geometric features from the local context of supervoxels. Combined with the detrended features of the local context, the Random Forest (RF) classifier will be adopted to obtain the initial semantic labeling results of trees from point clouds. Afterwards, a local context-based regularization is iteratively performed to achieve global optimum on a global graphical model, in order to spatially smoothing the semantic labeling results. Finally, a graph-based segmentation is conducted to separate individual trees according to the semantic labeling results. The use of supervoxel structure can preserve the geometric boundaries of objects in the scene, and compared with point-based solutions, the supervoxel-based method can largely decrease the number of basic elements during the processing. Besides, the introduction of supervoxel contexts can extract the local information of an object making the feature extraction more robust and representative. Detrended geometric features can get over the redundant and in-salient information in the local context, so that discriminative features are obtained. Benefiting from the regularization process, the spatial smoothing is obtained based on initial labeling results from classic classifications such as RF classification. As a result, misclassification errors are removed to a large degree and semantic labeling results are thus smoothed. Based on the constructed global graphical model during the spatially smoothing process, a graph-based segmentation is applied to partition the graphical model for the clustering the instances of trees. The experiments on two test datasets have shown promising results, with an accuracy of the semantic labeling of trees reaching around 0.9. The segmentation of trees using graph-based algorithm also show acceptable results, with trees having simple structures and sparse distributions correctly separated, but for those cramped trees with complex structures, the points are over- or under-segmented.","PeriodicalId":197319,"journal":{"name":"2018 10th IAPR Workshop on Pattern Recognition in Remote Sensing (PRRS)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2018-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114851195","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2018-08-01DOI: 10.1109/PRRS.2018.8486355
Mahdi Hasanlau, S. T. Seydi
This paper investigated the performance of different automatic binary threshold selection methods on hyperspectral change detection. For this purpose, 10 recent and most common algorithm for binary threshold selection implemented namely and evaluated. To evaluate these methods first, the sub-space based hyperspectral change detection method applied on the multi-temporal hyperspectral dataset. In the second part, the gray level change map converts to binary change map by mentioned thresholding methods. In this work, real-world hyperspectral dataset utilized to evaluate the related performance of threshold selection methods. The results show that Active-Contour method has high efficiency in comparison to other methods with overall accuracy more than 93.53% and a kappa coefficient of 0.851.
{"title":"Sensitivity Analysis on Performance of Different Unsupervised Threshold Selection Methods in Hyperspectral Change Detection","authors":"Mahdi Hasanlau, S. T. Seydi","doi":"10.1109/PRRS.2018.8486355","DOIUrl":"https://doi.org/10.1109/PRRS.2018.8486355","url":null,"abstract":"This paper investigated the performance of different automatic binary threshold selection methods on hyperspectral change detection. For this purpose, 10 recent and most common algorithm for binary threshold selection implemented namely and evaluated. To evaluate these methods first, the sub-space based hyperspectral change detection method applied on the multi-temporal hyperspectral dataset. In the second part, the gray level change map converts to binary change map by mentioned thresholding methods. In this work, real-world hyperspectral dataset utilized to evaluate the related performance of threshold selection methods. The results show that Active-Contour method has high efficiency in comparison to other methods with overall accuracy more than 93.53% and a kappa coefficient of 0.851.","PeriodicalId":197319,"journal":{"name":"2018 10th IAPR Workshop on Pattern Recognition in Remote Sensing (PRRS)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2018-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129640519","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2018-08-01DOI: 10.1109/PRRS.2018.8486236
Csaba Beleznai, Daniel Steininger, G. Croonen, Elisabeth Broneder
Recognizing humans from aerial views represents an increasingly relevant endeavor; a trend mainly driven by the widespread use of unmanned aerial vehicles (UAVs). An accurate and real-time visual human recognition task, however, represents a scientific challenge because typical UAV imaging and computational capabilities and conditions introduce complexities and constraints. Motion blur, the non-specific top-view appearance of humans, low-image resolution and limited onboard computational resources are among the most important limiting factors to be considered. In this paper we propose a run-time-efficient multi-modal detection framework performing clustering and recognition on thermal infrared, passive stereo depth and intensity channels in order to cope with the above complexities and to achieve accurate human detection results. Thermal infrared and depth data are used to generate proposals in combination with an explicit, tree-structured shape representation driven clustering scheme. Generated proposals are used as an input for a discriminatively trained deep classification step to recognize humans. The proposed clustering and classification scheme is validated in qualitative and quantitative terms on four large aerial datasets representing complex (small objects, clutter, occlusions) situations.
{"title":"Multi-Modal Human Detection from Aerial Views by Fast Shape-Aware Clustering and Classification","authors":"Csaba Beleznai, Daniel Steininger, G. Croonen, Elisabeth Broneder","doi":"10.1109/PRRS.2018.8486236","DOIUrl":"https://doi.org/10.1109/PRRS.2018.8486236","url":null,"abstract":"Recognizing humans from aerial views represents an increasingly relevant endeavor; a trend mainly driven by the widespread use of unmanned aerial vehicles (UAVs). An accurate and real-time visual human recognition task, however, represents a scientific challenge because typical UAV imaging and computational capabilities and conditions introduce complexities and constraints. Motion blur, the non-specific top-view appearance of humans, low-image resolution and limited onboard computational resources are among the most important limiting factors to be considered. In this paper we propose a run-time-efficient multi-modal detection framework performing clustering and recognition on thermal infrared, passive stereo depth and intensity channels in order to cope with the above complexities and to achieve accurate human detection results. Thermal infrared and depth data are used to generate proposals in combination with an explicit, tree-structured shape representation driven clustering scheme. Generated proposals are used as an input for a discriminatively trained deep classification step to recognize humans. The proposed clustering and classification scheme is validated in qualitative and quantitative terms on four large aerial datasets representing complex (small objects, clutter, occlusions) situations.","PeriodicalId":197319,"journal":{"name":"2018 10th IAPR Workshop on Pattern Recognition in Remote Sensing (PRRS)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2018-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125746746","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：