Pub Date : 2018-08-01DOI: 10.1109/PRRS.2018.8486287
Song Cui, Yanfei Zhong
Automatic matching of multi-modal remote sensing images remains a challenging task in remote sensing image analysis due to significant non-linear radiometric differences between these images. This paper introduces the phase congruency model with illumination and contrast invariance for image matching, and extends the model to a novel image registration method, named as multi-scale phase consistency (MS-PC). The Euclidean distance between MS-PC descriptors is used as similarity metric to achieve correspondences. The proposed method is evaluated with four pairs of multi-model remote sensing images. The experimental results show that MS-PC is more robust to the radiation differences between images, and performs better than two popular method (i.e. SIFT and SAR-SIFT) in both registration accuracy and tie points number.
{"title":"Multi-Modal Remote Sensing Image Registration Based on Multi-Scale Phase Congruency","authors":"Song Cui, Yanfei Zhong","doi":"10.1109/PRRS.2018.8486287","DOIUrl":"https://doi.org/10.1109/PRRS.2018.8486287","url":null,"abstract":"Automatic matching of multi-modal remote sensing images remains a challenging task in remote sensing image analysis due to significant non-linear radiometric differences between these images. This paper introduces the phase congruency model with illumination and contrast invariance for image matching, and extends the model to a novel image registration method, named as multi-scale phase consistency (MS-PC). The Euclidean distance between MS-PC descriptors is used as similarity metric to achieve correspondences. The proposed method is evaluated with four pairs of multi-model remote sensing images. The experimental results show that MS-PC is more robust to the radiation differences between images, and performs better than two popular method (i.e. SIFT and SAR-SIFT) in both registration accuracy and tie points number.","PeriodicalId":197319,"journal":{"name":"2018 10th IAPR Workshop on Pattern Recognition in Remote Sensing (PRRS)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130019209","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.2018.8486322
E. Michaelsen, U. Soergel
In man-made structures regularities and repetitions prevails. In particular in building facades lattices are common in which windows and other elements are repeated as well in vertical columns as in horizontal rows. In very-high-resolution space-borne radar images such lattices appear saliently. Even untrained arbitrary subjects see the structure instantaneously. However, automatic perceptual grouping is rarely attempted. This contribution applies a new lattice grouping method to such data. Utilization of knowledge about the particular mapping process of such radar data is distinguished from the use of Gestalt laws. The latter are universally applicable to all kinds of pictorial data. An example with so called permanent scatterers in the city of Berlin shows what can be achieved with automatic perceptual grouping alone, and what can be gained using domain knowledge. Keywords- perceptual grouping, SAR, permanent scatterers, façade recognition
{"title":"Reconstructing Lattices from Permanent Scatterers on Facades","authors":"E. Michaelsen, U. Soergel","doi":"10.1109/PRRS.2018.8486322","DOIUrl":"https://doi.org/10.1109/PRRS.2018.8486322","url":null,"abstract":"In man-made structures regularities and repetitions prevails. In particular in building facades lattices are common in which windows and other elements are repeated as well in vertical columns as in horizontal rows. In very-high-resolution space-borne radar images such lattices appear saliently. Even untrained arbitrary subjects see the structure instantaneously. However, automatic perceptual grouping is rarely attempted. This contribution applies a new lattice grouping method to such data. Utilization of knowledge about the particular mapping process of such radar data is distinguished from the use of Gestalt laws. The latter are universally applicable to all kinds of pictorial data. An example with so called permanent scatterers in the city of Berlin shows what can be achieved with automatic perceptual grouping alone, and what can be gained using domain knowledge. Keywords- perceptual grouping, SAR, permanent scatterers, façade recognition","PeriodicalId":197319,"journal":{"name":"2018 10th IAPR Workshop on Pattern Recognition in Remote Sensing (PRRS)","volume":"24 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129131599","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.2018.8486395
M. Rezaee, Yun Zhang, Rakesh K. Mishra, Fei Tong, Hengjian Tong
Acquiring information about forest stands such as individual tree species is crucial for monitoring forests. To date, such information is assessed by human interpreters using airborne or an Unmanned Aerial Vehicle (UAV), which is time/cost consuming. The recent advancement in remote sensing image acquisition, such as WorldView-3, has increased the spatial resolution up to 30 cm and spectral resolution up to 16 bands. This advancement has significantly increased the potential for Individual Tree Species Detection (ITSD). In order to use the single source Worldview-3 images, our proposed method first segments the image to delineate trees, and then detects trees using a VGG-16 network. We developed a pipeline for feeding the deep CNN network using the information from all the 8 visible-near infrareds' bands and trained it. The result is compared with two state-of-the-art ensemble classifiers namely Random Forest (RF) and Gradient Boosting (GB). Results demonstrate that the VGG-16 outperforms all the other methods reaching an accuracy of about 92.13%.
{"title":"Using a VGG-16 Network for Individual Tree Species Detection with an Object-Based Approach","authors":"M. Rezaee, Yun Zhang, Rakesh K. Mishra, Fei Tong, Hengjian Tong","doi":"10.1109/PRRS.2018.8486395","DOIUrl":"https://doi.org/10.1109/PRRS.2018.8486395","url":null,"abstract":"Acquiring information about forest stands such as individual tree species is crucial for monitoring forests. To date, such information is assessed by human interpreters using airborne or an Unmanned Aerial Vehicle (UAV), which is time/cost consuming. The recent advancement in remote sensing image acquisition, such as WorldView-3, has increased the spatial resolution up to 30 cm and spectral resolution up to 16 bands. This advancement has significantly increased the potential for Individual Tree Species Detection (ITSD). In order to use the single source Worldview-3 images, our proposed method first segments the image to delineate trees, and then detects trees using a VGG-16 network. We developed a pipeline for feeding the deep CNN network using the information from all the 8 visible-near infrareds' bands and trained it. The result is compared with two state-of-the-art ensemble classifiers namely Random Forest (RF) and Gradient Boosting (GB). Results demonstrate that the VGG-16 outperforms all the other methods reaching an accuracy of about 92.13%.","PeriodicalId":197319,"journal":{"name":"2018 10th IAPR Workshop on Pattern Recognition in Remote Sensing (PRRS)","volume":"5 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126826878","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.2018.8486164
Mengmeng Zhang, Wei Li, Xueling Wei, Xiang Li
Currently, how to efficiently exploit useful information from multi-source remote sensing data for better Earth observation becomes an interesting but challenging problem. In this paper, we propose an collaborative classification framework for hyperspectral image (HSI) and Light Detection and Ranging (LIDAR) data via image-to-image convolutional neural network (CNN). There is an image-to-image mapping, learning a representation from input source (i.e., HSI) to output source (i.e., LIDAR). Then, the extracted features are expected to own characteristics of both HSI and LIDAR data, and the collaborative classification is implemented by integrating hidden layers of the deep CNN. Experimental results on two real remote sensing data sets demonstrate the effectiveness of the proposed framework.
{"title":"Collaborative Classification of Hyperspectral and LIDAR Data Using Unsupervised Image-to-Image CNN","authors":"Mengmeng Zhang, Wei Li, Xueling Wei, Xiang Li","doi":"10.1109/PRRS.2018.8486164","DOIUrl":"https://doi.org/10.1109/PRRS.2018.8486164","url":null,"abstract":"Currently, how to efficiently exploit useful information from multi-source remote sensing data for better Earth observation becomes an interesting but challenging problem. In this paper, we propose an collaborative classification framework for hyperspectral image (HSI) and Light Detection and Ranging (LIDAR) data via image-to-image convolutional neural network (CNN). There is an image-to-image mapping, learning a representation from input source (i.e., HSI) to output source (i.e., LIDAR). Then, the extracted features are expected to own characteristics of both HSI and LIDAR data, and the collaborative classification is implemented by integrating hidden layers of the deep CNN. Experimental results on two real remote sensing data sets demonstrate the effectiveness of the proposed framework.","PeriodicalId":197319,"journal":{"name":"2018 10th IAPR Workshop on Pattern Recognition in Remote Sensing (PRRS)","volume":"15 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125191331","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.2018.8486193
Pel Pengcheng, Shi Yue, Wan ChengBo, Ma Xinming, Guo Wa, Qiao Rongbo
Since SVM is sensitive to the noises and outliers in the training set, a new SVM algorithm based on affinity Grey-Sigmoid kernel is proposed in the paper. The cluster membership is defined by the distance from the cluster center, but also defined by the affinity among samples. The affinity among samples is measured by the minimum super sphere which containing the maximum of the samples. Then the Grey degree of samples are defined by their position in the super sphere. Compared with the SVM based on traditional Sigmoid kernel, experimental results show that the Grey-Sigmoid kernel is more robust and efficient.
{"title":"The UAV Image Classification Method Based on the Grey-Sigmoid Kernel Function Support Vector Machine","authors":"Pel Pengcheng, Shi Yue, Wan ChengBo, Ma Xinming, Guo Wa, Qiao Rongbo","doi":"10.1109/PRRS.2018.8486193","DOIUrl":"https://doi.org/10.1109/PRRS.2018.8486193","url":null,"abstract":"Since SVM is sensitive to the noises and outliers in the training set, a new SVM algorithm based on affinity Grey-Sigmoid kernel is proposed in the paper. The cluster membership is defined by the distance from the cluster center, but also defined by the affinity among samples. The affinity among samples is measured by the minimum super sphere which containing the maximum of the samples. Then the Grey degree of samples are defined by their position in the super sphere. Compared with the SVM based on traditional Sigmoid kernel, experimental results show that the Grey-Sigmoid kernel is more robust and efficient.","PeriodicalId":197319,"journal":{"name":"2018 10th IAPR Workshop on Pattern Recognition in Remote Sensing (PRRS)","volume":"57 4","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114050358","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.2018.8486304
Meng Zhang, L. Hong
The spectral resolution and spatial resolution of hyperspectral images are continuously improving, providing rich information for interpreting remote sensing image. How to improve the image classification accuracy has become the focus of many studies. Recently, Deep learning is capable to extract discriminating high-level abstract features for image classification task, and some interesting results have been acquired in image processing. However, when deep learning is applied to the classification of hyperspectral remote sensing images, the spectral-based classification method is short of spatial and scale information; the image patch-based classification method ignores the rich spectral information provided by hyperspectral images. In this study, a multi-scale feature fusion hyperspectral image classification method based on deep learning was proposed. Firstly, multiscale features were obtained by multi-scale segmentation. Then multiscale features were input into the convolution neural network to extract high-level features. Finally, the high-level features were used for classification. Experimental results show that the classification results of the fusion multi-scale features are better than the single-scale features and regional feature classification results.
{"title":"Deep Learning Integrated with Multiscale Pixel and Object Features for Hyperspectral Image Classification","authors":"Meng Zhang, L. Hong","doi":"10.1109/PRRS.2018.8486304","DOIUrl":"https://doi.org/10.1109/PRRS.2018.8486304","url":null,"abstract":"The spectral resolution and spatial resolution of hyperspectral images are continuously improving, providing rich information for interpreting remote sensing image. How to improve the image classification accuracy has become the focus of many studies. Recently, Deep learning is capable to extract discriminating high-level abstract features for image classification task, and some interesting results have been acquired in image processing. However, when deep learning is applied to the classification of hyperspectral remote sensing images, the spectral-based classification method is short of spatial and scale information; the image patch-based classification method ignores the rich spectral information provided by hyperspectral images. In this study, a multi-scale feature fusion hyperspectral image classification method based on deep learning was proposed. Firstly, multiscale features were obtained by multi-scale segmentation. Then multiscale features were input into the convolution neural network to extract high-level features. Finally, the high-level features were used for classification. Experimental results show that the classification results of the fusion multi-scale features are better than the single-scale features and regional feature classification results.","PeriodicalId":197319,"journal":{"name":"2018 10th IAPR Workshop on Pattern Recognition in Remote Sensing (PRRS)","volume":"15 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133976399","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":"20 1","pages":"0"},"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.2018.8486185
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":"63 1","pages":"0"},"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.2018.8486325
L. Li, C. Yu, T. Sun, Z. Han, X. Tang
For the high-resolution borehole image obtained by digital panoramic borehole camera system, a method for recognizing soil layer based on color features is proposed. Due to the obvious difference in color between soil layer and common rock layer, a soil layer detection model based on HSV color space is established. The binarized image of soil layer is obtained by using this model. Secondly, the binary image is filtered to depress the noise effects. Then, the binarized image of the soil layer is divided and the density of pixels in each segmentation is calculated to determine the depth, area and direction of the soil layer, so that the identification of soil layer in the digital borehole image can be achieved. Through verifying this method with many actual borehole images and comparing them with the corresponding borehole radar images, the result illustrate that this method can identify all of the soil layer throughout the whole borehole digital optical image automatically and quickly. It provides a new reliable method for the automatic identification of borehole structural planes in engineering application.
{"title":"Automatic Identification of Soil Layer from Borehole Digital Optical Image and GPR Based on Color Features","authors":"L. Li, C. Yu, T. Sun, Z. Han, X. Tang","doi":"10.1109/PRRS.2018.8486325","DOIUrl":"https://doi.org/10.1109/PRRS.2018.8486325","url":null,"abstract":"For the high-resolution borehole image obtained by digital panoramic borehole camera system, a method for recognizing soil layer based on color features is proposed. Due to the obvious difference in color between soil layer and common rock layer, a soil layer detection model based on HSV color space is established. The binarized image of soil layer is obtained by using this model. Secondly, the binary image is filtered to depress the noise effects. Then, the binarized image of the soil layer is divided and the density of pixels in each segmentation is calculated to determine the depth, area and direction of the soil layer, so that the identification of soil layer in the digital borehole image can be achieved. Through verifying this method with many actual borehole images and comparing them with the corresponding borehole radar images, the result illustrate that this method can identify all of the soil layer throughout the whole borehole digital optical image automatically and quickly. It provides a new reliable method for the automatic identification of borehole structural planes in engineering application.","PeriodicalId":197319,"journal":{"name":"2018 10th IAPR Workshop on Pattern Recognition in Remote Sensing (PRRS)","volume":"69 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115272104","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.2018.8486195
Qingling Jia, Xue Wan, Baoqin Hei, Shengyang Li
Recent work has shown that convolutional neural network can solve the stereo matching problems in artificial scene successfully, such as buildings, roads and so on. However, whether it is suitable for remote sensing stereo image matching in featureless area, for example lunar surface, is uncertain. This paper exploits the ability of DispNet, an end-to-end disparity estimation algorithm based on convolutional neural network, for image matching in featureless lunar surface areas. Experiments using image pairs from NASA Polar Stereo Dataset demonstrate that DispNet has superior performance in the aspects of matching accuracy, the continuity of disparity and speed compared to three traditional stereo matching methods, SGM, BM and SAD. Thus it has the potential for the application in future planetary exploration tasks such as visual odometry for rover navigation and image matching for precise landing
{"title":"DispNet Based Stereo Matching for Planetary Scene Depth Estimation Using Remote Sensing Images","authors":"Qingling Jia, Xue Wan, Baoqin Hei, Shengyang Li","doi":"10.1109/PRRS.2018.8486195","DOIUrl":"https://doi.org/10.1109/PRRS.2018.8486195","url":null,"abstract":"Recent work has shown that convolutional neural network can solve the stereo matching problems in artificial scene successfully, such as buildings, roads and so on. However, whether it is suitable for remote sensing stereo image matching in featureless area, for example lunar surface, is uncertain. This paper exploits the ability of DispNet, an end-to-end disparity estimation algorithm based on convolutional neural network, for image matching in featureless lunar surface areas. Experiments using image pairs from NASA Polar Stereo Dataset demonstrate that DispNet has superior performance in the aspects of matching accuracy, the continuity of disparity and speed compared to three traditional stereo matching methods, SGM, BM and SAD. Thus it has the potential for the application in future planetary exploration tasks such as visual odometry for rover navigation and image matching for precise landing","PeriodicalId":197319,"journal":{"name":"2018 10th IAPR Workshop on Pattern Recognition in Remote Sensing (PRRS)","volume":"47 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124693711","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：