Through depth perception, humans have the ability to determine distances based on a single 2D image projected on their retina, where shape-from-shading (SFS) provides a mean to mimic such a phenomenon. The goal of this paper is to recover 3D facial shape from a single image of unknown general illumination, while relaxing the non-realistic assumption of Lambert Ian reflectance. Prior shape, albedo and reflectance models from real data, which are metric in nature, are incorporated into the shape recovery framework. Adopting a frequency-space based representation of the image irradiance equation, we propose an appearance model, termed as Harmonic Projection Images, which accounts explicitly for different human skin types as well as complex illumination conditions. Assuming skin reflectance obeys Torrance-Sparrow model, we prove analytically that it can be represented by at most 5th order harmonic basis whose closed form is provided. The recovery framework is a non-iterative approach which incorporates regression-like algorithm in the minimization process. Our experiments on synthetic and real images illustrate the robustness of our appearance model vis-a-vis illumination variation.
{"title":"Non-Lambertian Model-based Facial Shape Recovery from Single Image Under Unknown General Illumination","authors":"S. Elhabian, Eslam A. Mostafa, H. Rara, A. Farag","doi":"10.1109/CRV.2012.40","DOIUrl":"https://doi.org/10.1109/CRV.2012.40","url":null,"abstract":"Through depth perception, humans have the ability to determine distances based on a single 2D image projected on their retina, where shape-from-shading (SFS) provides a mean to mimic such a phenomenon. The goal of this paper is to recover 3D facial shape from a single image of unknown general illumination, while relaxing the non-realistic assumption of Lambert Ian reflectance. Prior shape, albedo and reflectance models from real data, which are metric in nature, are incorporated into the shape recovery framework. Adopting a frequency-space based representation of the image irradiance equation, we propose an appearance model, termed as Harmonic Projection Images, which accounts explicitly for different human skin types as well as complex illumination conditions. Assuming skin reflectance obeys Torrance-Sparrow model, we prove analytically that it can be represented by at most 5th order harmonic basis whose closed form is provided. The recovery framework is a non-iterative approach which incorporates regression-like algorithm in the minimization process. Our experiments on synthetic and real images illustrate the robustness of our appearance model vis-a-vis illumination variation.","PeriodicalId":372951,"journal":{"name":"2012 Ninth Conference on Computer and Robot Vision","volume":"285 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2012-05-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123034288","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}
High-level intuitive features (HLIF) that measure asymmetry of skin lesion images obtained using standard cameras are presented. These features can be used to help dermatologists objectively diagnose lesions as cancerous (melanoma) or benign with intuitive rationale. Existing work defines large sets of low-level statistical features for analysing skin lesions. The proposed HLIFs are designed such that smaller sets of HLIFs can capture more deterministic information than large sets of low-level features. Analytical reasoning is given for each feature to show how it aptly describes asymmetry. Promising experimental results show that classification using the proposed HLIF set, although only one-tenth the size of the existing state-of-the-art low-level feature set, labels the data with better or comparable success. The best classification is obtained by combining the low-level feature set with the HLIF set.
{"title":"Extracting High-Level Intuitive Features (HLIF) for Classifying Skin Lesions Using Standard Camera Images","authors":"R. Amelard, A. Wong, David A Clausi","doi":"10.1109/CRV.2012.59","DOIUrl":"https://doi.org/10.1109/CRV.2012.59","url":null,"abstract":"High-level intuitive features (HLIF) that measure asymmetry of skin lesion images obtained using standard cameras are presented. These features can be used to help dermatologists objectively diagnose lesions as cancerous (melanoma) or benign with intuitive rationale. Existing work defines large sets of low-level statistical features for analysing skin lesions. The proposed HLIFs are designed such that smaller sets of HLIFs can capture more deterministic information than large sets of low-level features. Analytical reasoning is given for each feature to show how it aptly describes asymmetry. Promising experimental results show that classification using the proposed HLIF set, although only one-tenth the size of the existing state-of-the-art low-level feature set, labels the data with better or comparable success. The best classification is obtained by combining the low-level feature set with the HLIF set.","PeriodicalId":372951,"journal":{"name":"2012 Ninth Conference on Computer and Robot Vision","volume":"7 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2012-05-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121048852","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}
We address the problem of repeated coverage of a target area, of any polygonal shape, by a team of robots having a limited visual range. Three distributed Cluster-based algorithms, and a method called Cyclic Coverage are introduced for the problem. The goal is to evaluate the performance of the repeated coverage algorithms under the effect of changes in the robots' visual range. A comprehensive set of performance metrics are considered, including the distance the robots travel, the frequency of visiting points in the target area, and the degree of balance in workload distribution among the robots. The Cyclic Coverage approach, used as a benchmark to compare the algorithms, produces optimal or near-optimal solutions for the single robot case under some criteria. The results show that the identity of the optimal repeated coverage algorithm depends on the metric and the robots' visual range.
{"title":"Multi-Robot Repeated Area Coverage: Performance Optimization Under Various Visual Ranges","authors":"Pooyan Fazli, Alireza Davoodi, Alan K. Mackworth","doi":"10.1109/CRV.2012.46","DOIUrl":"https://doi.org/10.1109/CRV.2012.46","url":null,"abstract":"We address the problem of repeated coverage of a target area, of any polygonal shape, by a team of robots having a limited visual range. Three distributed Cluster-based algorithms, and a method called Cyclic Coverage are introduced for the problem. The goal is to evaluate the performance of the repeated coverage algorithms under the effect of changes in the robots' visual range. A comprehensive set of performance metrics are considered, including the distance the robots travel, the frequency of visiting points in the target area, and the degree of balance in workload distribution among the robots. The Cyclic Coverage approach, used as a benchmark to compare the algorithms, produces optimal or near-optimal solutions for the single robot case under some criteria. The results show that the identity of the optimal repeated coverage algorithm depends on the metric and the robots' visual range.","PeriodicalId":372951,"journal":{"name":"2012 Ninth Conference on Computer and Robot Vision","volume":"14 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2012-05-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128672603","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}
The advent of the Kinect depth imager has opened the door to motion capture applications that would have been much more costly with previous technologies. In part, the Kinect achieves this by focusing on a very specific application domain, thus narrowing the requirement for the motion capture system. Specifically, Kinect motion capture works best within a small physical space while the camera is stationary. We seek to extend Kinect motion capture for use in athletic training - speed skating in particular - by placing the Kinect on a mobile, robotic platform to capture motion in situ. Athletes move over large distances, so the mobile platform addresses the limited viewing area of the Kinect. As the platform moves, we must also account for the now dynamic background against which the athlete performs. The result is a novel, visually-guided robotic platform that follows athletes, allowing us to capture motion and images that would not be possible with a treadmill. We describe the system in detail and give examples of the system capturing the motion of a speed skater at typical training speeds.
{"title":"In Situ Motion Capture of Speed Skating: Escaping the Treadmill","authors":"J. Boyd, Andrew Godbout, C. Thornton","doi":"10.1109/CRV.2012.68","DOIUrl":"https://doi.org/10.1109/CRV.2012.68","url":null,"abstract":"The advent of the Kinect depth imager has opened the door to motion capture applications that would have been much more costly with previous technologies. In part, the Kinect achieves this by focusing on a very specific application domain, thus narrowing the requirement for the motion capture system. Specifically, Kinect motion capture works best within a small physical space while the camera is stationary. We seek to extend Kinect motion capture for use in athletic training - speed skating in particular - by placing the Kinect on a mobile, robotic platform to capture motion in situ. Athletes move over large distances, so the mobile platform addresses the limited viewing area of the Kinect. As the platform moves, we must also account for the now dynamic background against which the athlete performs. The result is a novel, visually-guided robotic platform that follows athletes, allowing us to capture motion and images that would not be possible with a treadmill. We describe the system in detail and give examples of the system capturing the motion of a speed skater at typical training speeds.","PeriodicalId":372951,"journal":{"name":"2012 Ninth Conference on Computer and Robot Vision","volume":"583 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2012-05-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134118009","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}
The advent of RGB-D cameras which provide synchronized range and video data creates new opportunities for exploiting both sensing modalities for various robotic applications. This paper exploits the strengths of vision and range measurements and develops a novel robust algorithm for localization using RGB-D cameras. We show how correspondences established by matching visual SIFT features can effectively initialize the generalized ICP algorithm as well as demonstrate situations where such initialization is not viable. We propose an adaptive architecture which computes the pose estimate from the most reliable measurements in a given environment and present thorough evaluation of the resulting algorithm against a dataset of RGB-D benchmarks, demonstrating superior or comparable performance in the absence of the global optimization stage. Lastly we demonstrate the proposed algorithm on a challenging indoor dataset and demonstrate improvements where pose estimation from either pure range sensing or vision techniques perform poorly.
{"title":"Adaptive RGB-D Localization","authors":"M. Paton, J. Kosecka","doi":"10.1109/CRV.2012.11","DOIUrl":"https://doi.org/10.1109/CRV.2012.11","url":null,"abstract":"The advent of RGB-D cameras which provide synchronized range and video data creates new opportunities for exploiting both sensing modalities for various robotic applications. This paper exploits the strengths of vision and range measurements and develops a novel robust algorithm for localization using RGB-D cameras. We show how correspondences established by matching visual SIFT features can effectively initialize the generalized ICP algorithm as well as demonstrate situations where such initialization is not viable. We propose an adaptive architecture which computes the pose estimate from the most reliable measurements in a given environment and present thorough evaluation of the resulting algorithm against a dataset of RGB-D benchmarks, demonstrating superior or comparable performance in the absence of the global optimization stage. Lastly we demonstrate the proposed algorithm on a challenging indoor dataset and demonstrate improvements where pose estimation from either pure range sensing or vision techniques perform poorly.","PeriodicalId":372951,"journal":{"name":"2012 Ninth Conference on Computer and Robot Vision","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2012-05-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134619674","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}
Humanoid robots such as DLR's Justin are built with light-weight structures and flexible mechanical components. These generate positioning errors at the TCP (Tool-Center-Point) end-pose of the hand. The identification of these errors is essential for object manipulation and path planning. We proposed a verification routine to identify the bounds of the TCP end-pose errors by using the on-board stereo vision system. It involves estimating the pose of 3D point clouds of Justin's hand by using state-of-the-art 3D registration techniques. Partial models of the hand were generated by registering subsets of overlapping 3D point clouds. We proposed a method for the selection of overlapping point clouds of self-occluding objects (Justin's hand). It is based on a statistical analysis of the depth values. We applied an extended metaview registration method to the resulting subset of point clouds. The partial models were evaluated with detailed based surface consistency measures. The TCP end-pose errors estimated by using our method are consistent with ground-truth errors.
{"title":"3D Registration for Verification of Humanoid Justin's Upper Body Kinematics","authors":"Nadia Figueroa, Haider Ali, Florian Schmidt","doi":"10.1109/CRV.2012.43","DOIUrl":"https://doi.org/10.1109/CRV.2012.43","url":null,"abstract":"Humanoid robots such as DLR's Justin are built with light-weight structures and flexible mechanical components. These generate positioning errors at the TCP (Tool-Center-Point) end-pose of the hand. The identification of these errors is essential for object manipulation and path planning. We proposed a verification routine to identify the bounds of the TCP end-pose errors by using the on-board stereo vision system. It involves estimating the pose of 3D point clouds of Justin's hand by using state-of-the-art 3D registration techniques. Partial models of the hand were generated by registering subsets of overlapping 3D point clouds. We proposed a method for the selection of overlapping point clouds of self-occluding objects (Justin's hand). It is based on a statistical analysis of the depth values. We applied an extended metaview registration method to the resulting subset of point clouds. The partial models were evaluated with detailed based surface consistency measures. The TCP end-pose errors estimated by using our method are consistent with ground-truth errors.","PeriodicalId":372951,"journal":{"name":"2012 Ninth Conference on Computer and Robot Vision","volume":"7 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2012-05-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134495623","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}
This paper proposes an automatic pose-invariant face recognition system. In our approach, we consider the texture information around the facial features to compute the similarity measure between the probe and gallery images. The weight of each facial feature is dynamically estimated based on its robustness to the pose of the captured image. An approach to extract the 9 facial features used to initialize the Active shape model is proposed. The approach is not dependent on the texture around the facial feature only but incorporates the information obtained about the facial feature relations. Our face recognition system is tested on common datasets in pose evaluation CMU-PIE and FERET. The results show out-performance of the state of the art automatic face recognition systems.
{"title":"Dynamic Weighting of Facial Features for Automatic Pose-Invariant Face Recognition","authors":"Eslam A. Mostafa, A. Farag","doi":"10.1109/CRV.2012.61","DOIUrl":"https://doi.org/10.1109/CRV.2012.61","url":null,"abstract":"This paper proposes an automatic pose-invariant face recognition system. In our approach, we consider the texture information around the facial features to compute the similarity measure between the probe and gallery images. The weight of each facial feature is dynamically estimated based on its robustness to the pose of the captured image. An approach to extract the 9 facial features used to initialize the Active shape model is proposed. The approach is not dependent on the texture around the facial feature only but incorporates the information obtained about the facial feature relations. Our face recognition system is tested on common datasets in pose evaluation CMU-PIE and FERET. The results show out-performance of the state of the art automatic face recognition systems.","PeriodicalId":372951,"journal":{"name":"2012 Ninth Conference on Computer and Robot Vision","volume":"215 1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2012-05-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121201558","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}
We implement and quantitatively/qualitatively evaluate two optical flow methods that model occlusion. The Yuan et al. method [1] improves on the Horn and Schunck optical flow method at occlusion boundaries by using a dynamic coefficient (the Lagrange multiplier α) at each pixel that weighs the smoothness constraint relative to the optical flow constraint, by adopting a modified scheme to calculate average velocities and by using a “compensating” iterative algorithm to achieve higher computational efficiency. The Niu et al. method [2] is based on a modified version of the Lucas and Kanade optical flow method, that selects local intensity neighbourhoods, spatially and temporally, based on pixels that are on different sides of an occlusion boundary and then corrects any erroneous flow computed at occlusion boundaries. We present quantitative results for sinusoidal sequence with a known occlusion boundary. We also present qualitative evaluation of the methods on the Hamburg Taxi sequence and and the Trees sequence.
{"title":"Optical Flow at Occlusion","authors":"Jieyu Zhang, J. Barron","doi":"10.1109/CRV.2012.34","DOIUrl":"https://doi.org/10.1109/CRV.2012.34","url":null,"abstract":"We implement and quantitatively/qualitatively evaluate two optical flow methods that model occlusion. The Yuan et al. method [1] improves on the Horn and Schunck optical flow method at occlusion boundaries by using a dynamic coefficient (the Lagrange multiplier α) at each pixel that weighs the smoothness constraint relative to the optical flow constraint, by adopting a modified scheme to calculate average velocities and by using a “compensating” iterative algorithm to achieve higher computational efficiency. The Niu et al. method [2] is based on a modified version of the Lucas and Kanade optical flow method, that selects local intensity neighbourhoods, spatially and temporally, based on pixels that are on different sides of an occlusion boundary and then corrects any erroneous flow computed at occlusion boundaries. We present quantitative results for sinusoidal sequence with a known occlusion boundary. We also present qualitative evaluation of the methods on the Hamburg Taxi sequence and and the Trees sequence.","PeriodicalId":372951,"journal":{"name":"2012 Ninth Conference on Computer and Robot Vision","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2012-05-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129486840","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}
Improving laser range data acquisition speed is important for many robotic applications such as mapping and localization. One approach to reducing acquisition time is to acquire laser range data through a dynamically small subset of measurement locations. The reconstruction can then be performed based on the concept of compressed sensing (CS), where a sparse signal representation allows for signal reconstruction at sub-Nyquist measurements. Motivated by this, a novel multi-scale saliency-guided CS-based algorithm is proposed for an efficient robotic laser range data acquisition for robotic vision. The proposed system samples the objects of interest through an optimized probability density function derived based on multi-scale saliency rather than the uniform random distribution used in traditional CS systems. Experimental results with laser range data from indoor and outdoor environments show that the proposed approach requires less than half the samples needed by existing CS-based approaches while maintaining the same reconstruction performance. In addition, the proposed method offers significant improvement in reconstruction SNR compared to current CS-based approaches.
{"title":"Multi-Scale Saliency-Guided Compressive Sensing Approach to Efficient Robotic Laser Range Measurements","authors":"S. Schwartz, A. Wong, David A Clausi","doi":"10.1109/CRV.2012.8","DOIUrl":"https://doi.org/10.1109/CRV.2012.8","url":null,"abstract":"Improving laser range data acquisition speed is important for many robotic applications such as mapping and localization. One approach to reducing acquisition time is to acquire laser range data through a dynamically small subset of measurement locations. The reconstruction can then be performed based on the concept of compressed sensing (CS), where a sparse signal representation allows for signal reconstruction at sub-Nyquist measurements. Motivated by this, a novel multi-scale saliency-guided CS-based algorithm is proposed for an efficient robotic laser range data acquisition for robotic vision. The proposed system samples the objects of interest through an optimized probability density function derived based on multi-scale saliency rather than the uniform random distribution used in traditional CS systems. Experimental results with laser range data from indoor and outdoor environments show that the proposed approach requires less than half the samples needed by existing CS-based approaches while maintaining the same reconstruction performance. In addition, the proposed method offers significant improvement in reconstruction SNR compared to current CS-based approaches.","PeriodicalId":372951,"journal":{"name":"2012 Ninth Conference on Computer and Robot Vision","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2012-05-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131135361","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}
This paper describes a biologically motivated computational model, termed as early recurrent inhibition, to improve edge representation. The computation borrows the idea from the primate visual system that visual features are calculated in the two main visual pathways with different speeds and thus one can positively affect the other via early recurrent mechanisms. Based on the collected results, we conclude such a recurrent processing from area MT to the ventral layers of the primary visual area (V1) may be at play, and hypothesize that one effect of this recurrent mechanism is that V1 responses to high-spatial frequency edges are suppressed by signals sent from MT, leading to a cleaner edge representation. The operation is modeled as a weighted multiplicative inhibition process. Depending on the weighting methods, two types of inhibition are investigated, namely isotropic and anisotropic inhibition. To evaluate the inhibited edge representation, our model is attached to a contour operator to generate binary contour maps. Using real images, we quantitatively compared contours calculated by our work with those by a well-known biologically motivated model. Results clearly demonstrate that early recurrent inhibition has a positive and consistent influence on edge detection.
{"title":"Improved Edge Representation via Early Recurrent Inhibition","authors":"Xun Shi, John K. Tsotsos","doi":"10.1109/CRV.2012.13","DOIUrl":"https://doi.org/10.1109/CRV.2012.13","url":null,"abstract":"This paper describes a biologically motivated computational model, termed as early recurrent inhibition, to improve edge representation. The computation borrows the idea from the primate visual system that visual features are calculated in the two main visual pathways with different speeds and thus one can positively affect the other via early recurrent mechanisms. Based on the collected results, we conclude such a recurrent processing from area MT to the ventral layers of the primary visual area (V1) may be at play, and hypothesize that one effect of this recurrent mechanism is that V1 responses to high-spatial frequency edges are suppressed by signals sent from MT, leading to a cleaner edge representation. The operation is modeled as a weighted multiplicative inhibition process. Depending on the weighting methods, two types of inhibition are investigated, namely isotropic and anisotropic inhibition. To evaluate the inhibited edge representation, our model is attached to a contour operator to generate binary contour maps. Using real images, we quantitatively compared contours calculated by our work with those by a well-known biologically motivated model. Results clearly demonstrate that early recurrent inhibition has a positive and consistent influence on edge detection.","PeriodicalId":372951,"journal":{"name":"2012 Ninth Conference on Computer and Robot Vision","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2012-05-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132762135","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：