This paper presents a novel approach to modeling curiosity in a mobile robot, which is useful for monitoring and adaptive data collection tasks. We use ROST, a real time topic modeling framework to build a semantic perception model of the environment, using which, we plan a path through the locations in the world with high semantic information content. We demonstrate the approach using the Aqua robot in a variety of different scenarios, and find the robot be able to do tasks such as coral reef inspection, diver following, and sea floor exploration, without any prior training or preparation.
{"title":"Exploring Underwater Environments with Curiosity","authors":"Yogesh A. Girdhar, G. Dudek","doi":"10.1109/CRV.2014.22","DOIUrl":"https://doi.org/10.1109/CRV.2014.22","url":null,"abstract":"This paper presents a novel approach to modeling curiosity in a mobile robot, which is useful for monitoring and adaptive data collection tasks. We use ROST, a real time topic modeling framework to build a semantic perception model of the environment, using which, we plan a path through the locations in the world with high semantic information content. We demonstrate the approach using the Aqua robot in a variety of different scenarios, and find the robot be able to do tasks such as coral reef inspection, diver following, and sea floor exploration, without any prior training or preparation.","PeriodicalId":385422,"journal":{"name":"2014 Canadian Conference on Computer and Robot Vision","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2014-05-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123443535","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}
Particle swarm optimization (PSO) is a stochastic global optimization approach whose search characteristics are controlled by three parameters, inertial weight w, cognitive parameter c1 and social parameter c2. Large parameter values facilitate exploration by searching new horizons of solution space. On the other hand, small parameter values facilitate exploitation by searching the neighborhood. An appropriate value of these parameters provides a balance between exploration and exploitation and results in better performance. An adaptive parameter PSO (AP-PSO) algorithm is proposed in this work where the inertial weight is gradually decreased and values of the cognitive and social parameters depend on the fitness values. Good fitness values support exploitation and poor fitness values support exploration. The proposed algorithm has shown excellent performance on low dimensional system identification problems as well as high dimensional articulated human tracking (AHT) problems.
{"title":"A New Fitness Based Adaptive Parameter Particle Swarm Optimizer","authors":"S. Akhtar, E. Abdel-Rahman, Abdul-Rahim Ahmad","doi":"10.1109/CRV.2014.52","DOIUrl":"https://doi.org/10.1109/CRV.2014.52","url":null,"abstract":"Particle swarm optimization (PSO) is a stochastic global optimization approach whose search characteristics are controlled by three parameters, inertial weight w, cognitive parameter c1 and social parameter c2. Large parameter values facilitate exploration by searching new horizons of solution space. On the other hand, small parameter values facilitate exploitation by searching the neighborhood. An appropriate value of these parameters provides a balance between exploration and exploitation and results in better performance. An adaptive parameter PSO (AP-PSO) algorithm is proposed in this work where the inertial weight is gradually decreased and values of the cognitive and social parameters depend on the fitness values. Good fitness values support exploitation and poor fitness values support exploration. The proposed algorithm has shown excellent performance on low dimensional system identification problems as well as high dimensional articulated human tracking (AHT) problems.","PeriodicalId":385422,"journal":{"name":"2014 Canadian Conference on Computer and Robot Vision","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2014-05-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126003891","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}
In summer 2013, a Mercedes S-Class drove completely autonomously for about 100 km from Mannheim to Pforzheim, Germany, using only close-to-production sensors. In this project, called Mercedes Benz Intelligent Drive, stereo vision was one of the main sensing components. For the representation of free space and obstacles we relied on the so called Stixel World, a generic 3D intermediate representation which is computed from dense disparity images. In spite of the high performance of the Stixel World in most common traffic scenes, the availability of this technique is limited. For instance under adverse weather, rain or even spray water on the windshield results in erroneous disparity images which generate false Stixel results. This can lead to undesired behavior of autonomous vehicles. Our goal is to use the Stixel World for a robust free space estimation and a reliable obstacle detection even during difficult weather conditions. In this paper, we meet this challenge and fuse the Stixels incrementally into a reference grid map. Our new approach is formulated in a Bayesian manner and is based on existence estimation methods. We evaluate our new technique on a manually labeled database with emphasis on bad weather scenarios. The number of structures which are detected mistakenly within free space areas is reduced by a factor of two whereas the detection rate of obstacles increases at the same time.
{"title":"Stix-Fusion: A Probabilistic Stixel Integration Technique","authors":"M. Muffert, Nicolai Schneider, U. Franke","doi":"10.1109/CRV.2014.11","DOIUrl":"https://doi.org/10.1109/CRV.2014.11","url":null,"abstract":"In summer 2013, a Mercedes S-Class drove completely autonomously for about 100 km from Mannheim to Pforzheim, Germany, using only close-to-production sensors. In this project, called Mercedes Benz Intelligent Drive, stereo vision was one of the main sensing components. For the representation of free space and obstacles we relied on the so called Stixel World, a generic 3D intermediate representation which is computed from dense disparity images. In spite of the high performance of the Stixel World in most common traffic scenes, the availability of this technique is limited. For instance under adverse weather, rain or even spray water on the windshield results in erroneous disparity images which generate false Stixel results. This can lead to undesired behavior of autonomous vehicles. Our goal is to use the Stixel World for a robust free space estimation and a reliable obstacle detection even during difficult weather conditions. In this paper, we meet this challenge and fuse the Stixels incrementally into a reference grid map. Our new approach is formulated in a Bayesian manner and is based on existence estimation methods. We evaluate our new technique on a manually labeled database with emphasis on bad weather scenarios. The number of structures which are detected mistakenly within free space areas is reduced by a factor of two whereas the detection rate of obstacles increases at the same time.","PeriodicalId":385422,"journal":{"name":"2014 Canadian Conference on Computer and Robot Vision","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2014-05-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128270411","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}
A method for detecting local defects in moulded plastic parts is presented. The method uses deflectometry to produce a contrast enhanced image that is later processed in a novel algorithm. The method operates without the need for accurate mechanical models and is robust to changes in image resolution. Experimental results show that the method can detect subtle defects with over 90% accuracy on most parts.
{"title":"Robust Detection of Paint Defects in Moulded Plastic Parts","authors":"Cole Tarry, Michael Stachowsky, M. Moussa","doi":"10.1109/CRV.2014.48","DOIUrl":"https://doi.org/10.1109/CRV.2014.48","url":null,"abstract":"A method for detecting local defects in moulded plastic parts is presented. The method uses deflectometry to produce a contrast enhanced image that is later processed in a novel algorithm. The method operates without the need for accurate mechanical models and is robust to changes in image resolution. Experimental results show that the method can detect subtle defects with over 90% accuracy on most parts.","PeriodicalId":385422,"journal":{"name":"2014 Canadian Conference on Computer and Robot Vision","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2014-05-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134513715","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}
Automatic sign language recognition is an open problem that has received a lot of attention recently, not only because of its usefulness to signers, but also due to the numerous applications a sign classifier can have. In this article, we present a new feature extraction technique for hand pose recognition using depth and intensity images captured from a Microsoft Kinect sensor. We applied our technique to American Sign Language finger spelling classification using a Deep Belief Network, for which our feature extraction technique is tailored. We evaluated our results on a multi-user data set with two scenarios: one with all known users and one with an unseen user. We achieved 99% recall and precision on the first, and 77% recall and 79% precision on the second. Our method is also capable of real-time sign classification and is adaptive to any environment or lightning intensity.
{"title":"Sign Language Fingerspelling Classification from Depth and Color Images Using a Deep Belief Network","authors":"Lucas Rioux-Maldague, P. Giguère","doi":"10.1109/CRV.2014.20","DOIUrl":"https://doi.org/10.1109/CRV.2014.20","url":null,"abstract":"Automatic sign language recognition is an open problem that has received a lot of attention recently, not only because of its usefulness to signers, but also due to the numerous applications a sign classifier can have. In this article, we present a new feature extraction technique for hand pose recognition using depth and intensity images captured from a Microsoft Kinect sensor. We applied our technique to American Sign Language finger spelling classification using a Deep Belief Network, for which our feature extraction technique is tailored. We evaluated our results on a multi-user data set with two scenarios: one with all known users and one with an unseen user. We achieved 99% recall and precision on the first, and 77% recall and 79% precision on the second. Our method is also capable of real-time sign classification and is adaptive to any environment or lightning intensity.","PeriodicalId":385422,"journal":{"name":"2014 Canadian Conference on Computer and Robot Vision","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2014-05-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134265366","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}
A novel approach is proposed for real-time retrieval of images from a large database of overlapping images of an indoor environment. The procedure extracts visual features from images using selected computer vision techniques, and processes the extracted features to create a reduced list of features annotated with the frame numbers they appear in. This method is named landmark indexing. Unlike some state-of-the-art approaches, the proposed method does not need to consider large image adjacency graphs because the overlap of the images in the map sufficiently increases information gain, and mapping of similar features to the same landmark reduces the search space to improve search efficiency. Empirical evidence from experiments on real datasets shows better performance and accuracy than other approaches. Experiments are further performed by integrating the image retrieval technique into a 3D real-time navigation system. This system is tested in several indoor environments and all experiments show highly accurate localization results.
{"title":"Image Retrieval Using Landmark Indexing for Indoor Navigation","authors":"Dwaipayan Sinha, M. Ahmed, M. Greenspan","doi":"10.1109/CRV.2014.17","DOIUrl":"https://doi.org/10.1109/CRV.2014.17","url":null,"abstract":"A novel approach is proposed for real-time retrieval of images from a large database of overlapping images of an indoor environment. The procedure extracts visual features from images using selected computer vision techniques, and processes the extracted features to create a reduced list of features annotated with the frame numbers they appear in. This method is named landmark indexing. Unlike some state-of-the-art approaches, the proposed method does not need to consider large image adjacency graphs because the overlap of the images in the map sufficiently increases information gain, and mapping of similar features to the same landmark reduces the search space to improve search efficiency. Empirical evidence from experiments on real datasets shows better performance and accuracy than other approaches. Experiments are further performed by integrating the image retrieval technique into a 3D real-time navigation system. This system is tested in several indoor environments and all experiments show highly accurate localization results.","PeriodicalId":385422,"journal":{"name":"2014 Canadian Conference on Computer and Robot Vision","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2014-05-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125138081","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}
Terrance Devries, Kumar Biswaranjan, Graham W. Taylor
Recently, deep neural networks have been shown to perform competitively on the task of predicting facial expression from images. Trained by gradient-based methods, these networks are amenable to "multi-task" learning via a multiple term objective. In this paper we demonstrate that learning representations to predict the position and shape of facial landmarks can improve expression recognition from images. We show competitive results on two large-scale datasets, the ICML 2013 Facial Expression Recognition challenge, and the Toronto Face Database.
{"title":"Multi-task Learning of Facial Landmarks and Expression","authors":"Terrance Devries, Kumar Biswaranjan, Graham W. Taylor","doi":"10.1109/CRV.2014.21","DOIUrl":"https://doi.org/10.1109/CRV.2014.21","url":null,"abstract":"Recently, deep neural networks have been shown to perform competitively on the task of predicting facial expression from images. Trained by gradient-based methods, these networks are amenable to \"multi-task\" learning via a multiple term objective. In this paper we demonstrate that learning representations to predict the position and shape of facial landmarks can improve expression recognition from images. We show competitive results on two large-scale datasets, the ICML 2013 Facial Expression Recognition challenge, and the Toronto Face Database.","PeriodicalId":385422,"journal":{"name":"2014 Canadian Conference on Computer and Robot Vision","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2014-05-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"117035602","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}
Defect detection approaches based on template differencing require precise alignment of the input and template image, however, such alignment is easily affected by the presence of defects. Often, non-trivial pre/post-processing steps and/or manual parameter tuning are needed to remove false alarms, complicating the system and hampering automation. In this work, we explicitly address alignment and defect extraction jointly, and provide a general iterative algorithm to improve both their performance to pixel-wise accuracy. We achieve this by utilizing and extending the robust rank minimization and alignment method of [12]. We propose an effective and efficient optimization algorithm to decompose a template-guided image matrix into a low-rank part relating to alignment-refined defect-free images and an explicit error component containing the defects of interest. Our algorithm is fully automatic, training-free, only needs trivial pre/post-processing procedures, and has few parameters. The rank minimization formulation only requires a linearly correlated template image, and a template-guided approach relieves the common assumption of small defects, making our system very general. We demonstrate the performance of our novel approach qualitatively and quantitatively on a real-world data-set with defects of varying appearance.
{"title":"Direct Matrix Factorization and Alignment Refinement: Application to Defect Detection","authors":"Zhen Qin, P. Beek, Xu Chen","doi":"10.1109/CRV.2014.26","DOIUrl":"https://doi.org/10.1109/CRV.2014.26","url":null,"abstract":"Defect detection approaches based on template differencing require precise alignment of the input and template image, however, such alignment is easily affected by the presence of defects. Often, non-trivial pre/post-processing steps and/or manual parameter tuning are needed to remove false alarms, complicating the system and hampering automation. In this work, we explicitly address alignment and defect extraction jointly, and provide a general iterative algorithm to improve both their performance to pixel-wise accuracy. We achieve this by utilizing and extending the robust rank minimization and alignment method of [12]. We propose an effective and efficient optimization algorithm to decompose a template-guided image matrix into a low-rank part relating to alignment-refined defect-free images and an explicit error component containing the defects of interest. Our algorithm is fully automatic, training-free, only needs trivial pre/post-processing procedures, and has few parameters. The rank minimization formulation only requires a linearly correlated template image, and a template-guided approach relieves the common assumption of small defects, making our system very general. We demonstrate the performance of our novel approach qualitatively and quantitatively on a real-world data-set with defects of varying appearance.","PeriodicalId":385422,"journal":{"name":"2014 Canadian Conference on Computer and Robot Vision","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2014-05-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114339417","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}
Only a small part of the large intensities interval found in high dynamic range scenes can be captured with usual image sensors. This is why delivered images may contain under or overexposed pixels. A popular approach to overcome this problem is to take several images using different exposure parameters, and then fuse them into one single image. This exposure fusion is mostly performed as a weighted average between the corresponding pixels. The challenge is to find weights that produce best fused image quality and in a minimum amount of operations to meet real time requirements. In this paper we present a supervised learning method to estimate generalized exposure fusion weights and we demonstrate how they can be used to fuse any exposures very fast. Subjective and objective comparisons with some relevant works are conducted to prove the effectiveness of the proposed method.
{"title":"Generalized Exposure Fusion Weights Estimation","authors":"Mohammed Elamine Moumene, R. Nourine, D. Ziou","doi":"10.1109/CRV.2014.8","DOIUrl":"https://doi.org/10.1109/CRV.2014.8","url":null,"abstract":"Only a small part of the large intensities interval found in high dynamic range scenes can be captured with usual image sensors. This is why delivered images may contain under or overexposed pixels. A popular approach to overcome this problem is to take several images using different exposure parameters, and then fuse them into one single image. This exposure fusion is mostly performed as a weighted average between the corresponding pixels. The challenge is to find weights that produce best fused image quality and in a minimum amount of operations to meet real time requirements. In this paper we present a supervised learning method to estimate generalized exposure fusion weights and we demonstrate how they can be used to fuse any exposures very fast. Subjective and objective comparisons with some relevant works are conducted to prove the effectiveness of the proposed method.","PeriodicalId":385422,"journal":{"name":"2014 Canadian Conference on Computer and Robot Vision","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2014-05-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129114664","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}
In this paper, a novel algorithm for finding the optimal correspondence between two sets of image features has been introduced. The proposed algorithm pays attention not only to the similarity between features but also to the spatial layout of every matched feature and its neighbors. Unlike related methods that use geometrical relations between the neighboring features, the proposed method employees topology that survives against different types of deformations like scaling and rotation, resulting in more robust matching. The features are expressed as an undirected graph where every node represents a local feature and every edge represents adjacency between them. The topology of the resulting graph can be considered as a robust global feature of the represented object. The matching process is modeled as a graph matching problem, which in turn is formulated as a variation of the quadratic assignment problem. In this variation, a number of parameters are used to control the significance of global vs. local features to tune the performance and customize the model. The experimental results show a significant improvement in the number of correct matches using the proposed method compared to different methods.
{"title":"A More Robust Feature Correspondence for more Accurate Image Recognition","authors":"Shady Y. El-Mashad, A. Shoukry","doi":"10.1109/CRV.2014.32","DOIUrl":"https://doi.org/10.1109/CRV.2014.32","url":null,"abstract":"In this paper, a novel algorithm for finding the optimal correspondence between two sets of image features has been introduced. The proposed algorithm pays attention not only to the similarity between features but also to the spatial layout of every matched feature and its neighbors. Unlike related methods that use geometrical relations between the neighboring features, the proposed method employees topology that survives against different types of deformations like scaling and rotation, resulting in more robust matching. The features are expressed as an undirected graph where every node represents a local feature and every edge represents adjacency between them. The topology of the resulting graph can be considered as a robust global feature of the represented object. The matching process is modeled as a graph matching problem, which in turn is formulated as a variation of the quadratic assignment problem. In this variation, a number of parameters are used to control the significance of global vs. local features to tune the performance and customize the model. The experimental results show a significant improvement in the number of correct matches using the proposed method compared to different methods.","PeriodicalId":385422,"journal":{"name":"2014 Canadian Conference on Computer and Robot Vision","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2014-05-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128143857","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：