Xueyun Chen, Shiming Xiang, Cheng-Lin Liu, Chunhong Pan
Deep convolutional Neural Networks (DNN) is the state-of-the-art machine learning method. It has been used in many recognition tasks including handwritten digits, Chinese words and traffic signs, etc. However, training and test DNN are time-consuming tasks. In practical vehicle detection application, both speed and accuracy are required. So increasing the speeds of DNN while keeping its high accuracy has significant meaning for many recognition and detection applications. We introduce parallel branches into the DNN. The maps of the layers of DNN are divided into several parallel branches, each branch has the same number of maps. There are not direct connections between different branches. Our parallel DNN (PNN) keeps the same structure and dimensions of the DNN, reducing the total number of connections between maps. The more number of branches we divide, the more swift the speed of the PNN is, the conventional DNN becomes a special form of PNN which has only one branch. Experiments on large vehicle database showed that the detection accuracy of PNN dropped slightly with the speed increasing. Even the fastest PNN (10 times faster than DNN), whose branch has only two maps, fully outperformed the traditional methods based on features (such as HOG, LBP). In fact, PNN provides a good solution way for compromising the speed and accuracy requirements in many applications.
{"title":"Vehicle Detection in Satellite Images by Parallel Deep Convolutional Neural Networks","authors":"Xueyun Chen, Shiming Xiang, Cheng-Lin Liu, Chunhong Pan","doi":"10.1109/ACPR.2013.33","DOIUrl":"https://doi.org/10.1109/ACPR.2013.33","url":null,"abstract":"Deep convolutional Neural Networks (DNN) is the state-of-the-art machine learning method. It has been used in many recognition tasks including handwritten digits, Chinese words and traffic signs, etc. However, training and test DNN are time-consuming tasks. In practical vehicle detection application, both speed and accuracy are required. So increasing the speeds of DNN while keeping its high accuracy has significant meaning for many recognition and detection applications. We introduce parallel branches into the DNN. The maps of the layers of DNN are divided into several parallel branches, each branch has the same number of maps. There are not direct connections between different branches. Our parallel DNN (PNN) keeps the same structure and dimensions of the DNN, reducing the total number of connections between maps. The more number of branches we divide, the more swift the speed of the PNN is, the conventional DNN becomes a special form of PNN which has only one branch. Experiments on large vehicle database showed that the detection accuracy of PNN dropped slightly with the speed increasing. Even the fastest PNN (10 times faster than DNN), whose branch has only two maps, fully outperformed the traditional methods based on features (such as HOG, LBP). In fact, PNN provides a good solution way for compromising the speed and accuracy requirements in many applications.","PeriodicalId":365633,"journal":{"name":"2013 2nd IAPR Asian Conference on Pattern Recognition","volume":"314 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2013-11-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122751171","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}
G. Schaefer, B. Krawczyk, M. E. Celebi, H. Iyatomi
Malignant melanoma, the deadliest form of skin cancer, is one of the most rapidly increasing cancers in the world. Early diagnosis is crucial, since if detected early, it can be cured through a simple excision. In this paper, we present an effective approach to melanoma classification from dermoscopic images of skin lesions. First, we perform automatic border detection to delineate the lesion from the background skin. Shape features are then extracted from this border, while colour and texture features are obtained based on a division of the image into clinically significant regions. The derived features are then used in a pattern classification stage for which we employ a dedicated ensemble learning approach to address the class imbalance in the training data. Our classifier committee trains individual classifiers on balanced subspaces, removes redundant predictors based on a diversity measure and combines the remaining classifiers using a neural network fuser. Experimental results on a large dataset of dermoscopic skin lesion images show our approach to work well, to provide both high sensitivity and specificity, and the use of our classifier ensemble to lead to statistically better recognition performance.
{"title":"Melanoma Classification Using Dermoscopy Imaging and Ensemble Learning","authors":"G. Schaefer, B. Krawczyk, M. E. Celebi, H. Iyatomi","doi":"10.1109/ACPR.2013.102","DOIUrl":"https://doi.org/10.1109/ACPR.2013.102","url":null,"abstract":"Malignant melanoma, the deadliest form of skin cancer, is one of the most rapidly increasing cancers in the world. Early diagnosis is crucial, since if detected early, it can be cured through a simple excision. In this paper, we present an effective approach to melanoma classification from dermoscopic images of skin lesions. First, we perform automatic border detection to delineate the lesion from the background skin. Shape features are then extracted from this border, while colour and texture features are obtained based on a division of the image into clinically significant regions. The derived features are then used in a pattern classification stage for which we employ a dedicated ensemble learning approach to address the class imbalance in the training data. Our classifier committee trains individual classifiers on balanced subspaces, removes redundant predictors based on a diversity measure and combines the remaining classifiers using a neural network fuser. Experimental results on a large dataset of dermoscopic skin lesion images show our approach to work well, to provide both high sensitivity and specificity, and the use of our classifier ensemble to lead to statistically better recognition performance.","PeriodicalId":365633,"journal":{"name":"2013 2nd IAPR Asian Conference on Pattern Recognition","volume":"42 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2013-11-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122930802","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}
Tadashi Hyuga, H. Wada, Tomoyoshi Aizawa, Yoshihisa Ijiri, M. Kawade
In this demonstration, we will show our Optical Character Recognition(OCR) technique. Character deformation and touching problems often occur during high-speed printing process in the machine vision industry. As a result, it is difficult for OCR system to segment and recognize characters properly. To solve these problems, we propose a novel OCR technique which is robust against deformation and touching. It splits regions of characters simply and excessively, recognizes all segments and merged regions, and obtains optimal segments using graph theory.
{"title":"Deformed and Touched Characters Recognition","authors":"Tadashi Hyuga, H. Wada, Tomoyoshi Aizawa, Yoshihisa Ijiri, M. Kawade","doi":"10.1109/ACPR.2013.193","DOIUrl":"https://doi.org/10.1109/ACPR.2013.193","url":null,"abstract":"In this demonstration, we will show our Optical Character Recognition(OCR) technique. Character deformation and touching problems often occur during high-speed printing process in the machine vision industry. As a result, it is difficult for OCR system to segment and recognize characters properly. To solve these problems, we propose a novel OCR technique which is robust against deformation and touching. It splits regions of characters simply and excessively, recognizes all segments and merged regions, and obtains optimal segments using graph theory.","PeriodicalId":365633,"journal":{"name":"2013 2nd IAPR Asian Conference on Pattern Recognition","volume":"39 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2013-11-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127876425","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}
S. Yoshinaga, Atsushi Shimada, H. Nagahara, R. Taniguchi, Kouichiro Kajitani, Takeshi Naito
Many background models have been proposed to adapt to "illumination changes" and "dynamic changes" such as swaying motion of tree branches. However, the problem of background maintenance in complex environment, where foreground objects pass in front of stationary objects which cease moving, is still far from being completely solved. To address this problem, we propose a framework for multi-layered background modeling, in which we conserve the background models for stationary objects hierarchically in addition to the one for the initial background. To realize this framework, we also propose a spatio-temporal background model based on the similarity in the intensity changes among pixels. Experimental results on complex scenes, such as a bus stop and an intersection, show that our proposed method can adapt to both appearances and disappearances of stationary objects thanks to the multi-layered background modeling framework.
{"title":"Multi-layered Background Modeling for Complex Environment Surveillance","authors":"S. Yoshinaga, Atsushi Shimada, H. Nagahara, R. Taniguchi, Kouichiro Kajitani, Takeshi Naito","doi":"10.1109/ACPR.2013.83","DOIUrl":"https://doi.org/10.1109/ACPR.2013.83","url":null,"abstract":"Many background models have been proposed to adapt to \"illumination changes\" and \"dynamic changes\" such as swaying motion of tree branches. However, the problem of background maintenance in complex environment, where foreground objects pass in front of stationary objects which cease moving, is still far from being completely solved. To address this problem, we propose a framework for multi-layered background modeling, in which we conserve the background models for stationary objects hierarchically in addition to the one for the initial background. To realize this framework, we also propose a spatio-temporal background model based on the similarity in the intensity changes among pixels. Experimental results on complex scenes, such as a bus stop and an intersection, show that our proposed method can adapt to both appearances and disappearances of stationary objects thanks to the multi-layered background modeling framework.","PeriodicalId":365633,"journal":{"name":"2013 2nd IAPR Asian Conference on Pattern Recognition","volume":"19 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2013-11-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127767480","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}
Detecting the banknote serial number is an important task in business transaction. In this paper, we propose a new banknote number recognition method. The preprocessing of each banknote image is used to locate position of the banknote number image. Each number image is divided into non-overlapping partitions and the average gray value of each partition is used as feature vector for recognition. The optimal kernel function is obtained by the semi-definite programming (SDP). The experimental results show that the proposed method outperforms MASK, BP, HMM, Single SVM classifiers.
{"title":"New Banknote Number Recognition Algorithm Based on Support Vector Machine","authors":"S. Gai, Guowei Yang, S. Zhang, M. Wan","doi":"10.1109/ACPR.2013.115","DOIUrl":"https://doi.org/10.1109/ACPR.2013.115","url":null,"abstract":"Detecting the banknote serial number is an important task in business transaction. In this paper, we propose a new banknote number recognition method. The preprocessing of each banknote image is used to locate position of the banknote number image. Each number image is divided into non-overlapping partitions and the average gray value of each partition is used as feature vector for recognition. The optimal kernel function is obtained by the semi-definite programming (SDP). The experimental results show that the proposed method outperforms MASK, BP, HMM, Single SVM classifiers.","PeriodicalId":365633,"journal":{"name":"2013 2nd IAPR Asian Conference on Pattern Recognition","volume":"170 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2013-11-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131465893","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}
Image segmentation is a fundamental task of image processing that consists in partitioning the image by grouping pixels into homogeneous regions. We propose a novel segmentation algorithm that consists in combining many runs of a simple and fast randomized segmentation algorithm. Our algorithm also yields a soft-edge closed contour detector. We describe the theoretical probabilistic framework and report on our implementation that experimentally corroborates that performance increases with the number of runs.
{"title":"Consensus Region Merging for Image Segmentation","authors":"F. Nielsen, R. Nock","doi":"10.1109/ACPR.2013.142","DOIUrl":"https://doi.org/10.1109/ACPR.2013.142","url":null,"abstract":"Image segmentation is a fundamental task of image processing that consists in partitioning the image by grouping pixels into homogeneous regions. We propose a novel segmentation algorithm that consists in combining many runs of a simple and fast randomized segmentation algorithm. Our algorithm also yields a soft-edge closed contour detector. We describe the theoretical probabilistic framework and report on our implementation that experimentally corroborates that performance increases with the number of runs.","PeriodicalId":365633,"journal":{"name":"2013 2nd IAPR Asian Conference on Pattern Recognition","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2013-11-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115378011","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 an ObjectCode method is presented for fast template matching. Firstly, Local Binary Patterns are adopted to get the patterns for the template and the search image, respectively. Then, a selection strategy is proposed to choose a small portion of pixels (on average 1.87%) from the template, whose patterns are concatenated to form an ObjectCode representing the characteristics of the interested target region. For the candidates in the search image, we get the candidate codes using the selected pixels from the template accordingly. Finally, the similarities between the ObjectCode and the candidate codes are calculated efficiently by a new distance measure based on Hamming distance. Extensive experiments demonstrated that our method is 13.7 times faster than FFT-based template matching and 1.1 times faster than Two-stage Partial Correlation Elimination (TPCE) with similar performances, thus is a fast alternative for current template matching methods.
{"title":"A Fast Alternative for Template Matching: An ObjectCode Method","authors":"Yiping Shen, Shuxiao Li, Chenxu Wang, Hongxing Chang","doi":"10.1109/ACPR.2013.80","DOIUrl":"https://doi.org/10.1109/ACPR.2013.80","url":null,"abstract":"In this paper an ObjectCode method is presented for fast template matching. Firstly, Local Binary Patterns are adopted to get the patterns for the template and the search image, respectively. Then, a selection strategy is proposed to choose a small portion of pixels (on average 1.87%) from the template, whose patterns are concatenated to form an ObjectCode representing the characteristics of the interested target region. For the candidates in the search image, we get the candidate codes using the selected pixels from the template accordingly. Finally, the similarities between the ObjectCode and the candidate codes are calculated efficiently by a new distance measure based on Hamming distance. Extensive experiments demonstrated that our method is 13.7 times faster than FFT-based template matching and 1.1 times faster than Two-stage Partial Correlation Elimination (TPCE) with similar performances, thus is a fast alternative for current template matching methods.","PeriodicalId":365633,"journal":{"name":"2013 2nd IAPR Asian Conference on Pattern Recognition","volume":"85 1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2013-11-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115734955","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}
Indirect immunofluorescence imaging is a fundamental technique for detecting antinuclear antibodies in HEp-2 cells. This is particularly useful for the diagnosis of autoimmune diseases and other important pathological conditions involving the immune system. HEp-2 cells can be categorised into six groups: homogeneous, fine speckled, coarse speckled, nucleolar, cytoplasmic, and Centro mere cells, which give indications on different autoimmune diseases. This categorisation is typically performed by manual evaluation which is time consuming and subjective. In this paper, we present a method for automatic classification of HEp-2 cells using local binary pattern (LBP) based texture descriptors and ensemble classification. In our approach, we utilise multi-dimensional LBP (MD-LBP) histograms, which record multi-scale texture information while maintaining the relationships between the scales. Our dedicated ensemble classification approach is based on a set of heterogeneous base classifiers obtained through application of different feature selection algorithms, a diversity based pruning stage and a neural network classifier fuser. We test our algorithm on the ICPR 2012 HEp-2 contest benchmark dataset, and demonstrate it to outperform all algorithms that were entered in the competition as well as to exceed the performance of a human expert.
{"title":"HEp-2 Cell Classification Using Multi-dimensional Local Binary Patterns and Ensemble Classification","authors":"G. Schaefer, N. Doshi, B. Krawczyk","doi":"10.1109/ACPR.2013.175","DOIUrl":"https://doi.org/10.1109/ACPR.2013.175","url":null,"abstract":"Indirect immunofluorescence imaging is a fundamental technique for detecting antinuclear antibodies in HEp-2 cells. This is particularly useful for the diagnosis of autoimmune diseases and other important pathological conditions involving the immune system. HEp-2 cells can be categorised into six groups: homogeneous, fine speckled, coarse speckled, nucleolar, cytoplasmic, and Centro mere cells, which give indications on different autoimmune diseases. This categorisation is typically performed by manual evaluation which is time consuming and subjective. In this paper, we present a method for automatic classification of HEp-2 cells using local binary pattern (LBP) based texture descriptors and ensemble classification. In our approach, we utilise multi-dimensional LBP (MD-LBP) histograms, which record multi-scale texture information while maintaining the relationships between the scales. Our dedicated ensemble classification approach is based on a set of heterogeneous base classifiers obtained through application of different feature selection algorithms, a diversity based pruning stage and a neural network classifier fuser. We test our algorithm on the ICPR 2012 HEp-2 contest benchmark dataset, and demonstrate it to outperform all algorithms that were entered in the competition as well as to exceed the performance of a human expert.","PeriodicalId":365633,"journal":{"name":"2013 2nd IAPR Asian Conference on Pattern Recognition","volume":"40 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2013-11-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125335923","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}
When we act in a group with family members, friends, colleagues, each group member often play the respective role to achieve a goal that all group members have in common. This paper focuses on leadership among various kinds of roles observed in a social group and proposes a method to estimate a leader based on an interaction analysis. In order to estimate a leader in a group, we extract pointing actions of each person and measure how other people change their actions triggered by the pointing actions, i.e. how much influence the pointing actions have. When we can see the tendency that one specific person makes pointing actions and the actions have a high influence on another member, it is very likely that the person is a leader in a group. The proposed method is based on this intuition and measures the influence of pointing actions using their motion trajectories. We demonstrate that the proposed method has a potential for estimating the leadership through a comparison between the computed influence measures and subjective evaluations using some actual videos taken in a science museum.
{"title":"Group Leadership Estimation Based on Influence of Pointing Actions","authors":"H. Habe, K. Kajiwara, Ikuhisa Mitsugami, Y. Yagi","doi":"10.1109/ACPR.2013.181","DOIUrl":"https://doi.org/10.1109/ACPR.2013.181","url":null,"abstract":"When we act in a group with family members, friends, colleagues, each group member often play the respective role to achieve a goal that all group members have in common. This paper focuses on leadership among various kinds of roles observed in a social group and proposes a method to estimate a leader based on an interaction analysis. In order to estimate a leader in a group, we extract pointing actions of each person and measure how other people change their actions triggered by the pointing actions, i.e. how much influence the pointing actions have. When we can see the tendency that one specific person makes pointing actions and the actions have a high influence on another member, it is very likely that the person is a leader in a group. The proposed method is based on this intuition and measures the influence of pointing actions using their motion trajectories. We demonstrate that the proposed method has a potential for estimating the leadership through a comparison between the computed influence measures and subjective evaluations using some actual videos taken in a science museum.","PeriodicalId":365633,"journal":{"name":"2013 2nd IAPR Asian Conference on Pattern Recognition","volume":"55 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2013-11-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122024960","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}
Natural image matting is a useful and challenging task when processing image or editing video. It aims at solving the problem of accurately extracting the foreground object of arbitrary shape from an image by use of user-provided extra information, such as trimap. In this paper, we present a new sampling criterion based on random search for image matting. This improved random search algorithm can effectively avoid leaving good samples out and can also deal well with the relation between nearby samples and distant samples. In addition, an effective cost function is adopted to evaluate the candidate samples. The experimental results show that our method can produce high-quality mattes.
{"title":"Improving Sampling Criterion for Alpha Matting","authors":"Jun Cheng, Z. Miao","doi":"10.1109/ACPR.2013.145","DOIUrl":"https://doi.org/10.1109/ACPR.2013.145","url":null,"abstract":"Natural image matting is a useful and challenging task when processing image or editing video. It aims at solving the problem of accurately extracting the foreground object of arbitrary shape from an image by use of user-provided extra information, such as trimap. In this paper, we present a new sampling criterion based on random search for image matting. This improved random search algorithm can effectively avoid leaving good samples out and can also deal well with the relation between nearby samples and distant samples. In addition, an effective cost function is adopted to evaluate the candidate samples. The experimental results show that our method can produce high-quality mattes.","PeriodicalId":365633,"journal":{"name":"2013 2nd IAPR Asian Conference on Pattern Recognition","volume":"12 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2013-11-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123384406","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：