Pub Date : 2018-10-01DOI: 10.1109/AIPR.2018.8707373
Aleem Khaliq, M. Musci, M. Chiaberge
For maize crop, biophysical parameters such as canopy height and above ground biomass are the crucial agro-ecological indicator that can be used to describe the crop growth, photosynthetic efficiency and carbon stock. Remote sensing is widely used approach and most appropriate source in terms of area coverage that can be used to monitor vegetative conditions over the large area. In this study, sentinel-2 multispectral imagery is used to calculate spectral vegetation indices over the different maize growth period using some visible bands including near infrared spectrum. The relationship has been established and analyzed between maize biophysical variables (height of the canopy and above ground biomass) collected during the field measurements and derived spectral vegetation indices using simple linear regression and pearson correlation to exploit the possibility of using satellite imagery for estimation of crop biophysical parameters.
{"title":"Analyzing relationship between maize height and spectral indices derived from remotely sensed multispectral imagery","authors":"Aleem Khaliq, M. Musci, M. Chiaberge","doi":"10.1109/AIPR.2018.8707373","DOIUrl":"https://doi.org/10.1109/AIPR.2018.8707373","url":null,"abstract":"For maize crop, biophysical parameters such as canopy height and above ground biomass are the crucial agro-ecological indicator that can be used to describe the crop growth, photosynthetic efficiency and carbon stock. Remote sensing is widely used approach and most appropriate source in terms of area coverage that can be used to monitor vegetative conditions over the large area. In this study, sentinel-2 multispectral imagery is used to calculate spectral vegetation indices over the different maize growth period using some visible bands including near infrared spectrum. The relationship has been established and analyzed between maize biophysical variables (height of the canopy and above ground biomass) collected during the field measurements and derived spectral vegetation indices using simple linear regression and pearson correlation to exploit the possibility of using satellite imagery for estimation of crop biophysical parameters.","PeriodicalId":230582,"journal":{"name":"2018 IEEE Applied Imagery Pattern Recognition Workshop (AIPR)","volume":"20 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122261054","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 : 2018-10-01DOI: 10.1109/AIPR.2018.8707405
G. Rotich, Sathyanarayanan N. Aakur, R. Minetto, Maurício Pamplona Segundo, Sudeep Sarkar
The geospatial land recognition is often cast as a local-region based classification problem. We show in this work, that prior knowledge, in terms of global semantic relationships among detected regions, allows us to leverage semantics and visual features to enhance land use classification in aerial imagery. To this end, we first estimate the top-k labels for each region using an ensemble of CNNs called Hydra. Twelve different models based on two state-of-the-art CNN architectures, ResNet and DenseNet, compose this ensemble. Then, we use Grenander’s canonical pattern theory formalism coupled with the common-sense knowledge base, ConceptNet, to impose context constraints on the labels obtained by deep learning algorithms. These constraints are captured in a multi-graph representation involving generators and bonds with a flexible topology, unlike an MRF or Bayesian networks, which have fixed structures. Minimizing the energy of this graph representation results in a graphical representation of the semantics in the given image. We show our results on the recent fMoW challenge dataset. It consists of 1,047,691 images with 62 different classes of land use, plus a false detection category. The biggest improvement in performance with the use of semantics was for false detections. Other categories with significantly improved performance were: zoo, nuclear power plant, park, police station, and space facility. For the subset of fMow images with multiple bounding boxes the accuracy is 72.79% without semantics and 74.06% with semantics. Overall, without semantic context, the classification performance was 77.04%. With semantics, it reached 77.98%. Considering that less than 20% of the dataset contained more than one ROI for context, this is a significant improvement that shows the promise of the proposed approach.
{"title":"Using Semantic Relationships among Objects for Geospatial Land Use Classification","authors":"G. Rotich, Sathyanarayanan N. Aakur, R. Minetto, Maurício Pamplona Segundo, Sudeep Sarkar","doi":"10.1109/AIPR.2018.8707405","DOIUrl":"https://doi.org/10.1109/AIPR.2018.8707405","url":null,"abstract":"The geospatial land recognition is often cast as a local-region based classification problem. We show in this work, that prior knowledge, in terms of global semantic relationships among detected regions, allows us to leverage semantics and visual features to enhance land use classification in aerial imagery. To this end, we first estimate the top-k labels for each region using an ensemble of CNNs called Hydra. Twelve different models based on two state-of-the-art CNN architectures, ResNet and DenseNet, compose this ensemble. Then, we use Grenander’s canonical pattern theory formalism coupled with the common-sense knowledge base, ConceptNet, to impose context constraints on the labels obtained by deep learning algorithms. These constraints are captured in a multi-graph representation involving generators and bonds with a flexible topology, unlike an MRF or Bayesian networks, which have fixed structures. Minimizing the energy of this graph representation results in a graphical representation of the semantics in the given image. We show our results on the recent fMoW challenge dataset. It consists of 1,047,691 images with 62 different classes of land use, plus a false detection category. The biggest improvement in performance with the use of semantics was for false detections. Other categories with significantly improved performance were: zoo, nuclear power plant, park, police station, and space facility. For the subset of fMow images with multiple bounding boxes the accuracy is 72.79% without semantics and 74.06% with semantics. Overall, without semantic context, the classification performance was 77.04%. With semantics, it reached 77.98%. Considering that less than 20% of the dataset contained more than one ROI for context, this is a significant improvement that shows the promise of the proposed approach.","PeriodicalId":230582,"journal":{"name":"2018 IEEE Applied Imagery Pattern Recognition Workshop (AIPR)","volume":"24 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132580165","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 : 2018-10-01DOI: 10.1109/AIPR.2018.8707418
S. Recker, C. Gribble, M. Butkiewicz
We discuss the precision autonomous landing features of the Joint Tactical Aerial Resupply Vehicle (JTARV) platform. Autonomous navigation for aerial vehicles demands that computer vision algorithms provide not only relevant, actionable information, but that they do so in a timely manner—i.e., the algorithms must operate in real-time. This requirement for high performance dictates optimization at every level, which is the focus of our on-going research and development efforts for adding autonomous features to JTARV. Autonomous precision landing capabilities are enabled by high-performance deep learning and structure-from-motion techniques optimized for NVIDIA mobile GPUs. The system uses a single downward-facing camera to guide the vehicle to a coded photogrammetry target, ultimately enabling fully autonomous aerial resupply for troops on the ground. This paper details the system architecture and perception system design and evaluates performance on a scale vehicle. Results demonstrate that the system is capable of landing on stationary targets within relatively narrow spaces.
{"title":"Autonomous Precision Landing for the Joint Tactical Aerial Resupply Vehicle","authors":"S. Recker, C. Gribble, M. Butkiewicz","doi":"10.1109/AIPR.2018.8707418","DOIUrl":"https://doi.org/10.1109/AIPR.2018.8707418","url":null,"abstract":"We discuss the precision autonomous landing features of the Joint Tactical Aerial Resupply Vehicle (JTARV) platform. Autonomous navigation for aerial vehicles demands that computer vision algorithms provide not only relevant, actionable information, but that they do so in a timely manner—i.e., the algorithms must operate in real-time. This requirement for high performance dictates optimization at every level, which is the focus of our on-going research and development efforts for adding autonomous features to JTARV. Autonomous precision landing capabilities are enabled by high-performance deep learning and structure-from-motion techniques optimized for NVIDIA mobile GPUs. The system uses a single downward-facing camera to guide the vehicle to a coded photogrammetry target, ultimately enabling fully autonomous aerial resupply for troops on the ground. This paper details the system architecture and perception system design and evaluates performance on a scale vehicle. Results demonstrate that the system is capable of landing on stationary targets within relatively narrow spaces.","PeriodicalId":230582,"journal":{"name":"2018 IEEE Applied Imagery Pattern Recognition Workshop (AIPR)","volume":"30 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116442487","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 : 2018-10-01DOI: 10.1109/AIPR.2018.8707407
Shunsuke Kitada, Ryunosuke Kotani, H. Iyatomi
For analysing and/or understanding languages having no word boundaries based on morphological analysis such as Japanese, Chinese, and Thai, it is desirable to perform appropriate word segmentation before word embeddings. But it is inherently difficult in these languages. In recent years, various language models based on deep learning have made remarkable progress, and some of these methodologies utilizing character-level features have successfully avoided such a difficult problem. However, when a model is fed character-level features of the above languages, it often causes overfitting due to a large number of character types. In this paper, we propose a CE-CLCNN, character-level convolutional neural networks using a character encoder to tackle these problems. The proposed CE-CLCNN is an end-to-end learning model and has an image-based character encoder, i.e. the CE-CLCNN handles each character in the target document as an image. Through various experiments, we found and confirmed that our CE-CLCNN captured closely embedded features for visually and semantically similar characters and achieves state-of-the-art results on several open document classification tasks. In this paper we report the performance of our CE-CLCNN with the Wikipedia title estimation task and analyse the internal behaviour.
{"title":"End-to-End Text Classification via Image-based Embedding using Character-level Networks","authors":"Shunsuke Kitada, Ryunosuke Kotani, H. Iyatomi","doi":"10.1109/AIPR.2018.8707407","DOIUrl":"https://doi.org/10.1109/AIPR.2018.8707407","url":null,"abstract":"For analysing and/or understanding languages having no word boundaries based on morphological analysis such as Japanese, Chinese, and Thai, it is desirable to perform appropriate word segmentation before word embeddings. But it is inherently difficult in these languages. In recent years, various language models based on deep learning have made remarkable progress, and some of these methodologies utilizing character-level features have successfully avoided such a difficult problem. However, when a model is fed character-level features of the above languages, it often causes overfitting due to a large number of character types. In this paper, we propose a CE-CLCNN, character-level convolutional neural networks using a character encoder to tackle these problems. The proposed CE-CLCNN is an end-to-end learning model and has an image-based character encoder, i.e. the CE-CLCNN handles each character in the target document as an image. Through various experiments, we found and confirmed that our CE-CLCNN captured closely embedded features for visually and semantically similar characters and achieves state-of-the-art results on several open document classification tasks. In this paper we report the performance of our CE-CLCNN with the Wikipedia title estimation task and analyse the internal behaviour.","PeriodicalId":230582,"journal":{"name":"2018 IEEE Applied Imagery Pattern Recognition Workshop (AIPR)","volume":"39 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129278296","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 : 2018-10-01DOI: 10.1109/AIPR.2018.8707426
Nagaswathi Amamcherla, A. Turlapaty, B. Gokaraju
A surface electromyographic signal can provide information on neuromuscular activity and can be used as an input in a myoelectric control system for applications such as orthotic exoskeletons. In this process, a key step is to extract useful information from the EMG signals using the pattern recognition tools. Our research focus is on identification of a set of relevant features for efficient EMG signal classification. Specifically in this work, from the pre-processed myoelectric signals, we extracted auto regression coefficients, different time-domain features such as Hjorth features, integral absolute value, mean absolute value, root mean square and cepstral features. Next a subset consisting of a few selected features are fed to the multiclass SVM classifier. Using a radial basis function kernel a classification accuracy of 92.3% has been achieved.
{"title":"A Machine Learning System for Classification of EMG Signals to Assist Exoskeleton Performance","authors":"Nagaswathi Amamcherla, A. Turlapaty, B. Gokaraju","doi":"10.1109/AIPR.2018.8707426","DOIUrl":"https://doi.org/10.1109/AIPR.2018.8707426","url":null,"abstract":"A surface electromyographic signal can provide information on neuromuscular activity and can be used as an input in a myoelectric control system for applications such as orthotic exoskeletons. In this process, a key step is to extract useful information from the EMG signals using the pattern recognition tools. Our research focus is on identification of a set of relevant features for efficient EMG signal classification. Specifically in this work, from the pre-processed myoelectric signals, we extracted auto regression coefficients, different time-domain features such as Hjorth features, integral absolute value, mean absolute value, root mean square and cepstral features. Next a subset consisting of a few selected features are fed to the multiclass SVM classifier. Using a radial basis function kernel a classification accuracy of 92.3% has been achieved.","PeriodicalId":230582,"journal":{"name":"2018 IEEE Applied Imagery Pattern Recognition Workshop (AIPR)","volume":"7 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127217453","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 : 2018-10-01DOI: 10.1109/AIPR.2018.8707385
Hiroki Tani, Ryunosuke Kotani, S. Kagiwada, H. Uga, H. Iyatomi
Recent machine learning approaches have shown promising results in the field of automated plant diagnosis. However, all of the systems were designed to diagnose single infections, thus they do not assume multiple infections. In this paper, we created our original on-site cucumber leaf dataset including multiple infections to build a practical plant diagnosis system. Our dataset has a total of 48,311 cucumber leaf images (38,821 leaves infected with any of 11 kinds of diseases, 1,814 leaves infected with multiple diseases, and 7,676 healthy leaves). We developed a convolutional neural networks (CNN) classifier having the sigmoid function with a tunable threshold on each node in the last output layer. Our model attained on average a 95.5% classification accuracy on the entire dataset. On only multiple infected cases, the result was 85.9% and it accurately identified at least one disease in 1,808 out of the total of 1,814 (99.7%).
{"title":"Diagnosis of Multiple Cucumber Infections with Convolutional Neural Networks","authors":"Hiroki Tani, Ryunosuke Kotani, S. Kagiwada, H. Uga, H. Iyatomi","doi":"10.1109/AIPR.2018.8707385","DOIUrl":"https://doi.org/10.1109/AIPR.2018.8707385","url":null,"abstract":"Recent machine learning approaches have shown promising results in the field of automated plant diagnosis. However, all of the systems were designed to diagnose single infections, thus they do not assume multiple infections. In this paper, we created our original on-site cucumber leaf dataset including multiple infections to build a practical plant diagnosis system. Our dataset has a total of 48,311 cucumber leaf images (38,821 leaves infected with any of 11 kinds of diseases, 1,814 leaves infected with multiple diseases, and 7,676 healthy leaves). We developed a convolutional neural networks (CNN) classifier having the sigmoid function with a tunable threshold on each node in the last output layer. Our model attained on average a 95.5% classification accuracy on the entire dataset. On only multiple infected cases, the result was 85.9% and it accurately identified at least one disease in 1,808 out of the total of 1,814 (99.7%).","PeriodicalId":230582,"journal":{"name":"2018 IEEE Applied Imagery Pattern Recognition Workshop (AIPR)","volume":"22 2 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116593748","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 : 2018-10-01DOI: 10.1109/AIPR.2018.8707371
David Kornish, Soundararajan Ezekiel, Maria Scalzo-Cornacchia
In recent years, Deep Convolutional Neural Networks (DCNNs) have gained lots of attention and won many competitions in machine learning, object detection, image classification, and pattern recognition. The breakthroughs in the development of graphical processing units have made it possible to train DCNNs quickly for state-of-the-art tasks such as image classification, speech recognition, and many others. However, to solve complex problems, these multilayered convolutional neural networks become increasingly large, complex, and abstract. We propose methods to improve the performance of neural networks while reducing their dimensionality, enabling a better understanding of the learning process. To leverage the extensive training, as well as strengths of several pretrained models, we explored new approaches for combining features from fully connected layers of models with heterogeneous architectures. The proposed approach combines features extracted from the penultimate fully connected layer from three different DCNNs. We merge the features of all three DCNNs together and apply principal component analysis or linear discriminant analysis. Our approach aims to reduce the dimensionality of the feature vector and find the smallest feature vector dimension that can maintain the classifier performance. For this task we use a linear Support Vector Machine as a classifier. We also investigate whether it is advantageous to fuse only penultimate fully connected layers, or to perform fusion based on other fully connected layers using multiple homogenous or heterogeneous networks. The results show that the fusion method outperformed both individual networks in terms of accuracy and computational time in all of our various trial sizes. Overall our fusion methods are faster and more accurate than individual networks in both training and testing. Finally, we compared heterogeneous with homogenous fusion methods and the results show heterogeneous methods outperform homogeneous methods.
{"title":"Fusion based Heterogeneous Convolutional Neural Networks Architecture","authors":"David Kornish, Soundararajan Ezekiel, Maria Scalzo-Cornacchia","doi":"10.1109/AIPR.2018.8707371","DOIUrl":"https://doi.org/10.1109/AIPR.2018.8707371","url":null,"abstract":"In recent years, Deep Convolutional Neural Networks (DCNNs) have gained lots of attention and won many competitions in machine learning, object detection, image classification, and pattern recognition. The breakthroughs in the development of graphical processing units have made it possible to train DCNNs quickly for state-of-the-art tasks such as image classification, speech recognition, and many others. However, to solve complex problems, these multilayered convolutional neural networks become increasingly large, complex, and abstract. We propose methods to improve the performance of neural networks while reducing their dimensionality, enabling a better understanding of the learning process. To leverage the extensive training, as well as strengths of several pretrained models, we explored new approaches for combining features from fully connected layers of models with heterogeneous architectures. The proposed approach combines features extracted from the penultimate fully connected layer from three different DCNNs. We merge the features of all three DCNNs together and apply principal component analysis or linear discriminant analysis. Our approach aims to reduce the dimensionality of the feature vector and find the smallest feature vector dimension that can maintain the classifier performance. For this task we use a linear Support Vector Machine as a classifier. We also investigate whether it is advantageous to fuse only penultimate fully connected layers, or to perform fusion based on other fully connected layers using multiple homogenous or heterogeneous networks. The results show that the fusion method outperformed both individual networks in terms of accuracy and computational time in all of our various trial sizes. Overall our fusion methods are faster and more accurate than individual networks in both training and testing. Finally, we compared heterogeneous with homogenous fusion methods and the results show heterogeneous methods outperform homogeneous methods.","PeriodicalId":230582,"journal":{"name":"2018 IEEE Applied Imagery Pattern Recognition Workshop (AIPR)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129869916","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 : 2018-10-01DOI: 10.1109/AIPR.2018.8707411
Jon Patman, Sabrina C. J. Michael, Marvin M. F. Lutnesky, K. Palaniappan
Video object tracking has been used with great success in numerous applications ranging from autonomous vehicle navigation to medical image analysis. A broad and emerging domain for exploration is in the field of automatic video-based animal behavior understanding. Interesting, yet difficult-to-test hypotheses, can be evaluated by high-throughput processing of animal movements and interactions collected from both laboratory and field experiments. In this paper we describe BioSense, a new standalone software platform and user interface that provides researchers with an open-source framework for collecting and quantitatively analyzing video data characterizing animal movement and behavior (e.g. spatial location, velocity, region preference, etc.). BioSense is capable of tracking multiple objects in real-time, using various object detection methods suitable for a range of environments and animals. Real-time operation also provides a tactical approach to object tracking by allowing users the ability to manipulate the control of the software while seeing visual feedback immediately. We evaluate the capabilities of BioSense in a series of video tracking benchmarks representative of the challenges present in animal behavior research.
{"title":"BioSense: Real-Time Object Tracking for Animal Movement and Behavior Research","authors":"Jon Patman, Sabrina C. J. Michael, Marvin M. F. Lutnesky, K. Palaniappan","doi":"10.1109/AIPR.2018.8707411","DOIUrl":"https://doi.org/10.1109/AIPR.2018.8707411","url":null,"abstract":"Video object tracking has been used with great success in numerous applications ranging from autonomous vehicle navigation to medical image analysis. A broad and emerging domain for exploration is in the field of automatic video-based animal behavior understanding. Interesting, yet difficult-to-test hypotheses, can be evaluated by high-throughput processing of animal movements and interactions collected from both laboratory and field experiments. In this paper we describe BioSense, a new standalone software platform and user interface that provides researchers with an open-source framework for collecting and quantitatively analyzing video data characterizing animal movement and behavior (e.g. spatial location, velocity, region preference, etc.). BioSense is capable of tracking multiple objects in real-time, using various object detection methods suitable for a range of environments and animals. Real-time operation also provides a tactical approach to object tracking by allowing users the ability to manipulate the control of the software while seeing visual feedback immediately. We evaluate the capabilities of BioSense in a series of video tracking benchmarks representative of the challenges present in animal behavior research.","PeriodicalId":230582,"journal":{"name":"2018 IEEE Applied Imagery Pattern Recognition Workshop (AIPR)","volume":"7 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128405266","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 : 2018-10-01DOI: 10.1109/AIPR.2018.8707424
Hyun Jung, Christian Suloway, Tianyi Miao, E. Edmondson, D. Morcock, C. Deleage, Yanling Liu, Jack R. Collins, C. Lisle
Characterization of collagen deposition in immunostained images is relevant to various pathological conditions, particularly in human immunodeficiency virus (HIV) infection. Accurate segmentation of these collagens and extracting representative features of underlying diseases are important steps to achieve quantitative diagnosis. While a first order statistic derived from the segmented collagens can be useful in representing pathological evolutions at different timepoints, it fails to capture morphological changes and spatial arrangements. In this work, we demonstrate a complete pipeline for extracting key histopathology features representing underlying disease progression from histopathology whole-slide images (WSIs) via integration of deep learning and graph theory. A convolutional neural network is trained and utilized for histopathological WSI segmentation. Parallel processing is applied to convert 100K ~ 150K segmented collagen fibrils into a single collective attributed relational graph, and graph theory is applied to extract topological and relational information from the collagenous framework. Results are in good agreement with the expected pathogenicity induced by collagen deposition, highlighting potentials in clinical applications for analyzing various meshwork-structures in whole-slide histology images.
{"title":"Integration of Deep Learning and Graph Theory for Analyzing Histopathology Whole-slide Images","authors":"Hyun Jung, Christian Suloway, Tianyi Miao, E. Edmondson, D. Morcock, C. Deleage, Yanling Liu, Jack R. Collins, C. Lisle","doi":"10.1109/AIPR.2018.8707424","DOIUrl":"https://doi.org/10.1109/AIPR.2018.8707424","url":null,"abstract":"Characterization of collagen deposition in immunostained images is relevant to various pathological conditions, particularly in human immunodeficiency virus (HIV) infection. Accurate segmentation of these collagens and extracting representative features of underlying diseases are important steps to achieve quantitative diagnosis. While a first order statistic derived from the segmented collagens can be useful in representing pathological evolutions at different timepoints, it fails to capture morphological changes and spatial arrangements. In this work, we demonstrate a complete pipeline for extracting key histopathology features representing underlying disease progression from histopathology whole-slide images (WSIs) via integration of deep learning and graph theory. A convolutional neural network is trained and utilized for histopathological WSI segmentation. Parallel processing is applied to convert 100K ~ 150K segmented collagen fibrils into a single collective attributed relational graph, and graph theory is applied to extract topological and relational information from the collagenous framework. Results are in good agreement with the expected pathogenicity induced by collagen deposition, highlighting potentials in clinical applications for analyzing various meshwork-structures in whole-slide histology images.","PeriodicalId":230582,"journal":{"name":"2018 IEEE Applied Imagery Pattern Recognition Workshop (AIPR)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130385483","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 : 2018-10-01DOI: 10.1109/AIPR.2018.8707438
AL Samed, I. Karagoz, Ali Dogan
A star detection algorithm determines the position and magnitude of stars on an observed space scene. In this study, a robust star detection algorithm is presented that filters the noise out in astronomical images and accurately estimates the centroid of stars in a way that preserving their native circular shapes. The proposed algorithm suggests the usage of different filters including global and local filters as well as morphological operations. The global filter has been utilized to eliminate the blurring effect of the images due to system-induced noises with Point Spread Function (PSF) characteristics while the local filter aims to remove the noises with Gaussian distribution. The local filter should perform optimum noise reduction as well as not damaging the structure of the stars, therefore, a PCA (Principal Component Analysis) based denoising filter have been preferred to use. Although the PCA method is even good at preserving the mass integrity of stars, it may also have disruptive effects on the shape of them. Morphological operations help to restore this deformation. In order to verify the proposed algorithm, different types of noises having the Gaussian characteristics with different variance values have been inserted to astronomical star images to simulate the varied conditions of near space. Structural Similarity Index (SSIM) and Peak Signal to Noise Ratio (PSNR) parameters have been used as a performance metrics to show the accuracy of the filtering process. Furthermore, to demonstrate the overall accuracy of this method against to noise, the Mean Error of Centroid Estimation (MECE) has been achieved by means of the Monte Carlo analysis. Also, the performance of this algorithm has been compared with similar algorithms and the results show that this algorithm outperforms others.
{"title":"An Improved Star Detection Algorithm Using a Combination of Statistical and Morphological Image Processing Techniques","authors":"AL Samed, I. Karagoz, Ali Dogan","doi":"10.1109/AIPR.2018.8707438","DOIUrl":"https://doi.org/10.1109/AIPR.2018.8707438","url":null,"abstract":"A star detection algorithm determines the position and magnitude of stars on an observed space scene. In this study, a robust star detection algorithm is presented that filters the noise out in astronomical images and accurately estimates the centroid of stars in a way that preserving their native circular shapes. The proposed algorithm suggests the usage of different filters including global and local filters as well as morphological operations. The global filter has been utilized to eliminate the blurring effect of the images due to system-induced noises with Point Spread Function (PSF) characteristics while the local filter aims to remove the noises with Gaussian distribution. The local filter should perform optimum noise reduction as well as not damaging the structure of the stars, therefore, a PCA (Principal Component Analysis) based denoising filter have been preferred to use. Although the PCA method is even good at preserving the mass integrity of stars, it may also have disruptive effects on the shape of them. Morphological operations help to restore this deformation. In order to verify the proposed algorithm, different types of noises having the Gaussian characteristics with different variance values have been inserted to astronomical star images to simulate the varied conditions of near space. Structural Similarity Index (SSIM) and Peak Signal to Noise Ratio (PSNR) parameters have been used as a performance metrics to show the accuracy of the filtering process. Furthermore, to demonstrate the overall accuracy of this method against to noise, the Mean Error of Centroid Estimation (MECE) has been achieved by means of the Monte Carlo analysis. Also, the performance of this algorithm has been compared with similar algorithms and the results show that this algorithm outperforms others.","PeriodicalId":230582,"journal":{"name":"2018 IEEE Applied Imagery Pattern Recognition Workshop (AIPR)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116163541","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：