Imperial College London

Dr Panagiota (Tania) Stathaki

Faculty of EngineeringDepartment of Electrical and Electronic Engineering

Reader in Signal Processing
 
 
 
//

Contact

 

+44 (0)20 7594 6229t.stathaki Website

 
 
//

Assistant

 

Miss Vanessa Rodriguez-Gonzalez +44 (0)20 7594 6267

 
//

Location

 

812Electrical EngineeringSouth Kensington Campus

//

Summary

 

Publications

Publication Type
Year
to

167 results found

Liu T, Luo W, Ma L, Huang J-J, Stathaki T, Dai Tet al., 2020, Coupled network for robust pedestrian detection with gated multi-layer feature extraction and deformable occlusion handling, IEEE Transactions on Image Processing, Vol: 30, Pages: 754-766, ISSN: 1057-7149

Pedestrian detection methods have been significantly improved with the development of deep convolutional neural networks. Nevertheless, detecting ismall-scaled pedestrians and occluded pedestrians remains a challenging problem. In this paper, we propose a pedestrian detection method with a couple-network to simultaneously address these two issues. One of the sub-networks, the gated multi-layer feature extraction sub-network, aims to adaptively generate discriminative features for pedestrian candidates in order to robustly detect pedestrians with large variations on scale. The second sub-network targets on handling the occlusion problem of pedestrian detection by using deformable regional region of interest (RoI)-pooling. We investigate two different gate units for the gated sub-network, namely, the channel-wise gate unit and the spatio-wise gate unit, which can enhance the representation ability of the regional convolutional features among the channel dimensions or across the spatial domain, repetitively. Ablation studies have validated the effectiveness of both the proposed gated multi-layer feature extraction sub-network and the deformable occlusion handling sub-network. With the coupled framework, our proposed pedestrian detector achieves promising results on both two pedestrian datasets, especially on detecting small or occluded pedestrians. On the CityPersons dataset, the proposed detector achieves the lowest missing rates (i.e. 40.78% and 34.60%) on detecting small and occluded pedestrians, surpassing the second best comparison method by 6.0% and 5.87%, respectively.

Journal article

Ren G, Dai T, Barmpoutis P, Stathaki Tet al., 2020, Salient Object Detection Combining a Self-Attention Module and a Feature Pyramid Network, ELECTRONICS, Vol: 9

Journal article

Barmpoutis P, Stathaki T, Dimitropoulos K, Grammalidis Net al., 2020, Early Fire Detection Based on Aerial 360-Degree Sensors, Deep Convolution Neural Networks and Exploitation of Fire Dynamic Textures, REMOTE SENSING, Vol: 12

Journal article

Liu T, Huang J, Dai T, Stathaki T, Ren Get al., 2020, Gated Multi-layer Convolutional Feature Extraction Network for Robust Pedestrian Detection, ICASSP 2020, Publisher: IEEE

Conference paper

Ulku I, Barmpoutis P, Stathaki T, Akagunduz Eet al., 2020, Comparison of Single Channel Indices for U-Net Based Segmentation of Vegetation in Satellite Images, 12th International Conference on Machine Vision (ICMV), Publisher: SPIE-INT SOC OPTICAL ENGINEERING, ISSN: 0277-786X

Conference paper

Barmpoutis P, Kamperidou V, Stathaki T, 2020, Estimation of Extent of Trees' and Biomass' Infestation of the Suburban Forest of Thessaloniki (Seich Sou) using UAV Imagery and Combining R-CNNs and Multichannel Texture Analysis, 12th International Conference on Machine Vision (ICMV), Publisher: SPIE-INT SOC OPTICAL ENGINEERING, ISSN: 0277-786X

Conference paper

Protopapadakis E, Voulodimos A, Doulamis A, Doulamis N, Stathaki Tet al., 2019, Automatic crack detection for tunnel inspection using deep learning and heuristic image post-processing, APPLIED INTELLIGENCE, Vol: 49, Pages: 2793-2806, ISSN: 0924-669X

Journal article

Konstantinidis D, Stathaki T, Argyriou V, 2019, Phase amplified correlation for improved sub-pixel motion estimation, IEEE Transactions on Image Processing, Vol: 28, Pages: 3089-3101, ISSN: 1057-7149

Phase correlation (PC) is widely employed by several sub-pixel motion estimation techniques in an attempt to accurately and robustly detect the displacement between two images. To achieve sub-pixel accuracy, these techniques employ interpolation methods and function-fitting approaches on the cross-correlation function derived from the PC core. However, such motion estimation techniques still present a lower bound of accuracy that cannot be overcome. To allow room for further improvements, we propose in this paper the enhancement of the sub-pixel accuracy of motion estimation techniques by employing a completely different approach: the concept of motion magnification. To this end, we propose the novel phase amplified correlation (PAC) that integrates motion magnification between two compared images inside the phase correlation part of frequencybased motion estimation algorithms and thus directly substitutes the PC core. The experimentation on magnetic resonance (MR) images and real video sequences demonstrates the ability of the proposed PAC core to make subtle motions highly distinguishable and improve the sub-pixel accuracy of frequency-based motion estimation techniques.

Journal article

Barmpoutis P, Stathaki T, Kamperidou V, 2019, MONITORING OF TREES' HEALTH CONDITION USING A UAV EQUIPPED WITH LOW-COST DIGITAL CAMERA, 44th IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), Publisher: IEEE, Pages: 8291-8295, ISSN: 1520-6149

Conference paper

Barmpoutis P, Stathaki T, Gonzalez MI, 2019, A Region-based Fusion Scheme for Human Detection in Autonomous Navigation Applications, 45th Annual Conference of the IEEE Industrial Electronics Society (IECON), Publisher: IEEE, Pages: 5566-5571, ISSN: 1553-572X

Conference paper

Liu T, Stathaki T, 2018, Faster R-CNN for Robust Pedestrian Detection using Semantic Segmentation Network, Frontiers in Neurorobotics

Journal article

Voulodimos A, Doulamis N, Bebis G, Stathaki Tet al., 2018, Recent Developments in Deep Learning for Engineering Applications, COMPUTATIONAL INTELLIGENCE AND NEUROSCIENCE, Vol: 2018, ISSN: 1687-5265

Journal article

Alexiadis DS, Mitianoudis N, Stathaki T, 2018, Multidimensional directional steerable filters - Theory and application to 3D flow estimation, IMAGE AND VISION COMPUTING, Vol: 71, Pages: 38-67, ISSN: 0262-8856

Journal article

Stathaki P, ElMikaty M, 2018, Car detection in aerial images of dense urban areas, IEEE Transactions on Aerospace and Electronic Systems, Vol: 54, Pages: 51-63, ISSN: 0018-9251

With the ever-increasing demand in the analysis and understanding of aerial images in order to remotely recognise targets, this paper introduces a robust system for the detection and localisation of cars in images captured by air vehicles and satellites. The system adopts a sliding-window approach. It compromises a window-evaluation and a window-classification sub-systems. The performance of the proposed framework was evaluated on the Vaihingen dataset. Results demonstrate its superiority to the state of the art.

Journal article

Liu T, Stathaki T, 2017, Enhanced pedestrian detection using deep learning based semantic image segmentation, Digital Signal Processing (DSP) 2017, Publisher: IEEE

Pedestrian detection and semantic segmentation arehighly correlated tasks which can be jointly used for betterperformance. In this paper, we propose a pedestrian detectionmethod making use of semantic labeling to improve pedestriandetection results. A deep learning based semantic segmentationmethod is used to pixel-wise label images into 11 common classes.Semantic segmentation results which encodes high-level imagerepresentation are used as additional feature channels to beintegrated with the low-level HOG+LUV features. Some falsepositives, such as falsely detected pedestrians located on a tree,can be easier eliminated by making use of the semantic cues.Boosted forest is used for training the integrated feature channelsin a cascaded manner for hard negatives mining. Experimentson the Caltech-USA pedestrian dataset show improvements ondetection accuracy by using the additional semantic cues.

Conference paper

ElMikaty M, Stathaki P, 2017, Detection of cars in high-resolution aerial images of complex urban environments, IEEE Transactions on Geoscience and Remote Sensing, Vol: 55, Pages: 5913-5924, ISSN: 0196-2892

Detection of small targets, more specifically cars, in aerial images of urban scenes, has various applications in several domains, such as surveillance, military, remote sensing, and others. This is a tremendously challenging problem, mainly because of the significant interclass similarity among objects in urban environments, e.g., cars and certain types of nontarget objects, such as buildings' roofs and windows. These nontarget objects often possess very similar visual appearance to that of cars making it hard to separate the car and the noncar classes. Accordingly, most past works experienced low precision rates at high recall rates. In this paper, a novel framework is introduced that achieves a higher precision rate at a given recall than the state of the art. The proposed framework adopts a sliding-window approach and it consists of four stages, namely, window evaluation, extraction and encoding of features, classification, and postprocessing. This paper introduces a new way to derive descriptors that encode the local distributions of gradients, colors, and texture. Image descriptors characterize the aforementioned cues using adaptive cell distributions, wherein the distribution of cells within a detection window is a function of its dominant orientation, and hence, neither the rotation of the patch under examination nor the computation of descriptors at different orientations is required. The performance of the proposed framework has been evaluated on the challenging Vaihingen and Overhead Imagery Research data sets. Results demonstrate the superiority of the proposed framework to the state of the art.

Journal article

ElMikaty M, Stathaki P, 2017, Detection of cars in complex urban areas, IAPR Conference on Machine Vision Applications, Publisher: IEEE

Detection of cars in airborne images of typical urbanareas has various applications in several domains,such as surveillance, military and remote sensing. Itis a tremendously-challenging problem, mainly becauseof the significant inter-class similarity among variousobjects in urban environments. In this paper, a novelframework is introduced that adopts a sliding-windowapproach and it depicts, in a novel way, the local distributionof gradients, colours and texture. A linear supportvector machine classifier is used to differentiatebetween descriptors that belong to cars and descriptorsthat belong to other objects in a hyperspace of 3838dimensions. Descriptors are computed over a newlyproposedadaptive distribution of cells that enables theuse of various rotation-variant image descriptors. Theproposed framework has been evaluated on the Vaihingendataset and results corroborate its superiority as itachieves a higher precision for a given recall than thestate of the art.

Conference paper

Huang, Liu T, Dragotti, Stathakiet al., 2017, SRHRF+: Self-Example Enhanced Single Image Super-Resolution Using Hierarchical Random Forests, Computer Vision and Pattern Recognition Workshops

Conference paper

Liu T, Stathaki, 2017, Fast Head-Shoulder Proposal for Scare-aware Pedestrian Detection, International Conference on Pervasive Technologies Related to Assistive Environments

Conference paper

Liu T, Stathaki T, 2017, Fast Head-Shoulder Proposal for Deformable Part Model based Pedestrian Detection, International Conference on Digital Signal Processing, Publisher: IEEE, Pages: 457-461, ISSN: 2165-3577

In this paper we propose a fast head-shoulder detector as a means to facilitating faster pedestrian detection. The proposed approach is based on the observation that human head-shoulder regions share relatively robust features. The purpose is to address the problem of high computational speed of the deformable part model (DPM) detector by selecting candidate regions with higher likelihood to contain pedestrians. The proposed head-shoulder detector is based on the simple, yet effective normed gradient features. Head-shoulder detector outputs regions which are strong candidates for the presence of pedestrians and therefore, pedestrian detection processes are performed only within these regions, avoiding exhaustive sliding window search across the entire test image. Additionally, a two-pedestrian detector is applied to reinforce the detection accuracy especially in scenarios where pedestrians are close to each other. Our experiments on the INRIA dataset indicate that the proposed pedestrian detection method achieves comparable detection rate to the DPM detector, with improved speed of implementation.

Conference paper

Stathaki P, Konstantinidis D, Argyriou V, Grammalidis Net al., 2017, Building detection using enhanced HOG-LBP features and region refinement processes, IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, Vol: 10, Pages: 888-905, ISSN: 1939-1404

Building detection from two-dimensional high-resolution satellite images is a computer vision, photogrammetry, and remote sensing task that has arisen in the last decades with the advances in sensors technology and can be utilized in several applications that require the creation of urban maps or the study of urban changes. However, the variety of irrelevant objects that appear in an urban environment and resemble buildings, and the significant variations in the shape and generally the appearance of buildings render building detection a quite demanding task. As a result, automated methods that can robustly detect buildings in satellite images are necessary. To this end, we propose a building detection method that consists of two modules. The first module is a feature detector that extracts histograms of oriented gradients (HOG) and local binary patterns (LBP) from image regions. Using a novel approach, a support vector machine classifier is trained with the introduction of a special denoising distance measure for the computation of distances between HOG-LBP descriptors before their classification to the building or nonbuilding class. The second module consists of a set of region refinement processes that employs the output of the HOG-LBP detector in the form of detected rectangular image regions. Image segmentation is performed and a novel building recognition methodology is proposed to accurately identify building regions, while simultaneously discard false detections of the first module of the proposed method. We demonstrate that the proposed methodology can robustly detect buildings from satellite images and outperforms state-of-the-art building detection methods.

Journal article

Liu T, Stathaki P, Copin V, 2016, Human detection from ground truth cameras through combined use of Histogram of Oriented Gradients and body part models, International Conference and Computer Vision Theory and Applications

Conference paper

Stathaki P, Konstantinidis D, Argyriou V, Grammalidis Net al., 2016, A 3D feature for building segmentation based on shape-from-shading, International Conference and Computer Vision Theory and Applications

Conference paper

Stathaki P, Konstantinidis D, Argyriou V, Grammalidis Net al., 2016, A probabilistic feature fusion for building detection in satellite images, International Conference and Computer Vision Theory and Applications

Conference paper

Zhao X, Jiang Y, Stathaki T, Zhang Het al., 2016, Gait recognition method for arbitrary straight walking paths using appearance conversion machine, Neurocomputing, Vol: 173, Pages: 530-540, ISSN: 0925-2312

We investigate the problem of multi-view human gait recognition along any straight walking paths. It is observed that the gait appearance changes as the view changes while certain amount of correlated information exists among different views. Taking advantage of that type of correlation, a multi-view gait recognition method is proposed in this paper. First, we estimate the viewing angle of the monitor equipment in terms of the probe subject. To this end, our method considers this as a classification problem, where the classification signals are the viewing angles, and the classification features are the elements of the transformation matrix that is estimated by the Transformation Invariant Low-Rank Texture (TILT) algorithm. Then, the gallery gait appearances are converted to the view of the probe subject using the proposed Appearance Conversion Machine (ACM), where the gait features of the spatially neighbouring pixels of the gait feature are considered as the correlated information of the two views. In the end, a similarity measurement is applied on the converted gait appearance and the testing gait appearance. Experiments on the CASIA-B multi-view gait database show that the proposed gait recognition method outperforms the state-of-the-art under most views.

Journal article

Izhar LI, Elamvazuthi I, Stathaki T, Howell K, Omar Zet al., 2016, Multimodal Image Registration for Potential Diagnosis and Monitoring of Morphoea using a Hybrid NGC Method, 1st International Conference on Biomedical Engineering (IBIOMED) - Empowering Biomedical Technology for Better Future, Publisher: IEEE, Pages: 47-50

Conference paper

Zhao X, Jiang Y, Stathaki T, 2016, A novel low false alarm rate pedestrian detection framework based on single depth images, Image and Vision Computing, Vol: 45, Pages: 11-21, ISSN: 1872-8138

Pedestrian detection is an important image understanding problem with many potential applications. There has been little success in creating an algorithm which exhibits a high detection rate while keeping the false alarm in a relatively low rate. This paper presents a method designed to resolve this problem. The proposed method uses the Kinect or any similar type of sensors which facilitate the extraction of a distinct foreground. Then potential regions, which are candidates for the presence of human(s), are detected by employing the widely used Histogram of Oriented Gradients (HOG) technique, which performs well in terms of good detection rates but suffers from significantly high false alarm rates. Our method applies a sequence of operations to eliminate the false alarms produced by the HOG detector based on investigating the fine details of local shape information. Local shape information can be identified by efficient utilization of the edge points which, in this work, are used to formulate the so called Shape Context (SC) model. The proposed detection framework is divided in four sequential stages, with each stage aiming at refining the detection results of the previous stage. In addition, our approach employs a pre-evaluation stage to pre-screen and restrict further detection results. Extensive experimental results on the dataset created by the authors, involves 673 images collected from 11 different scenes, demonstrate that the proposed method eliminates a large percentage of the false alarms produced by the HOG pedestrian detector.

Journal article

Izhar LI, Stathaki T, Howell K, 2015, Global Based Thermal Image Registration for Diagnosis of Morphoea, International Conference for Innovation in Biomedical Engineering and Life Sciences (ICIBEL), Publisher: SPRINGER, Pages: 76-81, ISSN: 1680-0737

Analyzing and interpreting of thermograms have been increasingly employed in the diagnosis and monitoring of diseases thanks to its non-invasive, non-harmful nature and low cost. This paper presents a thermal image analysis system based on image registration for morphoea disease diagnosis. A novel system is proposed to improve the diagnosis and monitoring of morphoea based on integration with the published lines of Blaschko. In this application, image registration based on global and local registration methods are found inevitable. A modified normalized gradient cross-correlation (NGC) method to reduce large geometrical differences between two multimodal images of different subjects that are represented by smooth gray edge maps is proposed for the global registration approach. It is shown in this paper that the NGC method outperforms phase correlation (PC) method by a lower rate of misregistration. This demonstrates that by using the gradients of the gray edge maps, the performance of the PC based image registration method can be greatly improved.

Conference paper

Petrou ZI, Manakos I, Stathaki T, 2015, Remote sensing for biodiversity monitoring: a review of methods for biodiversity indicator extraction and assessment of progress towards international targets, Biodiversity and Conservation, Vol: 24, Pages: 2333-2363, ISSN: 1572-9710

Recognizing the imperative need for biodiversity protection, the Convention onBiological Diversity (CBD) has recently established new targets towards 2020, the so-calledAichi targets, and updated proposed sets of indicators to quantitatively monitor the progresstowards these targets. Remote sensing has been increasingly contributing to timely, accurate,and cost-effective assessment of biodiversity-related characteristics and functionsduring the last years. However, most relevant studies constitute individual research efforts,rarely related with the extraction of widely adopted CBD biodiversity indicators. Furthermore,systematic operational use of remote sensing data by managing authorities has stillbeen limited. In this study, the Aichi targets and the related CBD indicators whose monitoringcan be facilitated by remote sensing are identified. For each headline indicator anumber of recent remote sensing approaches able for the extraction of related propertiesare reviewed. Methods cover a wide range of fields, including: habitat extent and conditionmonitoring; species distribution; pressures from unsustainable management, pollution andclimate change; ecosystem service monitoring; and conservation status assessment of protectedareas. The advantages and limitations of different remote sensing data and algorithmsare discussed. Sorting of the methods based on their reported accuracies is attempted, whenpossible. The extensive literature survey aims at reviewing highly performing methods thatcan be used for large-area, effective, and timely biodiversity assessment, to encourage themore systematic use of remote sensing solutions in monitoring progress towards the Aichitargets, and to decrease the gaps between the remote sensing and management communities.

Journal article

Stathaki P, Omar Z, Hamzah N, 2015, Adaptive Chebyshev fusion of vegetation imagery based on SVM classifier, International Conference on Enhancing Collaborative Networks for the Innovation Ecosystem

Conference paper

This data is extracted from the Web of Science and reproduced under a licence from Thomson Reuters. You may not copy or re-distribute this data in whole or in part without the written consent of the Science business of Thomson Reuters.

Request URL: http://wlsprd.imperial.ac.uk:80/respub/WEB-INF/jsp/search-html.jsp Request URI: /respub/WEB-INF/jsp/search-html.jsp Query String: respub-action=search.html&id=00169011&limit=30&person=true