Towards the retrieval of accurate OD matrices from Bluetooth data : lessons learned from 2 years of data

The Bluetooth technology is being increasingly used to track vehicles throughout their trips, within urban networks and across freeway stretches. One important opportunity offered by this type of data is the measurement of Origin-Destination patterns, emerging from the aggregation and clustering of individual trips. In order to obtain accurate estimations, however, a number of issues need to be addressed, through data filtering and correction techniques. These issues mainly stem from the use of the Bluetooth technology amongst drivers, and the physical properties of the Bluetooth sensors themselves. First, not all cars are equipped with discoverable Bluetooth devices and the Bluetooth-enabled vehicles may belong to some small socio-economic groups of users. Second, the Bluetooth datasets include data from various transport modes; such as pedestrian, bicycles, cars, taxi driver, buses and trains. Third, the Bluetooth sensors may fail to detect all of the nearby Bluetooth-enabled vehicles. As a consequence, the exact journey for some vehicles may become a latent pattern that will need to be extracted from the data. Finally, sensors that are in close proximity to each other may have overlapping detection areas, thus making the task of retrieving the correct travelled path even more challenging. The aim of this paper is twofold. We first give a comprehensive overview of the aforementioned issues. Further, we propose a methodology that can be followed, in order to cleanse, correct and aggregate Bluetooth data. We postulate that the methods introduced by this paper are the first crucial steps that need to be followed in order to compute accurate Origin-Destination matrices in urban road networks.

Detecting anomalous events at railway level crossings

Collisions between pedestrians and vehicles continue to be a major problem throughout the world. Pedestrians trying to cross roads and railway tracks without any caution are often highly susceptible to collisions with vehicles and trains. Continuous financial, human and other losses have prompted transport related organizations to come up with various solutions addressing this issue. However, the quest for new and significant improvements in this area is still ongoing. This work addresses this issue by building a general framework using computer vision techniques to automatically monitor pedestrian movements in such high-risk areas to enable better analysis of activity, and the creation of future alerting strategies. As a result of rapid development in the electronics and semi-conductor industry there is extensive deployment of CCTV cameras in public places to capture video footage. This footage can then be used to analyse crowd activities in those particular places. This work seeks to identify the abnormal behaviour of individuals in video footage. In this work we propose using a Semi-2D Hidden Markov Model (HMM), Full-2D HMM and Spatial HMM to model the normal activities of people. The outliers of the model (i.e. those observations with insufficient likelihood) are identified as abnormal activities. Location features, flow features and optical flow textures are used as the features for the model. The proposed approaches are evaluated using the publicly available UCSD datasets, and we demonstrate improved performance using a Semi-2D Hidden Markov Model compared to other state of the art methods. Further we illustrate how our proposed methods can be applied to detect anomalous events at rail level crossings.

Histogram of weighted local directions for gait recognition

In this paper, we explore the effectiveness of patch-based gradient feature extraction methods when applied to appearance-based gait recognition. Extending existing popular feature extraction methods such as HOG and LDP, we propose a novel technique which we term the Histogram of Weighted Local Directions (HWLD). These 3 methods are applied to gait recognition using the GEI feature, with classification performed using SRC. Evaluations on the CASIA and OULP datasets show significant improvements using these patch-based methods over existing implementations, with the proposed method achieving the highest recognition rate for the respective datasets. In addition, the HWLD can easily be extended to 3D, which we demonstrate using the GEV feature on the DGD dataset, observing improvements in performance.

Scene invariant multi camera crowd counting

Automated crowd counting has become an active field of computer vision research in recent years. Existing approaches are scene-specific, as they are designed to operate in the single camera viewpoint that was used to train the system. Real world camera networks often span multiple viewpoints within a facility, including many regions of overlap. This paper proposes a novel scene invariant crowd counting algorithm that is designed to operate across multiple cameras. The approach uses camera calibration to normalise features between viewpoints and to compensate for regions of overlap. This compensation is performed by constructing an 'overlap map' which provides a measure of how much an object at one location is visible within other viewpoints. An investigation into the suitability of various feature types and regression models for scene invariant crowd counting is also conducted. The features investigated include object size, shape, edges and keypoints. The regression models evaluated include neural networks, K-nearest neighbours, linear and Gaussian process regresion. Our experiments demonstrate that accurate crowd counting was achieved across seven benchmark datasets, with optimal performance observed when all features were used and when Gaussian process regression was used. The combination of scene invariance and multi camera crowd counting is evaluated by training the system on footage obtained from the QUT camera network and testing it on three cameras from the PETS 2009 database. Highly accurate crowd counting was observed with a mean relative error of less than 10%. Our approach enables a pre-trained system to be deployed on a new environment without any additional training, bringing the field one step closer toward a 'plug and play' system.

The association of single nucleotide polymorphisms with endometrial cancer risk and prognosis

Endometrial cancer is one of the most common female diseases in developed nations and is the most commonly diagnosed gynaecological cancer in Australia. The disease is commonly classified by histology: endometrioid or non-endometrioid endometrial cancer. While non-endometrioid endometrial cancers are accepted to be high-grade, aggressive cancers, endometrioid cancers (comprising 80% of all endometrial cancers diagnosed) generally carry a favourable patient prognosis. However, endometrioid endometrial cancer patients endure significant morbidity due to surgery and radiotherapy used for disease treatment, and patients with recurrent disease have a 5-year survival rate of less than 50%. Genetic analysis of women with endometrial cancer could uncover novel markers associated with disease risk and/or prognosis, which could then be used to identify women at high risk and for the use of specialised treatments. Proteases are widely accepted to play an important role in the development and progression of cancer. This PhD project hypothesised that SNPs from two protease gene families, the matrix metalloproteases (MMPs, including their tissue inhibitors, TIMPs) and the tissue kallikrein-related peptidases (KLKs) would be associated with endometrial cancer susceptibility and/or prognosis. In the first part of this study, optimisation of the genotyping techniques was performed. Results from previously published endometrial cancer genetic association studies were attempted to be validated in a large, multicentre replication set (maximum cases n = 2,888, controls n = 4,483, 3 studies). The rs11224561 progesterone receptor SNP (PGR, A/G) was observed to be associated with increased endometrial cancer risk (per A allele OR 1.31, 95% CI 1.12-1.53; p-trend = 0.001), a result which was initially reported among a Chinese sample set. Previously reported associations for the remaining 8 SNPs investigated for this section of the PhD study were not confirmed, thereby reinforcing the importance of validation of genetic association studies. To examine the effect of SNPs from the MMP and KLK families on endometrial cancer risk, we selected the most significantly associated MMP and KLK SNPs from genome-wide association study analysis (GWAS) to be genotyped in the GWAS replication set (cases n = 4,725, controls n = 9,803, 13 studies). The significance of the MMP24 rs932562 SNP was unchanged after incorporation of the stage 2 samples (Stage 1 per allele OR 1.18, p = 0.002; Combined Stage 1 and 2 OR 1.09, p = 0.002). The rs10426 SNP, located 3' to KLK10 was predicted by bioinformatic analysis to effect miRNA binding. This SNP was observed in the GWAS stage 1 result to exhibit a recessive effect on endometrial cancer risk, a result which was not validated in the stage 2 sample set (Stage 1 OR 1.44, p = 0.007; Combined Stage 1 and 2 OR 1.14, p = 0.08). Investigation of the regions imputed surrounding the MMP, TIMP and KLK genes did not reveal any significant targets for further analysis. Analysis of the case data from the endometrial cancer GWAS to identify genetic variation associated with cancer grade did not reveal SNPs from the MMP, TIMP or KLK genes to be statistically significant. However, the representation of SNPs from the MMP, TIMP and KLK families by the GWAS genotyping platform used in this PhD project was examined and observed to be very low, with the genetic variation of four genes (MMP23A, MMP23B, MMP28 and TIMP1) not captured at all by this technique. This suggests that comprehensive candidate gene association studies will be required to assess the role of SNPs from these genes with endometrial cancer risk and prognosis. Meta-analysis of gene expression microarray datasets curated as part of this PhD study identified a number of MMP, TIMP and KLK genes to display differential expression by endometrial cancer status (MMP2, MMP10, MMP11, MMP13, MMP19, MMP25 and KLK1) and histology (MMP2, MMP11, MMP12, MMP26, MMP28, TIMP2, TIMP3, KLK6, KLK7, KLK11 and KLK12). In light of these findings these genes should be prioritised for future targeted genetic association studies. Two SNPs located 43.5 Mb apart on chromosome 15 were observed from the GWAS analysis to be associated with increased endometrial cancer grade, results that were validated in silico in two independent datasets. One of these SNPs, rs8035725 is located in the 5' untranslated region of a MYC promoter binding protein DENND4A (Stage 1 OR 1.15, p = 9.85 x 10P -5 P, combined Stage 1 and in silico validation OR 1.13, p = 5.24 x 10P -6 P). This SNP has previously been reported to alter the expression of PTPLAD1, a gene involved in the synthesis of very long fatty acid chains and in the Rac1 signaling pathway. Meta-analysis of gene expression microarray data found PTPLAD1 to display increased expression in the aggressive non-endometrioid histology compared with endometrioid endometrial cancer, suggesting that the causal SNP underlying the observed genetic association may influence expression of this gene. Neither rs8035725 nor significant SNPs identified by imputation were predicted bioinformatically to affect transcription factor binding sites, indicating that further studies are required to assess their potential effect on other regulatory elements. The other grade- associated SNP, rs6606792, is located upstream of an inferred pseudogene, ELMO2P1 (Stage 1 OR 1.12, p = 5 x 10P -5 P; combined Stage 1 and in silico validation OR 1.09, p = 3.56 x 10P -5 P). Imputation of the ±1 Mb region surrounding this SNP revealed a cluster of significantly associated variants which are predicted to abolish various transcription factor binding sites, and would be expected to decrease gene expression. ELMO2P1 was not included on the microarray platforms collected for this PhD, and so its expression could not be investigated. However, the high sequence homology of ELMO2P1 with ELMO2, a gene important to cell motility, indicates that ELMO2 could be the parent gene for ELMO2P1 and as such, ELMO2P1 could function to regulate the expression of ELMO2. Increased expression of ELMO2 was seen to be associated with increasing endometrial cancer grade, as well as with aggressive endometrial cancer histological subtypes by microarray meta-analysis. Thus, it is hypothesised that SNPs in linkage disequilibrium with rs6606792 decrease the transcription of ELMO2P1, reducing the regulatory effect of ELMO2P1 on ELMO2 expression. Consequently, ELMO2 expression is increased, cell motility is enhanced leading to an aggressive endometrial cancer phenotype. In summary, these findings have identified several areas of research for further study. The results presented in this thesis provide evidence that a SNP in PGR is associated with risk of developing endometrial cancer. This PhD study also reports two independent loci on chromosome 15 to be associated with increased endometrial cancer grade, and furthermore, genes associated with these SNPs to be differentially expressed according in aggressive subtypes and/or by grade. The studies reported in this thesis support the need for comprehensive SNP association studies on prioritised MMP, TIMP and KLK genes in large sample sets. Until these studies are performed, the role of MMP, TIMP and KLK genetic variation remains unclear. Overall, this PhD study has contributed to the understanding of genetic variation involvement in endometrial cancer susceptibility and prognosis. Importantly, the genetic regions highlighted in this study could lead to the identification of novel gene targets to better understand the biology of endometrial cancer and also aid in the development of therapeutics directed at treating this disease.

Closing the case of APOE in multiple sclerosis : no association with disease risk in over 29,000 subjects

Background Single nucleotide polymorphisms (SNPs) rs429358 (ε4) and rs7412 (ε2), both invoking changes in the amino-acid sequence of the apolipoprotein E (APOE) gene, have previously been tested for association with multiple sclerosis (MS) risk. However, none of these studies was sufficiently powered to detect modest effect sizes at acceptable type-I error rates. As both SNPs are only imperfectly captured on commonly used microarray genotyping platforms, their evaluation in the context of genome-wide association studies has been hindered until recently. Methods We genotyped 12 740 subjects hitherto not studied for their APOE status, imputed raw genotype data from 8739 subjects from five independent genome wide association studies datasets using the most recent high-resolution reference panels, and extracted genotype data for 8265 subjects from previous candidate gene assessments. Results Despite sufficient power to detect associations at genome-wide significance thresholds across a range of ORs, our analyses did not support a role of rs429358 or rs7412 on MS susceptibility. This included meta-analyses of the combined data across 13 913 MS cases and 15 831 controls (OR=0.95, p=0.259, and OR 1.07, p=0.0569, for rs429358 and rs7412, respectively). Conclusion Given the large sample size of our analyses, it is unlikely that the two APOE missense SNPs studied here exert any relevant effects on MS susceptibility.

Introducing standardized protocols for anthropological measurement of virtual subadult crania using computed tomography

Objectives This study introduces and assesses the precision of a standardized protocol for anthropometric measurement of the juvenile cranium using three-dimensional surface rendered models, for implementation in forensic investigation or paleodemographic research. Materials and methods A subset of multi-slice computed tomography (MSCT) DICOM datasets (n=10) of modern Australian subadults (birth—10 years) was accessed from the “Skeletal Biology and Forensic Anthropology Virtual Osteological Database” (n>1200), obtained from retrospective clinical scans taken at Brisbane children hospitals (2009–2013). The capabilities of Geomagic Design X™ form the basis of this study; introducing standardized protocols using triangle surface mesh models to (i) ascertain linear dimensions using reference plane networks and (ii) calculate the area of complex regions of interest on the cranium. Results The protocols described in this paper demonstrate high levels of repeatability between five observers of varying anatomical expertise and software experience. Intra- and inter-observer error was indiscernible with total technical error of measurement (TEM) values ≤0.56 mm, constituting <0.33% relative error (rTEM) for linear measurements; and a TEM value of ≤12.89 mm2, equating to <1.18% (rTEM) of the total area of the anterior fontanelle and contiguous sutures. Conclusions Exploiting the advances of MSCT in routine clinical assessment, this paper assesses the application of this virtual approach to acquire highly reproducible morphometric data in a non-invasive manner for human identification and population studies in growth and development. The protocols and precision testing presented are imperative for the advancement of “virtual anthropology” into routine Australian medico-legal death investigation.

Generalised gamma bidding model

A generalised gamma bidding model is presented, which incorporates many previous models. The log likelihood equations are provided. Using a new method of testing, variants of the model are fitted to some real data for construction contract auctions to find the best fitting models for groupings of bidders. The results are examined for simplifying assumptions, including all those in the main literature. These indicate no one model to be best for all datasets. However, some models do appear to perform significantly better than others and it is suggested that future research would benefit from a closer examination of these.

Continuous appearance-based localisation and mapping

This thesis presents a novel approach to mobile robot navigation using visual information towards the goal of long-term autonomy. A novel concept of a continuous appearance-based trajectory is proposed in order to solve the limitations of previous robot navigation systems, and two new algorithms for mobile robots, CAT-SLAM and CAT-Graph, are presented and evaluated. These algorithms yield performance exceeding state-of-the-art methods on public benchmark datasets and large-scale real-world environments, and will help enable widespread use of mobile robots in everyday applications.

An evaluation of different features and learning models for anomalous event detection

The huge amount of CCTV footage available makes it very burdensome to process these videos manually through human operators. This has made automated processing of video footage through computer vision technologies necessary. During the past several years, there has been a large effort to detect abnormal activities through computer vision techniques. Typically, the problem is formulated as a novelty detection task where the system is trained on normal data and is required to detect events which do not fit the learned ‘normal’ model. There is no precise and exact definition for an abnormal activity; it is dependent on the context of the scene. Hence there is a requirement for different feature sets to detect different kinds of abnormal activities. In this work we evaluate the performance of different state of the art features to detect the presence of the abnormal objects in the scene. These include optical flow vectors to detect motion related anomalies, textures of optical flow and image textures to detect the presence of abnormal objects. These extracted features in different combinations are modeled using different state of the art models such as Gaussian mixture model(GMM) and Semi- 2D Hidden Markov model(HMM) to analyse the performances. Further we apply perspective normalization to the extracted features to compensate for perspective distortion due to the distance between the camera and objects of consideration. The proposed approach is evaluated using the publicly available UCSD datasets and we demonstrate improved performance compared to other state of the art methods.

Metrics for Understanding Communication on Twitter

As the systematic investigation of Twitter as a communications platform continues, the question of developing reliable comparative metrics for the evaluation of public, communicative phenomena on Twitter becomes paramount. What is necessary here is the establishment of an accepted standard for the quantitative description of user activities on Twitter. This needs to be flexible enough in order to be applied to a wide range of communicative situations, such as the evaluation of individual users’ and groups of users’ Twitter communication strategies, the examination of communicative patterns within hashtags and other identifiable ad hoc publics on Twitter (Bruns & Burgess, 2011), and even the analysis of very large datasets of everyday interactions on the platform. By providing a framework for quantitative analysis on Twitter communication, researchers in different areas (e.g., communication studies, sociology, information systems) are enabled to adapt methodological approaches and to conduct analyses on their own. Besides general findings about communication structure on Twitter, large amounts of data might be used to better understand issues or events retrospectively, detect issues or events in an early stage, or even to predict certain real-world developments (e.g., election results; cf. Tumasjan, Sprenger, Sandner, & Welpe, 2010, for an early attempt to do so).

Interestingness Measures for Multi-Level Association Rules

Association rule mining is one technique that is widely used when querying databases, especially those that are transactional, in order to obtain useful associations or correlations among sets of items. Much work has been done focusing on efficiency, effectiveness and redundancy. There has also been a focusing on the quality of rules from single level datasets with many interestingness measures proposed. However, with multi-level datasets now being common there is a lack of interestingness measures developed for multi-level and cross-level rules. Single level measures do not take into account the hierarchy found in a multi-level dataset. This leaves the Support-Confidence approach, which does not consider the hierarchy anyway and has other drawbacks, as one of the few measures available. In this chapter we propose two approaches which measure multi-level association rules to help evaluate their interestingness by considering the database’s underlying taxonomy. These measures of diversity and peculiarity can be used to help identify those rules from multi-level datasets that are potentially useful.

Semi-binary based video features for activity representation

Efficient and effective feature detection and representation is an important consideration when processing videos, and a large number of applications such as motion analysis, 3D scene understanding, tracking etc. depend on this. Amongst several feature description methods, local features are becoming increasingly popular for representing videos because of their simplicity and efficiency. While they achieve state-of-the-art performance with low computational complexity, their performance is still too limited for real world applications. Furthermore, rapid increases in the uptake of mobile devices has increased the demand for algorithms that can run with reduced memory and computational requirements. In this paper we propose a semi binary based feature detectordescriptor based on the BRISK detector, which can detect and represent videos with significantly reduced computational requirements, while achieving comparable performance to the state of the art spatio-temporal feature descriptors. First, the BRISK feature detector is applied on a frame by frame basis to detect interest points, then the detected key points are compared against consecutive frames for significant motion. Key points with significant motion are encoded with the BRISK descriptor in the spatial domain and Motion Boundary Histogram in the temporal domain. This descriptor is not only lightweight but also has lower memory requirements because of the binary nature of the BRISK descriptor, allowing the possibility of applications using hand held devices.We evaluate the combination of detectordescriptor performance in the context of action classification with a standard, popular bag-of-features with SVM framework. Experiments are carried out on two popular datasets with varying complexity and we demonstrate comparable performance with other descriptors with reduced computational complexity.

Towards bio-inspired place recognition over multiple spatial scales

This paper presents a new multi-scale place recognition system inspired by the recent discovery of overlapping, multi-scale spatial maps stored in the rodent brain. By training a set of Support Vector Machines to recognize places at varying levels of spatial specificity, we are able to validate spatially specific place recognition hypotheses against broader place recognition hypotheses without sacrificing localization accuracy. We evaluate the system in a range of experiments using cameras mounted on a motorbike and a human in two different environments. At 100% precision, the multiscale approach results in a 56% average improvement in recall rate across both datasets. We analyse the results and then discuss future work that may lead to improvements in both robotic mapping and our understanding of sensory processing and encoding in the mammalian brain.

Acquiring user information needs for recommender systems

Most recommender systems attempt to use collaborative filtering, content-based filtering or hybrid approach to recommend items to new users. Collaborative filtering recommends items to new users based on their similar neighbours, and content-based filtering approach tries to recommend items that are similar to new users' profiles. The fundamental issues include how to profile new users, and how to deal with the over-specialization in content-based recommender systems. Indeed, the terms used to describe items can be formed as a concept hierarchy. Therefore, we aim to describe user profiles or information needs by using concepts vectors. This paper presents a new method to acquire user information needs, which allows new users to describe their preferences on a concept hierarchy rather than rating items. It also develops a new ranking function to recommend items to new users based on their information needs. The proposed approach is evaluated on Amazon book datasets. The experimental results demonstrate that the proposed approach can largely improve the effectiveness of recommender systems.