Leveraging Multiple Features for Image Retrieval and Matching

The goal of image retrieval and matching is to find and locate object instances in images from a large-scale image database. While visual features are abundant, how to combine them to improve performance by individual features remains a challenging task. In this work, we focus on leveraging multiple features for accurate and efficient image retrieval and matching. We first propose two graph-based approaches to rerank initially retrieved images for generic image retrieval. In the graph, vertices are images while edges are similarities between image pairs. Our first approach employs a mixture Markov model based on a random walk model on multiple graphs to fuse graphs. We introduce a probabilistic model to compute the importance of each feature for graph fusion under a naive Bayesian formulation, which requires statistics of similarities from a manually labeled dataset containing irrelevant images. To reduce human labeling, we further propose a fully unsupervised reranking algorithm based on a submodular objective function that can be efficiently optimized by greedy algorithm. By maximizing an information gain term over the graph, our submodular function favors a subset of database images that are similar to query images and resemble each other. The function also exploits the rank relationships of images from multiple ranked lists obtained by different features. We then study a more well-defined application, person re-identification, where the database contains labeled images of human bodies captured by multiple cameras. Re-identifications from multiple cameras are regarded as related tasks to exploit shared information. We apply a novel multi-task learning algorithm using both low level features and attributes. A low rank attribute embedding is joint learned within the multi-task learning formulation to embed original binary attributes to a continuous attribute space, where incorrect and incomplete attributes are rectified and recovered. To locate objects in images, we design an object detector based on object proposals and deep convolutional neural networks (CNN) in view of the emergence of deep networks. We improve a Fast RCNN framework and investigate two new strategies to detect objects accurately and efficiently: scale-dependent pooling (SDP) and cascaded rejection classifiers (CRC). The SDP improves detection accuracy by exploiting appropriate convolutional features depending on the scale of input object proposals. The CRC effectively utilizes convolutional features and greatly eliminates negative proposals in a cascaded manner, while maintaining a high recall for true objects. The two strategies together improve the detection accuracy and reduce the computational cost.

H∞ dynamic output feedback control of power systems with electric vehicles

This paper presents a H∞ dynamic output feedback control scheme for load frequency control (LFC) of interconnected power systems with multiple input timedelays. In this study, electric vehicles (EVs) are participated in the LFC to support reheated thermal power units to rapidly suppress load and frequency fluctuations. A mathematical model of an interconnected power system is first introduced. This model takes into consideration of the different time delays in control inputs; specifically the communication/information delays between the control center and the fleet of EVs. We then derive stabilization conditions in terms of feasible linear matrix inequalities (LMIs) for the proposed system and develop an effective algorithm to parameterize H∞ controllers ensuring stability of the closed-loop system with H∞ performance. Extensive simulations are given to show the effectiveness of the proposed control method.

Physical activity preferences for people living with multiple myeloma : a qualitative study.

Although physical activity (PA) has significant benefits for people living with multiple myeloma (MM), participation rates are low. Examination of PA preferences will provide important information to clinicians and assist in the development of interventions to increase participation in PA for people living with MM. OBJECTIVE: The aim of this study is to gain an in-depth understanding of the PA preferences for people living with MM, including the preferred role of clinicians. METHODS: Semistructured interviews were conducted with patients treated for MM within the preceding 2 to 12 months. Interviews were analyzed using content analysis, where coding categories were derived directly from the text data. RESULTS: Twenty-four interviews were conducted (women, 54%; age: mean [SD], 62 [8.8] years); 16 (67%) participants had an autologous stem cell transplant. Light- to moderate-intensity PA during and after treatment was feasible, with the strongest preference for a program 2 to 8 months after treatment. The timing of information delivery was important, as was input from clinicians and organizations with knowledge of MM. Preferences for location, structure, and timing of programs varied. CONCLUSIONS: Low- to moderate-intensity PA after treatment is likely to interest people with MM. Programs need to be flexible and consider individual differences in PA preferences, functional status, and treatment schedules. IMPLICATIONS: An individually tailored PA program should form part of clinical care, involving clinicians and organizations with expertise in MM. Options for home-based PA are also important. Further research, including a population-based study of people living with MM, is necessary to further quantify PA preferences.

Integration of electric vehicles for load frequency output feedback H ∞ control of smart grids

This study considers a novel application of electric vehicles (EVs) to quickly help reheated thermal turbine units to provide the stability fluctuated by load demands. A mathematical model of a power system with EVs is first derived. This model contains the dynamic interactions of EVs and multiple network-induced time delays. Then, a dynamic output feedback H∞ controller for load frequency control of power systems with multiple time delays in the control input is proposed. To address the multiple time delays issue, a refined Jensen-based inequality, which encompasses the Jensen inequality, is used to derive less conservative synthesis conditions in terms of tractable linear matrix inequalities. A procedure is given to parameterise an output feedback controller to guarantee stability and H∞ performance of the closed-loop system. Extensive simulations are conducted to validate the proposed control method.