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.

High Precision Plasma Etch for Pattern Transfer: Towards Fluorocarbon Based Atomic Layer Etching

A basic requirement of a plasma etching process is fidelity of the patterned organic materials. In photolithography, a He plasma pretreatment (PPT) based on high ultraviolet and vacuum ultraviolet (UV/VUV) exposure was shown to be successful for roughness reduction of 193nm photoresist (PR). Typical multilayer masks consist of many other organic masking materials in addition to 193nm PR. These materials vary significantly in UV/VUV sensitivity and show, therefore, a different response to the He PPT. A delamination of the nanometer-thin, ion-induced dense amorphous carbon (DAC) layer was observed. Extensive He PPT exposure produces volatile species through UV/VUV induced scissioning. These species are trapped underneath the DAC layer in a subsequent plasma etch (PE), causing a loss of adhesion. Next to stabilizing organic materials, the major goals of this work included to establish and evaluate a cyclic fluorocarbon (FC) based approach for atomic layer etching (ALE) of SiO2 and Si; to characterize the mechanisms involved; and to evaluate the impact of processing parameters. Periodic, short precursor injections allow precise deposition of thin FC films. These films limit the amount of available chemical etchant during subsequent low energy, plasma-based Ar+ ion bombardment, resulting in strongly time-dependent etch rates. In situ ellipsometry showcased the self-limited etching. X-ray photoelectron spectroscopy (XPS) confirms FC film deposition and mixing with the substrate. The cyclic ALE approach is also able to precisely etch Si substrates. A reduced time-dependent etching is seen for Si, likely based on a lower physical sputtering energy threshold. A fluorinated, oxidized surface layer is present during ALE of Si and greatly influences the etch behavior. A reaction of the precursor with the fluorinated substrate upon precursor injection was observed and characterized. The cyclic ALE approach is transferred to a manufacturing scale reactor at IBM Research. Ensuring the transferability to industrial device patterning is crucial for the application of ALE. In addition to device patterning, the cyclic ALE process is employed for oxide removal from Si and SiGe surfaces with the goal of minimal substrate damage and surface residues. The ALE process developed for SiO2 and Si etching did not remove native oxide at the level required. Optimizing the process enabled strong O removal from the surface. Subsequent 90% H2/Ar plasma allow for removal of C and F residues.

RISK-BASED MULTIOBJECTIVE PATH PLANNING AND DESIGN OPTIMIZATION FOR UNMANNED AERIAL VEHICLES

Safe operation of unmanned aerial vehicles (UAVs) over populated areas requires reducing the risk posed by a UAV if it crashed during its operation. We considered several types of UAV risk-based path planning problems and developed techniques for estimating the risk to third parties on the ground. The path planning problem requires making trade-offs between risk and flight time. Four optimization approaches for solving the problem were tested; a network-based approach that used a greedy algorithm to improve the original solution generated the best solutions with the least computational effort. Additionally, an approach for solving a combined design and path planning problems was developed and tested. This approach was extended to solve robust risk-based path planning problem in which uncertainty about wind conditions would affect the risk posed by a UAV.

CERVIXCHECK: A SPIRITUALLY-BASED SMS TEXT MESSAGING PILOT INTERVENTION TO INCREASE CERVICAL CANCER AWARENESS AND PAP TEST SCREENING INTENTION AMONG AFRICAN AMERICAN WOMEN

African American women account for a disproportionate burden of cervical cancer incidence and mortality rate when compared to non-Hispanic White women. Cervical cancer is one of the most preventable types of cancer, and women can be screened for it with a routine Pap test. Given that religion occupies an essential place in African American lives, framing health messages with important spiritual themes and delivering them through a popular communication delivery channel may allow for a more culturally-relevant and accessible technology-based approach to promoting cervical cancer educational content to African American women. Using community-engaged research as a framework, the purpose of this multiple methods study was to develop, pilot test, and evaluate the feasibility, acceptability, and initial efficacy of a spiritually-based SMS text messaging intervention to increase cervical cancer awareness and Pap test screening intention among African American women. The study recruited church-attending African American women ages 21-65 and was conducted in three phases. Phases 1 and 2 consisted of a series of focus group discussions (n=15), cognitive response interviews (n=8), and initial usability testing that were conducted to inform the intervention development and modifications. Phase 3 utilized a non-experimental one-group pretest-posttest design to pilot test the 16-day text messaging intervention (n=52). Of the individuals enrolled, forty-six completed the posttest (retention rate=88%). Findings provided evidence for the early feasibility, high acceptability, and some initial efficacy of the CervixCheck intervention. There were significant pre-post increases observed for knowledge about cervical cancer and the Pap test (p = .001) and subjective norms (p = .006). Additionally, results post-intervention revealed that 83% of participants reported being either “satisfied” or “very satisfied” with the program and 85% found the text messages either “useful” or “very useful”. 85% of the participants also indicated that they would “likely” or “very likely” share the information they learned from the intervention with the women around them, with 39% indicating that they had already shared some of the information they received with others they knew. A spiritually-based SMS text messaging intervention could be a culturally appropriate and cost-effective method of promoting cervical cancer early detection information to African American women.