Matthew N. Dailey

Shot boundary detection from videos using entropy and local descriptor

Video shot segmentation is an important step in key frame selection, video copy detection, video summarization, and video indexing for retrieval. Although some types of video data, e.g., live sports coverage, have abrupt shot boundaries that are easy to identify using simple heuristics, it is much more difficult to identify shot boundaries in other types such as cinematic movies. We propose an algorithm for shot boundary detection able to accurately identify not only abrupt shot boundaries, but also the fade-in and fade-out boundaries typical of cinematic movies. The algorithm is based on analysis of changes in the entropy of the gray scale intensity over consecutive frames and analysis of correspondences between SURF features over consecutive frames. In an experimental evaluation on the TRECVID-2007 shot boundary test set, the algorithm achieves substantial improvements over state of the art methods, with a precision of 97.8% and a recall of 99.3%.

Machine vision for counting fruit on mango tree canopies

Machine vision technologies hold the promise of enabling rapid and accurate fruit crop yield predictions in the field. The key to fulfilling this promise is accurate segmentation and detection of fruit in images of tree canopies. This paper proposes two new methods for automated counting of fruit in images of mango tree canopies, one using texture-based dense segmentation and one using shape-based fruit detection, and compares the use of these methods relative to existing techniques:—(i) a method based on K-nearest neighbour pixel classification and contour segmentation, and (ii) a method based on super-pixel over-segmentation and classification using support vector machines. The robustness of each algorithm was tested on multiple sets of images of mango trees acquired over a period of 3xa0years. These image sets were acquired under varying conditions (light and exposure), distance to the tree, average number of fruit on the tree, orchard and season. For images collected under the same conditions as the calibration images, estimated fruit numbers were within 16xa0% of actual fruit numbers, and the F1 measure of detection performance was above 0.68 for these methods. Results were poorer when models were used for estimating fruit numbers in trees of different canopy shape and when different imaging conditions were used. For fruit-background segmentation, K-nearest neighbour pixel classification based on colour and smoothness or pixel classification based on super-pixel over-segmentation, clustering of dense scale invariant feature transform features into visual words and bag-of-visual-word super-pixel classification using support vector machines was more effective than simple contrast and colour based segmentation. Pixel classification was best followed by fruit detection using an elliptical shape model or blob detection using colour filtering and morphological image processing techniques. Method results were also compared using precision–recall plots. Imaging at night under artificial illumination with careful attention to maintaining constant illumination conditions is highly recommended.

Simultaneous Localization and Mapping with Stereo Vision

In the simultaneous localization and mapping (SLAM) problem, a mobile robot must build a map of its environment while simultaneously determining its location within that map. We propose a new algorithm, for visual SLAM (VSLAM), in which the robots only sensory information is video imagery. Our approach combines stereo vision with a popular sequential Monte Carlo (SMC) algorithm, the Rao-Blackwellised particle filter, to simultaneously explore multiple hypotheses about the robots six degree-of-freedom trajectory through space and maintain a distinct stochastic map for each of those candidate trajectories. We demonstrate the algorithms effectiveness in mapping a large outdoor virtual reality environment in the presence of odometry error

Texture-based fruit detection

In this paper, a technique based on texture analysis is proposed for detecting green fruits on plants. The method involves interest point feature extraction and descriptor computation, interest point classification using support vector machines, candidate fruit point mapping, morphological closing and fruit region extraction. In an empirical study using low-cost web camera sensors suitable for use in mechanized systems, 24 combinations of interest point features and interest point descriptors were evaluated on two fruit types (pineapple and bitter melon). The method is highly accurate, with single-image detection rates of 85xa0% for pineapples and 100xa0% for bitter melons. The method is thus sufficiently accurate for precise location and monitoring of textured fruit in the field. Future work will explore combination of detection and tracking for further improved results.

Steganography in Thai text

Steganography, or communication through covert channels, is desirable when the mere existence of an encrypted message might provide useful information to eavesdroppers. Text is ideal for steganography due to its ubiquity. However, text communication channels do not necessarily provide sufficient redundancy for covert communication. We propose a steganographic scheme for Thai plain text documents that exploits redundancies in the way particular vowel, diacritical, and tonal symbols are composed in TIS-620, the standard Thai character set. The scheme is blind in that the original carrier text is not required for decoding. In an experimental evaluation, we find that the message embedding scheme allows 2.2 bytes of embedded covert text per kilobyte of carrier text on average, and that the document modifications are unnoticeable by casual observers. The method is thus a practical and effective method for covert communication over Thai plain text channels.

Joint localization and target tracking with a monocular camera

Abstract Localization capabilities are necessary for autonomous robots that need to keep track of their position with respect to a surrounding environment. A pursuit robot is an autonomous robot that tracks and pursues a moving target, requiring accurate localization relative to the target’s position and obstacles in the local environment. Small unmanned ground vehicles (SUGVs) equipped with a monocular camera and wheel encoders could act as effective pursuit robots, but the noisy 2D target position and size estimates from the monocular camera will in turn lead to overly noisy 3D target pose estimates. One possible approach to relative localization for pursuit robots is, rather than simply tracking and estimating a relative robot–target position in each frame, joint localization, in which the purser and target are both localized with respect to a common reference frame. In this paper, we propose a novel method for joint localization of a pursuit robot and arbitrary target. The proposed method fuses the pursuit robot’s kinematics and the target’s dynamics in a joint state space model. We show that predicting and correcting pursuer and target trajectories simultaneously produces improved results compared to standard filters for estimating relative target trajectories in a 3D coordinate system. For visual tracking, we also introduce an adaptive histogram matching threshold for suspending tracking when the target is lost in a cluttered environment. When tracking is suspended, rather than traversing the entire image to search for a reappearance of the target, we only search the part of the image segmented by histogram backprojection and correctly reinitialize the tracker. The experimental results show that the joint localization method outperforms standard localization methods and that the visual tracker for pursuit robot can deal effectively with target occlusions.

Steganography in Thai Text

Steganography, or communication through covert channels, is desirable when the mere existence of an encrypted message might provide useful information to eavesdroppers. Text is ideal for steganography due to its ubiquity. However, text communication channels do not necessarily provide sufficient redundancy for covert communication. We propose a steganographic scheme for Thai plain text documents that exploits redundancies in the way particular vowel, diacritical, and tonal symbols are composed in TIS-620, the standard Thai character set. The scheme is blind in that the original carrier text is not required for decoding. In an experimental evaluation, we find that the message embedding scheme allows 2.2 bytes of embedded covert text per kilobyte of carrier text on average, and that the document modifications are unnoticeable by casual observers. The method is thus a practical and effective method for covert communication over Thai plain text channels.