Two-dimensional string notation for representing video sequences

Most current work on video indexing concentrates on queries which operate over high level semantic information which must be entirely composed and entered manually. We propose an indexing system which is based on spatial information about key objects in a scene. These key objects may be detected automatically, with manual supervision, and tracked through a sequence using one of a number of recently developed techniques. This representation is highly compact and allows rapid resolution of queries specified by iconic example. A number of systems have been produced which use 2D string notations to index digital image libraries. Just as 2D strings provide a compact and tractable indexing notation for digital pictures, a sequence of 2D strings might provide an index for a video or image sequence. To improve further upon this we reduce the representation to the 2D string pair representing the initial frame, and a sequence of edits to these strings. This takes advantage of the continuity between frames to further reduce the size of the notation. By representing video sequences using string edits, a notation has been developed which is compact, and allows querying on the spatial relationships of objects to be performed without rebuilding the majority of the scene. Calculating ranks of objects directly from the edit sequence allows matching with minimal calculation, thus greatly reducing search time. This paper presents the edit sequence notation and algorithms for evaluating queries over image sequences. A number of optimizations which represent a considerably saving in search time is demonstrated in the paper.

Generating comprehensible summaries of rushes sequences based on robust feature matching

This paper describes our first attempt at tackling a pilot task in Trecvid: video summarization of rushes data [3]. Our method is based on the tight clustering produced via SIFT matching. In this first attempt, we try to examine how our approach performs without complex implementation in terms of concept detection and excerpt assembly (i.e, no picture-in-picture, split screen and special transitions). Although we do not perform very well in terms of concept inclusion, we rank very well in terms of the summary being easy to understand and relevancy of included segments.

Finding the optimal temporal partitioning of video sequences

The existing techniques for shot partitioning either process each shot boundary independently or proceed sequentially. The sequential process assumes the last shot boundary is correctly detected and utilizes the shot length distribution to adapt the threshold for detecting the next boundary. These techniques are only locally optimal and suffer from the strong assumption about the correct detection of the last boundary. Addressing these fundamental issues, in this paper, we aim to find the global optimal shot partitioning by utilizing Bayesian principles to model the probability of a particular video partition being the shot partition. A computationally efficient algorithm based on Dynamic Programming is then formulated. The experimental results on a large movie set show that our algorithm performs consistently better than the best adaptive-thresholding technique commonly used for the task.

Explicit state duration HMM for abnormality detection in sequences of human activity

The importance of explicit duration modelling for classification of sequences of human activity and the reliable and timely detection of duration abnormality was highlighted. The normal classes of behavior were designed to highlight the importance of modelling duration given the limitations of the tracking system. It was found that HMM was the weakest model for classification of the unseen normal sequences with 81% accuracy. Long term abnormality was investigated by artificially varying the duration of primary activity in a randomly selected test sequence. The incorporation of duration in models of human behavior is an important consideration for systems seeking to provide cognitive support and to detect deviation in the behavorial patterns.

Towards detecting patterns of human behaviour from image sequences

This paper details research that will explore the analysis of human behaviour via video surveillance. Digital computer images will be obtained from video footage of a real world scene, and positions of people in the scene will be identified and tracked through each frame in the sequence.

The noted positions will build into a pattern of motion that can be examined and classified. It is proposed that specific events, such as panic or fight situations, will have unique, and therefore identifying, characteristics that will enable automatic detection of such events.

It is envisaged that active cameras will be used when a situation of interest occurs, to enable more information to be extracted from the scene (e.g., panning to follow action, or zooming to enhance detail.)

Mining condensed sets of frequent episodes with more accurate frequencies from complex sequences

Many previous approaches to frequent episode discovery only accept simple sequences. Although a recent approach has been able to nd frequent episodes from complex sequences, the discovered sets are neither condensed nor accurate. This paper investigates the discovery of condensed sets of frequent episodes from complex sequences. We adopt a novel anti-monotonic frequency measure based on non-redundant occurrences, and dene a condensed set, nDaCF (the set of non-derivable approximately closed frequent episodes) within a given maximal error bound of support. We then introduce a series of effective pruning strategies, and develop a method, nDaCF-Miner, for discovering nDaCF sets. Experimental results show that, when the error bound is somewhat high, the discovered nDaCF sets are two orders of magnitude smaller than complete sets, and nDaCF-miner is more efficient than previous mining approaches. In addition, the nDaCF sets are more accurate than the sets found by previous approaches.

Effects of alterations of primer-binding site sequences on human immunodeficiency virus type 1 replication

The human immunodeficiency virus type 1 genomic RNA primer-binding site (PBS) sequence comprises 18 nucleotides which are complementary to those at the 3' end of the replication initiation primer tRNA(3Lys). To investigate the role of the PBS in viral replication, we either deleted the original wild-type PBS (complementary to tRNA(3Lys) or replaced it with DNA sequences complementary to either tRNA(1,2Lys) or tRNA(Phe). Transfection of COS cells with such molecular constructs yielded similar levels of viral progeny that were indistinguishable with regard to viral proteins and tRNA content. Virus particles derived from PBS-deleted molecular clones were noninfectious for MT-4, Jurkat, and CEM-T4 cells. However, infectious viruses were derived from constructs in which the PBS had been altered to sequences complementary to either tRNA(1,2Lys) or tRNA(Phe), although mutated forms showed significant lags in replication efficiency in comparison with wild types. Molecular analysis of reverse-transcribed DNA in cells infected by the mutated viruses indicated that both tRNA(1,2Lys) and tRNA(Phe) could function as primers for reverse transcription during the early stages of infection. Sequencing of full-length proviral DNA, obtained 6 days after infection, revealed the mutated PBS, indicating that a complete cycle of reverse transcription had occurred. During subsequent rounds of infection, reversion of the mutated PBS to wild-type sequences was observed, accompanied by increased production of viral gene products. Reversion to wild-type PBS sequences was confirmed both by specific PCR analysis, using distinct primer pairs, and by direct sequencing of amplified segments. We also performed endogenous in vitro reverse transcription experiments in which synthesis of minus-strand strong-stop viral DNA was primed from a synthetic RNA template containing a PBS complementary to various tRNA isoacceptors. These results showed that tRNA(3Lys) was a much more efficient primer of such reactions than either tRNA(1,2Lys) or tRNA(Phe).

The A-rich RNA sequences of HIV-1 pol are important for the synthesis of viral cDNA

The bias of A-rich codons in HIV-1 pol is thought to be a record of hypermutations in viral genomes that lack biological functions. Bioinformatic analysis predicted that A-rich sequences are generally associated with minimal local RNA structures. Using codon modifications to reduce the amount of A-rich sequences within HIV-1 genomes, we have reduced the flexibility of RNA sequences in pol to analyze the functional significance of these A-rich ‘structurally poor’ RNA elements in HIV-1 pol. Our data showed that codon modification of HIV-1 sequences led to a suppression of virus infectivity by 5–100-fold, and this defect does not correlate with, viral entry, viral protein expression levels, viral protein profiles or virion packaging of genomic RNA. Codon modification of HIV-1 pol correlated with an enhanced dimer stability of the viral RNA genome, which was associated with a reduction of viral cDNA synthesis both during HIV-1 infection and in a cell free reverse transcription assay. Our data provided direct evidence that the HIV-1 A-rich pol sequence is not merely an evolutionary artifact of enzyme-induced hypermutations, and that HIV-1 has adapted to rely on A-rich RNA sequences to support the synthesis of viral cDNA during reverse transcription, highlighting the utility of using ‘structurally poor’ RNA domains in regulating biological process.

Evidence from genetic and Lagrangian drifter data for transatlantic transport of small juvenile green turtles

Aim  A key life-history component for many animals is the need for movement between different geographical locations at particular times. Green turtle (Chelonia mydas) hatchlings disperse from their natal location to spend an early pelagic stage in the ocean, followed by a neritic stage where small juveniles settle in coastal areas. In this study, we combined genetic and Lagrangian drifter data to investigate the connectivity between natal and foraging locations. In particular we focus on the evidence for transatlantic transport. Location  Atlantic Ocean.

Methods  We used mitochondrial DNA (mtDNA) sequences (n = 1567) from foraging groups (n = 8) and nesting populations (n = 12) on both sides of the Atlantic. Genetic data were obtained for Cape Verde juvenile turtles, a foraging group not previously sampled for genetic study. Various statistical methods were used to explore spatial genetics and population genetic structure (e.g. exact tests of differentiation, Geneland and analysis of molecular variance). Many-to-many mixed stock analysis estimated the connectivity between nesting and foraging groups.

Results  Our key new finding is robust evidence for connectivity between a nesting population on the South American coast (25% of the Surinam nesting population are estimated to go to Cape Verde) and a foraging group off the coast of West Africa (38% of Cape Verde juveniles are estimated to originate from Surinam), thus extending the results of previous investigations by confirming that there is substantial transatlantic dispersal in both directions. Lagrangian drifter data demonstrated that transport by drift across the Atlantic within a few years is possible.

Main conclusions  Small juvenile green turtles seem capable of dispersing extensively, and can drop out of the pelagic phase on a transatlantic scale (the average distance between natal and foraging locations was 3048 km). Nevertheless, we also find support for the ‘closest-to-home’ hypothesis in that the degree of contribution from a nesting population to a foraging group is correlated with proximity. Larger-sized turtles appear to feed closer to their natal breeding grounds (the average distance was 1133 km), indicating that those that have been initially transported to far-flung foraging grounds may still be able to move nearer to home as they grow larger.