Abstract Over recent years, many approaches have been proposed for the denoising or semantic segmentation of X-ray computed tomography (CT) scans. In most cases, high-quality CT reconstructions are used; however, such reconstructions are not always available. When the X-ray exposure time has to be limited, undersampled tomograms (in terms of their component projections) are attained. This low number of projections offers low-quality reconstructions that are difficult to segment. Here, we consider CT time-series (i.e. 4D data), where the limited time for capturing fast-occurring temporal events results in the time-series tomograms being necessarily undersampled. Fortunately, in these collections, it is common practice to obtain representative highly sampled tomograms before or after the time-critical portion of the experiment. In this paper, we propose an end-to-end network that can learn to denoise and segment the time-series’ undersampled CTs, by training with the earlier highly sampled representative CTs. Our single network can offer two desired outputs while only training once, with the denoised output improving the accuracy of the final segmentation. Our method is able to outperform state-of-the-art methods in the task of semantic segmentation and offer comparable results in regard to denoising. Additionally, we propose a knowledge transfer scheme using synthetic tomograms. This not only allows accurate segmentation and denoising using less real-world data, but also increases segmentation accuracy. Finally, we make our datasets, as well as the code, publicly available.

There is an increase in consumption of agricultural produce as a result of the rapidly growing human population, particularly in developing nations. This has triggered high-quality plant phenotyping research to help with the breeding of high-yielding plants that can adapt to our continuously changing climate. Novel, low-cost, fully automated plant phenotyping systems, capable of infield deployment, are required to help identify quantitative plant phenotypes. The identification of quantitative plant phenotypes is a key challenge which relies heavily on the precise segmentation of plant images. Recently, the plant phenotyping community has started to use very deep convolutional neural networks (CNNs) to help tackle this fundamental problem. However, these very deep CNNs rely on some millions of model parameters and generate very large weight matrices, thus making them difficult to deploy infield on low-cost, resource-limited devices. We explore how to compress existing very deep CNNs for plant image segmentation, thus making them easily deployable infield and on mobile devices. In particular, we focus on applying these models to the pixel-wise segmentation of plants into multiple classes including background, a challenging problem in the plant phenotyping community. We combined two approaches (separable convolutions and SVD) to reduce model parameter numbers and weight matrices of these very deep CNN-based models. Using our combined method (separable convolution and SVD) reduced the weight matrix by up to 95% without affecting pixel-wise accuracy. These methods have been evaluated on two public plant datasets and one non-plant dataset to illustrate generality. We have successfully tested our models on a mobile device.

Plant phenotyping is the quantitative description of a plant's physiological, biochemical, and anatomical status which can be used in trait selection and helps to provide mechanisms to link underlying genetics with yield. Here, an active vision- based pipeline is presented which aims to contribute to reducing the bottleneck associated with phenotyping of architectural traits. The pipeline provides a fully automated response to photometric data acquisition and the recovery of three-dimensional (3D) models of plants without the dependency of botanical expertise, whilst ensuring a non-intrusive and non-destructive approach. Access to complete and accurate 3D models of plants supports computation of a wide variety of structural measurements. An Active Vision Cell (AVC) consisting of a camera-mounted robot arm plus combined software interface and a novel surface reconstruction algorithm is proposed. This pipeline provides a robust, flexible, and accurate method for automating the 3D reconstruction of plants. The reconstruction algorithm can reduce noise and provides a promising and extendable framework for high throughput phenotyping, improving current state-of-the-art methods. Furthermore, the pipeline can be applied to any plant species or form due to the application of an active vision framework combined with the automatic selection of key parameters for surface reconstruction.

We address the complex problem of reliably segmenting root structure from soil in X-ray Computed Tomography (CT) images. We utilise a deep learning approach, and propose a state-of-the-art multi-resolution architecture based on encoderdecoders. While previous work in encoder-decoders implies the use of multiple resolutions simply by downsampling and upsampling images, we make this process explicit, with branches of the network tasked separately with obtaining local high-resolution segmentation, and wider low-resolution contextual information. The complete network is a memory efficient implementation that is still able to resolve small root detail in large volumetric images. We compare against a number of different encoder-decoder based architectures from the literature, as well as a popular existing image analysis tool designed for root CT segmentation. We show qualitatively and quantitatively that a multi-resolution approach offers substantial accuracy improvements over a both a small receptive field size in a deep network, or a larger receptive field in a shallower network. We then further improve performance using an incremental learning approach, in which failures in the original network are used to generate harder negative training examples. Our proposed method requires no user interaction, is fully automatic, and identifies large and fine root material throughout the whole volume.

We propose a visual object tracking framework for the extraction of multiple interacting plant root systems from three-dimensional X-ray micro computed tomography images of plants grown in soil. Our method is based on a level set framework guided by a greyscale intensity distribution model to identify object boundaries in image cross-sections. Root objects are followed through the data volume, while updating the tracker’s appearance models to adapt to changing intensity values. In the presence of multiple root systems, multiple trackers can be used, but need to distinguish target objects from one another in order to correctly associate roots with their originating plants. Since root objects are expected to exhibit similar greyscale intensity distributions, shape information is used to constrain the evolving level set interfaces in order to lock trackers to their correct targets. The proposed method is tested on root systems of wheat plants grown in soil.

The emerging discipline of plant phenomics aims to measure key plant characteristics, or traits, though as yet the set of plant traits that should be measured by automated systems is not well defined. Methods capable of recovering generic representations of the 3D structure of plant shoots from images would provide a key technology underpinning quantification of a wide range of current and future physiological and morphological traits. We present a fully automatic approach to image-based 3D plant reconstruction which represents plants as series of small planar sections that together model the complex architecture of leaf surfaces. The initial boundary of each leaf patch is refined using a level set method, optimising the model based on image information, curvature constraints and the position of neighbouring surfaces. The reconstruction process makes few assumptions about the nature of the plant material being reconstructed. As such it is applicable to a wide variety of plant species and topologies, and can be extended to canopy-scale imaging. We demonstrate the effectiveness of our approach on real images of wheat and rice plants, an artificial plant with challenging architecture, as well as a novel virtual dataset that allows us to compute distance measures of reconstruction accuracy. We also illustrate the method’s potential to support the identification of individual leaves, and so the phenotyping of plant shoots, using a spectral clustering approach.

This paper proposes a novel approach to part-based tracking by replacing local matching of an appearance model by direct prediction of the displacement between local image patches and part locations. We propose to use cascaded regression with incremental learning to track generic objects without any prior knowledge of an object's structure or appearance. We exploit the spatial constraints between parts by implicitly learning the shape and deformation parameters of the object in an online fashion. We integrate a multiple temporal scale motion model to initialise our cascaded regression search close to the target and to allow it to cope with occlusions. Experimental results show that our tracker ranks first on the CVPR 2013 Benchmark.

Visual tracking frameworks have traditionally relied upon a single motion model such as Random Walk, and a fixed, embedded search method like Particle Filter. As a single motion model can't reliably handle various target motion types, the interest toward multiple motion models has grown over the years. The existence of multiple competing hypotheses or predictions by the multiple motion models opens up the possibility of a wider range of search methods. To search for the target in a fixed grid of equal sized cells, an integration of the Wang-Landau method and the Markov Chain Monte Carlo (MCMC) method has recently been introduced. In this paper, we generalize this search method to cells of variable size and location, where the cells are formed around the predictions generated by multiple motion models. The effectiveness of the proposed method is tested by adopting a multiple motion model tracker. Experiments show that the modified tracker has improved accuracy and better consistency over different runs compared to its original, and superior performance over state-of-the-art trackers in challenging video sequences.

Recent work in visual tracking has focussed on modelling target appearance, while using comparatively simple search methods to match those models to image data. Knowledge of the target's likely motion can both significantly reduce the search space and support more effective search strategies. We propose a new approach to target location which utilises sparse estimates of motion direction derived from local features to guide the generation of particles by a Markov Chain Monte Carlo (MCMC) based particle filter. The standard two-dimensional random walk is replaced by a series of one-dimensional searches in directions determined by the distribution of local feature motions. Two algorithms based on this approach are presented and evaluated. Experiments on both artificial and publically available, real image sequences show that the highest accuracy is obtained by sampling motion direction. The resulting algorithm successfully handles motion variations and reduces the likelihood that the tracker will be trapped in local extrem a when the target moves close to or is partially occluded by similar objects.

Optical Character Recognition (OCR) is an important technology. The Arabic language lacks both the variety of OCR systems and the depth of research relative to Roman scripts. A machine learning, Haar-Cascade classifier (HCC) approach was introduced by Viola and Jones (Viola and Jones 2001) to achieve rapid object detection based on a boosted cascade Haar-like features. Here, that approach is modified for the first time to suit Arabic glyph recognition. The HCC approach eliminates problematic steps in the pre-processing and recognition phases and, most importantly, the character segmentation stage. A recognizer was produced for each of the 61 Arabic glyphs that exist after the removal of diacritical marks. These recognizers were trained and tested on some 2,000 images each. The system was tested with real text images and produces a recognition rate for Arabic glyphs of 87%. The proposed method is fast, with an average document recognition time of 14.7 seconds compared with 15.8 seconds for commercial software.

We explore the idea of making aesthetic decorative patterns that contain multiple visual codes. We chart an iterative collaboration with ceramic designers and a restaurant to refine a recognition technology to work reliably on ceramics, produce a pattern book of designs, and prototype sets of tableware and a mobile app to enhance a dining experience. We document how the designers learned to work with and creatively exploit the technology, enriching their patterns with embellishments and backgrounds and developing strategies for embedding codes into complex designs. We discuss the potential and challenges of interacting with such patterns. We argue for a transition from designing 'codes to patterns' that reflects the skills of designers alongside the development of new technologies.

SUMMARY: The original RootTrace tool has proved successful in measuring primary root lengths across time series image data. Biologists have shown interest in using the tool to address further problems, namely counting lateral roots to use as parameters in screening studies, and measuring highly curved roots. To address this, the software has been extended to count emerged lateral roots, and the tracking model extended so that strongly curved and agravitropic roots can be now be recovered. Here, we describe the novel image analysis algorithms and user interface implemented within the RootTrace framework to handle such situations and evaluate the results. AVAILABILITY: The software is open source and available from http://sourceforge.net/projects/roottrace.

We explore the ways in which interfaces can be designed to deceive users so as to create the illusion of magic. We present a study of an experimental performance in which a magician used a computer vision system to conduct a series of illusions based on the well-known 'three cups' magic trick. We explain our findings in terms of the two broad strategies of misdirecting attention and setting false expectations, articulating specific tactics that were employed in each case. We draw on existing theories of collaborative and spectator interfaces, ambiguity and interpretation, and trajectories through experiences to explain our findings in broader HCI terms. We also extend and integrate current theory to provide refined sensitising concepts for analysing deceptive interactions.

Recent societal events and advances in computer vision technology have lead to the development of a variety of automatic surveillance systems. Despite their arguable success in the laboratory, fully automatic methods remain unsuitable for use in real life situations due to the complex nature of the context in which they must operate. Current techniques may, however, be immediately valuable if deployed as components of integrated human-automatic CCTV surveillance systems. It is therefore important to understand the potential of current automatic methods and provide design recommendations for semi-automatic systems, so that work to date can be exploited in full. An experiment was conducted to investigate the importance of the functionality and level of accuracy of, and feedback provided by, the automatic component of an integrated, semi-automatic CCTV surveillance system. The operatorspsila workload and spare attentional capacity was measured to investigate the effect of each of these factors. Results showed significant reduction in workload when reliable confidence information is fed back. Increases in accuracy and variation in functionality failed to produce evidence of change in workload.

This paper presents a new methodology for improving the tracking of multiple targets in complex scenes. The new method,Motion Parameter Sharing, incorporates social motion information into tracking predictions. This is achieved by allowing a tracker to share motion estimates within groups of targets which have previously been moving in a coordinated fashion. The method is intuitive and, as well as aiding the prediction estimates, allows the implicit formation of 'social groups' of targets as a side effect of the process. The underlying reasoning and method are presented, as well as a description of how the method fits into the framework of a typical Bayesian tracking system. This is followed by some preliminary results which suggest the method is more accurate and robust than algorithms which do not incorporate the social information available in multiple target scenarios.

Particle filtering provides a well-developed and widely adopted approach to visual tracking. For effective tracking in real-world environments the particle set must sample widely enough that it can represent alternative target states in areas of ambiguity. It must not, however, become diffuse, spreading across the image plane rather than clustering around the object(s) of interest. A key issue in the design of particle filter-based trackers is how to manage the spread of the particle set to balance these conflicting requirements. To be computationally efficient, balance must be achieved with as small a particle set as reasonably possible. A number of hybrid particle filter/mean-shift trackers have recently been proposed. We believe that their strength lies in their ability to alternately disperse and cluster particles together, providing both a degree of balance and a reduced particle set. We present a novel hybrid of the annealed particle filter and kernel mean-shift algorithms that emphasises this behaviour. The algorithm has been applied to a wide variety of artificial and real image sequences. The method has performance and efficiency advantages over both pure kernel mean-shift and particle filtering trackers and existing hybrid algorithms

The authors will introduce and demonstrate a novel computer vision based system for augmented performance. Unlike previous systems, which have primarily focused on 'high art' forms such as modern dance, this system is designed for use during a juggling performance. The system allows a juggler to interact with a computer through their movements, and the movements of the balls, to create audio and visual projections which respond to their performance. This system has been designed in an iterative process involving amateur and professional performers, in order to create a system which is truly accessible. In particular, this project takes inspiration from mass market interactive entertainment and has been developed to only use commodity hardware and to be easily distributable, in order to allow it to be used within the small, self funded groups common in the circus arts community.

By considering the spatial character of sensor-based interactive systems, this paper investigates how discussions of seams and seamlessness in ubiquitous computing neglect the complex spatial character that is constructed as a side-effect of deploying sensor technology within a space. Through a study of a torch (aka 'flashlight') based interface, we develop a framework for analysing this spatial character generated by sensor technology. This framework is then used to analyse and compare a range of other systems in which sensor technology is used, in order to develop a design spectrum that contrasts the revealing and hiding of a system's structure to users. Finally, we discuss the implications for interfaces situated in public spaces and consider the benefits of hiding structure from users.

It is widely accepted that automatic interpretation of drawings and documents are not achieved in the short to medium term and that manual digitisation is not an acceptable alternative. Despite this, there has been little systematic discussion of the role of the user in line drawing interpretation. We consider desirable properties of human interaction with line drawing interpretation systems, argue that descriptions are required of the inference structure of line drawing interpretation tasks and systems if appropriate and effective user involvement is to be achieved, and develop and compare the inference structures of two well-developed line drawing interpretation systems; one automatic, one interactive. Analysis of these representations highlights two areas in which interactivity could be improved. An interactive system addressing one of those areas is outlined and directions for future research are identified.

Movements of interfaces can be analyzed in terms of whether they are expected, sensed, and desired. Expected movements are those that users naturally perform; sensed are those that can be measured by a computer; and desired movements are those that are required by a given application. We show how a systematic comparison of expected, sensed, and desired movements, especially with regard to how they do not precisely overlap, can reveal potential problems with an interface and also inspire new features. We describe how this approach has been applied to the design of three interfaces: pointing flashlights at walls and posters in order to play sounds; the Augurscope II, a mobile augmented reality interface for outdoors; and the Drift Table, an item of furniture that uses load sensing to control the display of aerial photographs. We propose that this approach can help to build a bridge between the analytic and inspirational approaches to design and can help designers meet the challenges raised by a diversification of sensing technologies and interface forms, increased mobility, and an emerging focus on technologies for everyday life.

A combined tracking method using the Kalman filter and Hough transform is presented. An extended Kalman filter is used to model the parameters and motion of a set of lines detected in a Hough space The integration of these two techniques gives a number of advantages. The use of a Hough transform provides resilience to noise and partial occlusion, and the Kalman filter’s ability to predict future states is used to reduce the computational load of line detection. Analysis of the tracker from synthetic data shows that it is robust to noise, occlusion, and deviations from the constant motion model underlying the Kalman filter. Tracking results from video sequences illustrate its applicability to real-world domains. 1

We describe the design of tangible interfaces to the KidPad collaborative drawing tool. Our aims are to support the re-enactment of stories to audiences, and integration within real classroom environments. A six-month iterative design process, working with children and teachers in school, has produced the “magic carpet”, an interface that uses pressure mats and video-tracked and barcoded physical props to navigate a story in KidPad. Reflecting on this process, we propose four guidelines for the design of tangible interfaces for the classroom. (1) Use physical size and shysical props to encourage collaboration. (2) Be aware of how different interfaces emphasize different actions. (3) Be aware that superficial changes to the design can produce very different physical interactions. (4) Focus on open low-tech technologies rather than (over) polished products.

A methodology for the interpretation of images of engineering drawings is presented. The approach is based on the combination of schemata describing prototypical drawing constructs with a library of low-level image analysis routines and a set of explicit control rules applied by an LR(1) parser. The resulting system (Anon) integrates bottom-up and top-down processing strategies within a single, flexible framework modeled on the human perceptual cycle. Anon's structure and operation are described and discussed, and examples of its interpretation of real mechanical drawings are shown.>

schema-driven image analysis system capable of producing high level interpretations of greyscale images of mechanical engineering drawings. The system has been extended to incorporate a drawing memory into which completed schema instances are placed and which may be accessed as an integral part of ANON's knowledge directed visual search. This memory provides a basis for the integration into a coherent whole of the piecewise interpretations previously supplied by the system. It leads to a richer description of drawing content, improves efficiency and allows ANON to deal with partially complete interpretations. The new system's structure and operation are discussed and examples shown of it's interpretation of real mechanical drawings. 1 ANON [1,2] is a knowledge directed image analysis system