Phil F. Culverhouse

Time to automate identification

Taxonomists should work with specialists in pattern recognition, machine learning and artificial intelligence, say Norman MacLeod, Mark Benfield and Phil Culverhouse — more accuracy and less drudgery will result.

Recommendations on methods for the detection and control of biological pollution in marine coastal waters

Adverse effects of invasive alien species (IAS), or biological pollution, is an increasing problem in marine coastal waters, which remains high on the environmental management agenda. All maritime countries need to assess the size of this problem and consider effective mechanisms to prevent introductions, and if necessary and where possible to monitor, contain, control or eradicate the introduced impacting organisms. Despite this, and in contrast to more enclosed water bodies, the openness of marine systems indicates that once species are in an area then eradication is usually impossible. Most institutions in countries are aware of the problem and have sufficient governance in place for management. However, there is still a general lack of commitment and concerted action plans are needed to address this problem. This paper provides recommendations resulting from an international workshop based upon a large amount of experience relating to the assessment and control of biopollution.

Robust Adaptive Control of an Uninhabited Surface Vehicle

A robust adaptive autopilot for uninhabited surface vehicles (USV) based on a model predictive controller (MPC) is presented in this paper. The novel autopilot is capable of handling sudden changes in system dynamics. In real life situations, very often a sudden change in dynamics results in missions being aborted and the uninhabited vehicles have to be rescued before they cause damage to other marine craft in the vicinity. This problem has been suitably dealt with by this innovative design. The MPC adopts an online adaptive nature by utilising three algorithms, individually: gradient descent, least squares and weighted least squares (WLS). Even with random initialisation, significant improvements over the other algorithmic approach were achieved by WLS by maintaining the intermittent continuous values of system parameters and periodically reinitialising the covariance matrix. Also, a time frame of 25 seconds appears to be the optimum to reinitialise the parameters in simulation studies. This novel approach enables the autopilot to cope well with significant changes in the system dynamics and empowers USVs to accomplish their desired missions.

Novel Hybrid Adaptive Controller for Manipulation in Complex Perturbation Environments

In this paper we present a hybrid control scheme, combining the advantages of task-space and joint-space control. The controller is based on a human-like adaptive design, which minimises both control effort and tracking error. Our novel hybrid adaptive controller has been tested in extensive simulations, in a scenario where a Baxter robot manipulator is affected by external disturbances in the form of interaction with the environment and tool-like end-effector perturbations. The results demonstrated improved performance in the hybrid controller over both of its component parts. In addition, we introduce a novel method for online adaptation of learning parameters, using the fuzzy control formalism to utilise expert knowledge from the experimenter. This mechanism of meta-learning induces further improvement in performance and avoids the need for tuning through trial testing.

Human and machine factors in algae monitoring performance

Abstract We all take our visual systems for granted, and often assume we are always ‘near perfect’ observers. This is not the case; expert visual recognition is complex and can be error prone. Starting with examples that define the problem I will explore some of the issues of recognition where expert judgements are required. In addition to ‘expert’ effects, there are a number of cognitive factors that can severely affect performance, including fatigue, boredom, recency effects, positivity bias and short-term memory effects. Experimental evidence of the impact of these on performance are presented and discussed. The specimen identifications generated by experts are useful not only to ecology, but to researchers developing systems for automatic labelling of marine plankton. Comparisons of performance are presented, where human experts have been pitted against machines to label plankton. Consensus of opinion is important in reducing errors, yet it is the norm for experts to operate alone. The shortcomings of man and machines engaged in plankton recognition are reviewed and the future of automation is assessed.

Photonic gas sensors exploiting directly the optical properties of hybrid carbon nanotube localized surface plasmon structures

We investigate the modification of the optical properties of carbon nanotubes (CNTs) resulting from a chemical reaction triggered by the presence of a specific compound (gaseous carbon dioxide (CO2)) and show this mechanism has important consequences for chemical sensing. CNTs have attracted significant research interest because they can be functionalized for a particular chemical, yielding a specific physical response which suggests many potential applications in the fields of nanotechnology and sensing. So far, however, utilizing their optical properties for this purpose has proven to be challenging. We demonstrate the use of localized surface plasmons generated on a nanostructured thin film, resembling a large array of nano-wires, to detect changes in the optical properties of the CNTs. Chemical selectivity is demonstrated using CO2 in gaseous form at room temperature. The demonstrated methodology results additionally in a new, electrically passive, optical sensing configuration that opens up the possibilities of using CNTs as sensors in hazardous/explosive environments.

A novel application of motion analysis for detecting stress responses in embryos at different stages of development

BackgroundMotion analysis is one of the tools available to biologists to extract biologically relevant information from image datasets and has been applied to a diverse range of organisms. The application of motion analysis during early development presents a challenge, as embryos often exhibit complex, subtle and diverse movement patterns. A method of motion analysis able to holistically quantify complex embryonic movements could be a powerful tool for fields such as toxicology and developmental biology to investigate whole organism stress responses. Here we assessed whether motion analysis could be used to distinguish the effects of stressors on three early developmental stages of each of three species: (i) the zebrafish Danio rerio (stages 19 h, 21.5 h and 33 h exposed to 1.5% ethanol and a salinity of 5); (ii) the African clawed toad Xenopus laevis (stages 24, 32 and 34 exposed to a salinity of 20); and iii) the pond snail Radix balthica (stages E3, E4, E6, E9 and E11 exposed to salinities of 5, 10 and 15). Image sequences were analysed using Sparse Optic Flow and the resultant frame-to-frame motion parameters were analysed using Discrete Fourier Transform to quantify the distribution of energy at different frequencies. This spectral frequency dataset was then used to construct a Bray-Curtis similarity matrix and differences in movement patterns between embryos in this matrix were tested for using ANOSIM.ResultsSpectral frequency analysis of these motion parameters was able to distinguish stage-specific effects of environmental stressors in most cases, including Xenopus laevis at stages 24, 32 and 34 exposed to a salinity of 20, Danio rerio at 33 hpf exposed to 1.5% ethanol, and Radix balthica at stages E4, E9 and E11 exposed to salinities of 5, 10 and 15. This technique was better able to distinguish embryos exposed to stressors than analysis of manual quantification of movement and within species distinguished most of the developmental stages studied in the control treatments.ConclusionThis innovative use of motion analysis incorporates data quantifying embryonic movements at a range of frequencies and so provides an holistic analysis of an embryo’s movement patterns. This technique has potential applications for quantifying embryonic responses to environmental stressors such as exposure to pharmaceuticals or pollutants, and also as an automated tool for developmental staging of embryos.

Comparison of Human and Automatic Facial Emotions and Emotion Intensity Levels Recognition

In this paper we discuss the problem of human facial emotions and emotion intensity levels recognition using active appearance models (AAM) and support vector machines (SVM). AAM are used for appropriate feature extraction and SVM for convenient facial emotion and emotion level classification. Problems related to proper selection of data retrieved from AAM and SVM learning parameters settings are discussed too. Furthermore, we propose analysis of specially designed psychological experiment which led to alternative classifier evaluation methodology that uses the human visual system as a reference point. Finally, we analyze classification characteristics of proposed AAM-SVM classifier comparing to humans and show that our classifier give slightly more consistent labels to emotion categories than human subjects, while humans were more consistent at identifying emotion intensity level than SVM.

Committees, collectives and individuals: Expert visual classification by neural network

It has been established that committee classifiers, in which the outputs of different, individual network classifiers are combined in various ways, can produce better accuracy than the best individual in the committee. We describe results showing that these advantages are obtained when neural networks are applied to a taxonomic problem in marine science: the classification of images of marine phytoplankton. Significant benefits were found when individual networks, trained on different classes of input, having comparable individual performances, were combined. Combining networks of very different accuracy did not improve performance when measured against the best single network, but nor was it reduced. An alternative architecture, which we term a collective machine, in which the different data types are combined in a single network, was found to have significantly better accuracy than the committee machine architectures. The performance gains and resilience to non-discriminatory types of data suggest the techniques have great utility in the development of general purpose, network classifiers.

3D object recognition from static 2D views using multiple coarse data channels

Abstract A 3D object recognition system is described that employs novel multiresolution representation and coarse encoding of feature information. Modifications are bought to classic feature extraction methods by proposing the use of wavelet transform maxima for directing the actions of feature extraction modules. The reasons behind the use of a multi-channel architecture are described, together with the description of the feature extraction and coarse modules. The targeted field of application being automatic categorisation of natural objects, the proposed system is designed to run on ordinary hardware platforms and to process an input in a short timeframe. The system has been evaluated on a variety of 2D views of a set of 5 synthetic objects designed to present various degrees of similarity, as being rated by a panel of human subjects. Parallels between these ratings and the system’s behaviour are drawn. Additionally a small set of photomicrographs of fish larvae has been used to assess the system’s performance when presented with very similar, non-rigid shapes. For comparison, the parameters extracted from each image were fed into two categorisers, discriminant analysis and multilayer feedforward neural network with backpropagation of error. Experimental evidence is presented which demonstrates the efficacy of the methods. The satisfactory categorisation performances of the system are reported, and conclusions are drawn about the system’s behaviour.