Kristian Meissner

Network of evolutionary binary classifiers for classification and retrieval in macroinvertebrate databases

In this paper, we focus on advanced classification and data retrieval schemes that are instrumental when processing large taxonomical image datasets. With large number of classes, classification and an efficient retrieval of a particular benthic macroinvertebrate image within a dataset will surely pose a severe problem. To address this, we propose a novel network of evolutionary binary classifiers, which is scalable, dynamically adaptable and highly accurate for the classification and retrieval of large biological species-image datasets. The classification and retrieval results for the macroinvertebrate test data attain taxonomic accuracy that equals and even surpasses that of an average expert. Our findings are encouraging for aquatic biomonitoring where cost intensity of sample analysis currently poses a bottleneck for routine biomonitoring.

Multiple order gradient feature for macro-invertebrate identification using support vector machines

This paper investigates the feasibility of automated benthic macro-invertebrate taxon identification based on support vector machines and a novel gradient based feature. Biomonitoring can efficiently pinpoint subtle environmental changes and is therefore globally widely used in ecological status assessment. However, all biomonitoring is costintensive due to the expert work needed to identify organisms. To relieve this problem an automated image recognition system for benthic macro-invertebrate taxonomical analysis is proposed in this work. Using a novel approach, we present high accuracy classification results, suggesting that automated taxa recognition for benthic macro-invertebrates is viable. Our study indicates that automated image recognition techniques can match human taxonomic identification accuracy and greatly reduce the costs of future taxonomic analysis.

Evaluating the performance of artificial neural networks for the classification of freshwater benthic macroinvertebrates

Abstract Macroinvertebrates form an important functional component of aquatic ecosystems. Their ability to indicate various types of anthropogenic stressors is widely recognized which has made them an integral component of freshwater biomonitoring. The use of macroinvertebrates in biomonitoring is dependent on manual taxa identification which is currently a time-consuming and cost-intensive process conducted by highly trained taxonomical experts. Automated taxa identification of macroinvertebrates is a relatively recent research development. Previous studies have displayed great potential for solutions to this demanding data mining application. In this research we have a collection of 1350 images from eight different macroinvertebrate taxa and the aim is to examine the suitability of artificial neural networks (ANNs) for automated taxa identification of macroinvertebrates. More specifically, the focus is drawn on different training algorithms of Multi-Layer Perceptron (MLP), probabilistic neural network (PNN) and Radial Basis Function network (RBFN). We performed thorough experimental tests and we tested altogether 13 training algorithms for MLPs. The best classification accuracy of MLPs, 95.3%, was obtained by two conjugate gradient backpropagation variations and scaled conjugate gradient backpropagation. For PNN 92.8% and for RBFN 95.7% accuracies were achieved. The results show how important a proper choice of ANN is in order to obtain high accuracy in the automated taxa identification of macroinvertebrates and the obtained model can outperform the level of identification which is made by a taxonomist.

Predator-prey interactions in a variable environment: responses of a caddis larva and its blackfly prey to variations in stream flow.

Predator-prey studies in streams have traditionally focused on mayfly-stonefly interactions in relatively constant flow conditions. In reality, however, lotic prey encounter multiple types of predators, most of which are restricted to low-velocity microhabitats. By contrast, some invertebrate prey may occur in very high current velocities. For example, many blackfly species are able to feed at velocities of 100 cm s-1, whereas even moderate currents reduce the hunting efficiency of their invertebrate predators. The caddisfly larvae of the genus Rhyacophila, however, may be an exception to the pattern of reducing predator efficiency with increasing velocity. Using a combination of laboratory and field experiments and behavioral field observations, we examined the interaction between predatory Rhyacophila caddis larvae and larval blackflies along a velocity gradient of 20–120 cm s-1. In laboratory experiments, Rhyacophila preferred currents slower than 50 cm s-1 while blackflies exhibited a wide tolerance of currents and frequently occurred in currents exceeding 100 cm s-1. In direct field observations, total activity and distance moved by Rhyacophila were similar at all current velocity regimes tested, but frequency of predation attempts on blackflies was lowest at the highest velocities (> 100 cm s-1). In a field colonization study, blackflies avoided substrates with the slowest velocities (< 40 cm s-1), as also did the caddis larvae. Only velocities approaching 100 cm s-1 provide blackflies with refuge from predation by Rhyacophila. Being able to maneuver across a wide range of velocities, Rhyacophila may have more pervasive effects on their prey than other lotic invertebrate predators.

Statistical classification and proportion estimation - an application to a macroinvertebrate image database

We apply and compare a random Bayes forest classifier and three traditional classification methods to a dataset of complex benthic macroinvertebrate images of known taxonomical identity. Since in biomonitoring changes in benthic macroinvertebrate taxa proportions correspond to changes in water quality, their correct estimation is pivotal. As classification errors are passed on to the allocated proportions, we explore a correction method known as a confusion matrix correction. Classification methods were compared using the misclassification error and the χ2 distance measures of the true proportions to the allocated and to the corrected proportions. Using low misclassification error and smallest χ2 distance measures as performance criteria the classical Bayes classifier performed best followed closely by the random Bayes forest.

Understanding the statistical properties of the percent model affinity index can improve biomonitoring related decision making

The percent model affinity (PMA) index is used to measure the similarity of two probability profiles representing, for example, an ideal profile (i.e. reference condition) and a monitored profile (i.e. possibly impacted condition). The goal of this work is to study the effects of sample size, evenness, true value of the index and number of classes on the statistical properties of the estimator of the PMA index. We derive and extend previous formulas of the expectation and variance of the estimator for estimated monitored profile and fixed reference profile. Using the obtained extension, we find that the estimator is asymptotically unbiased, converging faster when the profiles differ. When both profiles are estimated, we calculate the expectation using transformation rules for expectation and in addition derive the formula for the estimator’s variance. Since the computation of the probabilities in the variance formula is slow, we study the behavior of the variance with simulation experiments and assess whether it could be approximated with the variance for the fixed reference profile. Finally, we provide a set of recommendations for the users of the PMA index to avoid the most common caveats of the index.

Data Enrichment in Fine-Grained Classification of Aquatic Macroinvertebrates

The types and numbers of benthic macroinvertebrates found in a water body reflect water quality. Therefore, macroinvertebrates are routinely monitored as a part of freshwater ecological quality assessment. The collected macroinvertebrate samples are identified by human experts, which is costly and time-consuming. Thus, developing automated identification methods that could partially replace the human effort is important. In our group, we have been working toward this goal and, in this paper, we improve our earlier results on automated macroinvertebrate classification obtained using deep Convolutional Neural Networks (CNNs). We apply simple data enrichment prior to CNN training. By rotations and mirroring, we create new images so as to increase the total size of the image database sixfold. We evaluate the effect of data enrichment on Caffe and MatConvNet CNN implementations. The networks are trained either fully on the macroinvertebrate data or first pretrained using ImageNet pictures and then fine-tuned using the macroinvertebrate data. The results show 3-6% improvement, when the enriched data are used. This is an encouraging result, because it significantly narrows the gap between automated techniques and human experts, while it leaves room for future improvements as even the size of the enriched data, about 60000 images, is small compared to data sizes typically required for efficient training of deep CNNs.

Learned vs. engineered features for fine-grained classification of aquatic macroinvertebrates

Aquatic macroinvertebrate biomonitoring is an efficient way of assessment of slow and subtle anthropogenic changes and their effect on water quality. It is imperative to have reliable identification and counts of the various taxa occurring in samples as these form the basis for the quality indices used to infer the ecological status of the aquatic ecosystem. In this paper, we try to close the gap between human taxa identification accuracy (typically 90–95% on 30–40 classes of macroinvertebrates) and results of automatic fine-grained classification by introducing a novel technique based on Convolutional Neural Networks (CNN). CNN learns optimal features for macroinvertebrate classification and achieves near human accuracy when tested on 29 macroinvertebrate classes. Moreover, we perform comparative evaluation of the learned features against the hand-crafted features, which have been commonly used in classical approaches, and confirm superiority of the learned deep features over the engineered ones.

Testate amoebae community analysis as a tool to assess biological impacts of peatland use

As most ecosystems, peatlands have been heavily exploited for different human purposes. For example, in Finland the majority is under forestry, agriculture or peat mining use. Peatlands play an important role in carbon storage, water cycle, and are a unique habitat for rare organisms. Such properties highlight their environmental importance and the need for their restoration. To monitor the success of peatland restoration sensitive indicators are needed. Here we test whether testate amoebae can be used as a reliable bioindicator for assessing peatland condition. To qualify as reliable indicators, responses in testate amoebae community structure to ecological changes must be stronger than random spatial and temporal variation. In this study, we simultaneously assessed differences between the effects of seasonality, intermediate scale spatial variation and land uses on living testate amoebae assemblages in natural, forested and restored peatlands. We expected the effects of seasonality on testate amoebae communities to be less pronounced than those of land use and within site variation. On average, natural sites harboured the highest richness and density, while the lowest numbers were found at forestry sites. Despite small changes observed in taxa dominance and differences in TA community structure between seasons and years at some sites, spatial heterogeneity, temperature, pH, nor water table depth seemed to significantly affect testate amoebae communities. Instead, observed differences were related to type of land use, which explained 75% of the community variation. Our results showed that testate amoebae community monitoring is a useful tool to evaluate impacts of human land use on boreal peatlands.

Benchmark database for fine-grained image classification of benthic macroinvertebrates

Abstract Managing the water quality of freshwaters is a crucial task worldwide. One of the most used methods to biomonitor water quality is to sample benthic macroinvertebrate communities, in particular to examine the presence and proportion of certain species. This paper presents a benchmark database for automatic visual classification methods to evaluate their ability for distinguishing visually similar categories of aquatic macroinvertebrate taxa. We make publicly available a new database, containing 64 types of freshwater macroinvertebrates, ranging in number of images per category from 7 to 577. The database is divided into three datasets, varying in number of categories (64, 29, and 9 categories). Furthermore, in order to accomplish a baseline evaluation performance, we present the classification results of Convolutional Neural Networks (CNNs) that are widely used for deep learning tasks in large databases. Besides CNNs, we experimented with several other well-known classification methods using deep features extracted from the data.