Nataliya Strokina

Joint Estimation of Bulk Flow Velocity and Angle Using a Lateral Line Probe

Measurement of complex natural flows, especially those occurring in rivers due to man-made structures, is often hampered by the limitations of existing flow measurement methods. Furthermore, there is a growing need for new measurement devices that are capable of measuring the hydrodynamic characteristics of complex natural flows required in environmental studies that often use fish as an indicator of ecological health. In this paper, we take the first step toward in situ natural flow measurements with a new biologically inspired probe design in conjunction with signal processing methods. The device presented in this paper is a dedicated hydrodynamically sensitive sensor array following the fish lateral line sensor modality. Low-level multidimensional sensor signals are transformed to the two key hydrodynamic primitives, bulk flow velocity and bulk flow angle. We show that this can be achieved via canonical signal transformation and kernel ridge regression, allowing velocity estimates with a less than 10 cm/s error. The approach provides robust velocity estimates not only when the sensor is ideally oriented parallel to the bulk flow, but also across the full range of angular deviations up to a completely orthogonal orientation by correcting the pressure field asymmetry for large angular deviations. Furthermore, we show that their joint estimation becomes feasible above a threshold current velocity of 0.45 m/s. The method demonstrated an error of 14 cm/s in velocity estimation in a river environment after training in laboratory conditions.

Flow feature extraction for underwater robot localization: Preliminary results

Underwater robots conventionally use vision and sonar sensors for perception purposes, but recently bio-inspired sensors that can sense flow have been developed. In literature, flow sensing has been shown to provide useful information about an underwater object and its surroundings. In the light of this, we develop an underwater landmark recognition technique which is based on the extraction and comparison of compact flow features. The proposed features are based on frequency spectrum of a pressure signal acquired by a piezo-resistive sensor. We report experiments in semi-natural (human-made flume with obstacles) and natural (river) underwater conditions where the proposed technique successfully recognizes previously visited locations.

Design and application of a fish-shaped lateral line probe for flow measurement

We introduce the lateral line probe (LLP) as a measurement device for natural flows. Hydraulic surveys in rivers and hydraulic structures are currently based on time-averaged velocity measurements using propellers or acoustic Doppler devices. The long-term goal is thus to develop a sensor system, which includes spatial gradients of the flow field along a fish-shaped sensor body. Interpreting the biological relevance of a collection of point velocity measurements is complicated by the fact that fish and other aquatic vertebrates experience the flow field through highly dynamic fluid-body interactions. To collect body-centric flow data, a bioinspired fish-shaped probe is equipped with a lateral line pressure sensing array, which can be applied both in the laboratory and in the field. Our objective is to introduce a new type of measurement device for body-centric data and compare its output to estimates of conventional point-based technologies. We first provide the calibration workflow for laboratory investigations. We then provide a review of two velocity estimation workflows, independent of calibration. Such workflows are required as existing field investigations consist of measurements in environments where calibration is not feasible. The mean difference for uncalibrated LLP velocity estimates from 0 to 50 cm/s under in a closed flow tunnel and open channel flume was within 4 cm/s when compared to conventional measurement techniques. Finally, spatial flow maps in a scale vertical slot fishway are compared for the LLP, direct measurements, and 3D numerical models where it was found that the LLP provided a slight overestimation of the current velocity in the jet and underestimated the velocity in the recirculation zone.

Framework for developing image-based dirt particle classifiers for dry pulp sheets

One important aspect of assessing the quality in pulp and papermaking is dirt particle counting and classification. Knowing the number and types of dirt particles present in pulp is useful for detecting problems in the production process as early as possible and for fixing them. Since manual quality control is a time-consuming and laborious task, the problem calls for an automated solution using machine vision techniques. However, the ground truth required to train an automated system is difficult to ascertain, since all of the dirt particles should be manually segmented and classified based on image information. This paper proposes a framework for developing and tuning dirt particle detection and classification systems. To avoid manual annotation, dry pulp sheets with a single dirt type in each were exploited to generate semisynthetic images with the ground truth information. To classify the dirt particles, a set of features were computed for each image segment. Sequential feature selection was employed to determine a close-to-optimal set of features to be used in classification. The framework was tested both with semisynthetically generated images based on real pulp sheets and with independent original real pulp sheets without any generation. The results of the experiments show that the semisynthetic procedure does not significantly change the properties of images and has little effect on the particle segmentation. The feature selection proved to be important when the number of dirt classes changes since it allows to improve the classification results. Using the standard classification methods, it is possible to obtain satisfactory results, although the methods modeling the data, such as the Bayesian classifier using the Gaussian Mixture Model, show better performance.

Detection of Curvilinear Structures by Tensor Voting Applied to Fiber Characterization

The paper presents a framework for the detection of curvilinear objects in images. Such objects are challenging to be described by a geometrical model, and although they appear in a number of applications, the problem of detecting curvilinear objects has drawn limited attention. The proposed approach starts with an edge detection algorithm after which the task of object detection becomes a problem of edge linking. A state-of-the-art local linking approach called tensor voting is used to estimate the edge point saliency describing the likelihood of a point belonging to a curve, and to extract the end points and junction points of these curves. After the tensor voting, the curves are grown from high-saliency seed points utilizing a linking method proposed in this paper. In the experimental part of the work, the method was systematically tested on pulp suspension images to characterize fibers based on their length and curl index. The fiber length was estimated with the accuracy of 71.5% and the fiber curvature with the accuracy of 70.7%.

Estimation of Flow Turbulence Metrics With a Lateral Line Probe and Regression

The time-averaged velocity of water flow is the most commonly measured metric for both laboratory and field applications. Its employment in scientific and engineering studies often leads to an oversimplification of the underlying flow physics. In reality, complex flows are ubiquitous, and commonly arise from fluid-body interactions with man-made structures, such as bridges as well as from natural flows along rocky river beds. Studying flows outside of laboratory conditions requires more detailed information in addition to time-averaged flow properties. The choice of in situ measuring device capable of delivering turbulence metrics is determined based on site accessibility, the required measuring period, and overall flow complexity. Current devices are suitable for measuring turbulence under controlled laboratory conditions, and thus there remains a technology gap for turbulence measurement in the field. In this paper, we show how a bioinspired fish-shaped probe outfitted with an artificial lateral line can be utilized to measure turbulence metrics under challenging conditions. The device and proposed signal processing methods are experimentally validated in a scale vertical slot fishway, which represents an extreme turbulent environment, such as those commonly encountered in the field. Optimal performance is achieved after 10 s of sampling using a standard deviation feature.

Comparison of Appearance-Based and Geometry-Based Bubble Detectors

Bubble detection is a complicated tasks since varying lighting conditions changes considerably the appearance of bubbles in liquid. The two common techniques to detect circular objects such as bubbles, the geometry-based and appearance-based approaches, have their advantages and weaknesses. The geometry-based methods often fail to detect small blob-like bubbles that do not match the used geometrical model, and appearance-based approaches are vulnerable to appearance changes caused by, e.g., illumination. In this paper, we compare a geometry-based concentric circular arrangements (CCA) and appearance-based sliding window methods as well as their combinations in terms of bubble detection, gas volume computation, and size distribution estimation. The best bubble detection performance was achieved with the sliding window method whereas the most precise volume estimate was produced by the CCA method. The combination of the two approaches gave only a minor advantage compared to the base methods.

Adaptive classification of dirt particles in papermaking process

In pulping and papermaking, dirt particles significantly affect the quality of paper. Knowledge of the dirt type helps to track the sources of the impurities which would considerably improve the paper making process. Dirt particle classification designed for this purpose should be adaptable because the dirt types are specific to the different processes of paper mills. This paper introduces a general approach for the adaptable classification system. The attention is paid to feature extraction and evaluation, in order to determine a suboptimal set of features for a certain data. The performance of standard classifiers on the provided data is presented, considering how the dirt particles or different types are classified. The effect of dirt particle grouping according to the particle size on the results of classification and feature evaluation is discussed. It is shown that the representative features of dirt particles from different size groups are different, which has an effect on the classification.

Image-based characterization of the pulp flows

Material flow characterization is important in the process industries and its further automation. In this study, close-to-laminar pulp suspension flows are analyzed based on double-exposure images captured in laboratory conditions. The correlation-based methods including autocorrelation and the particle image pattern technique were studied. During the experiments, synthetic and real test data with manual ground truth were used. The particle image pattern matching method showed better performance achieving the accuracy of 90.0% for the real data set with linear motion of the suspension and 79.2% for the data set with flow distortions.

Improving particle segmentation from process images with Wiener filtering

While there is growing interest in in-line measurements of paper making processes, the factory environment often restricts the acquisition of images. The in-line imaging of pulp suspension is often difficult due to constraints to camera and light positioning, resulting in images with uneven illumination and motion blur. This article presents an algorithm for segmenting fibers from suspension images and studies the performance of Wiener filtering in improving the sub-optimal images. Methods are presented for estimating the point spread function and noise-to-signal ratio for constructing the Wiener filter. It is shown that increasing the sharpness of the image improves the performance of the presented segmentation method.