Plankton Images Research Articles

Accurate automatic quantification of taxa-specific plankton abundance using dual classification with correction

MEPS Marine Ecology Progress Series Contact the journal Facebook Twitter RSS Mailing List Subscribe to our mailing list via Mailchimp HomeLatest VolumeAbout the JournalEditorsTheme Sections MEPS 306:51-61 (2006) - doi:10.3354/meps306051 Accurate automatic quantification of taxa-specific plankton abundance using dual classification with correction Qiao Hu1,*, Cabell Davis2 1Department of Applied Ocean Physics and Engineering, and 2Department of Biology, Woods Hole Oceanographic Institution, Woods Hole, Massachusetts 02543-1541, USA *Email: qhu@whoi.edu ABSTRACT: Optical imaging samplers are becoming widely used in plankton ecology, but image analysis methods have lagged behind image acquisition rates. Automated methods for analysis and recognition of plankton images have been developed, which are capable of real-time processing of incoming image data into major taxonomic groups. The limited accuracy of these methods can require significant manual post-processing to correct the automatically generated results, in order to obtain accurate estimates of plankton abundance patterns. We present here a dual-classification method in which each plankton image is first identified using a shaped-based feature set and a neural network classifier, and then a second time using a texture-based feature set and a support vector machine classifier. The plankton image is considered to belong to a given taxon only if the 2 identifications agree; otherwise it is labeled as unknown. This dual-classification method greatly reduces the false positive rate, and thus gives better abundance estimation in regions of low relative abundance. A confusion matrix is computed from a set of training images in order to determine the detection and false positives rates. These rates are used to correct abundances estimated from the automatic classification results. Aside from the manual sorting required to generate the initial training set of images, this dual-classification method is fully automatic and does not require subsequent manual correction of automatically sorted images. The resulting abundances agree closely with those obtained using manually sorted results. A set of images from a Video Plankton Recorder was used to evaluate this method and compare it with previously reported single-classifier results for major taxa. KEY WORDS: Plankton · Video · Sampling · Pattern recognition · Real-time · Rejection Full text in pdf format PreviousNextExport citation RSS - Facebook - Tweet - linkedIn Cited by Published in MEPS Vol. 306. Online publication date: January 11, 2006 Print ISSN: 0171-8630; Online ISSN: 1616-1599 Copyright © 2006 Inter-Research.

Transport of plankton and particles between the Chukchi and Beaufort Seas during summer 2002, described using a Video Plankton Recorder

A key goal of the Western Arctic Shelf Basin Interactions program is to understand how physical and biological processes together impact shelf–basin exchange of biological, chemical, and physical properties. High-resolution vertical distributions of plankton and particles were obtained using an Auto Video Plankton Recorder from 29 locations on the Chukchi Shelf, in the deep Beaufort Sea, and across the Beaufort–Chukchi Shelf-break during a cruise on the USCGC Healy in July–August, 2002. Coincident velocity estimates were collected using hull-mounted acoustic Doppler current profilers. Images of plankton and particles were extracted automatically and identified manually to taxa and type. Copepods, diatom chains, decaying diatoms, marine snow, and radiolarians were the most abundant categories observed. Distinct regional differences in abundance were observed that were associated with different oceanographic regimes and with the prevailing circulation in the region. Vertical distributions were closely associated with the physical structure of the water column. A sharp horizontal discontinuity in abundance of all categories between shelf and basin was observed, located over the shelf break and potentially established and maintained by transport of plankton and particles along-shelf to the east rather than northwards towards the basin. Barrow Canyon and the shelf and shelf-break east of Barrow Canyon had very high concentrations of plankton and particles, especially marine snow, that may have resulted from elevated production on the eastern Chukchi Shelf that subsequently was advected out of Barrow Canyon and to the east. Comparisons of downward flux, estimated from particle sinking rates based on individual marine snow particle size, and horizontal velocities suggested that much of the marine snow carbon was sinking to the benthos of the Chukchi Sea prior to being advected off-shelf. Velocities and plankton concentrations together indicated that little off-shelf flux of plankton or particles to the basin was occurring except in an eddy located off of the Beaufort Shelf.

Automatic plankton image recognition with co-occurrence matrices and Support Vector Machine

A long-standing problem in plankton ecology is sparseness of taxa-specific data. New optical imaging systems are becoming available which can acquire high-resolution data on the abun- dance and biomass of plankton taxa. The Video Plankton Recorder (VPR) has been designed and used for automatic sampling and visualization of major planktonic taxa at sea in real time, providing high-resolution data over a broad range of scales. Although these optical systems produce digital images of plankton that can be automatically identified by computer, the limited accuracy of auto- matic classification methods can reduce confidence in subsequent abundance estimates, especially in areas where a taxon is in low relative abundance. This paper describes an improved classification system for automatic identification of plankton taxa from digital images. Classifiers are trained from a set of images that were classified by human experts. The data set used to verify the classification system contains over 20 000 planktonic images manually sorted into 7 different categories. The new method uses co-occurrence matrices (COM) as the feature, and a Support Vector Machine (SVM) as the classifier. This new method is compared against a previous plankton recognition system, which used moment invariants, Fourier descriptors and granulometry as features and a learning vector quantization neural network as a classifier. The new method reduced the classification error rate from 39 to 28%. Subsequent plankton abundance estimates are improved by more than 50% in regions of low relative abundance. In general, the reduction in classification error was due to a combination of the use of COM and SVM.

Multiple competitive learning network fusion for object classification

This paper introduces a multiple competitive learning neural network fusion method for pattern recognition. By defining a confidence level measure for the learning vector quantization network classifier, we develop both a serial and a parallel network fusion algorithm to combine the discriminatory ability of different individually trained networks. We use two distinct feature vectors, gray-scale morphological granulometry and Fourier boundary descriptor, to demonstrate the efficacy of the classifier. The algorithms are applied on the classification of more than 8000 underwater plankton images. The classification accuracy for training data and for testing data are over 92% and 85%, respectively.

An in situ video camera for plankton studies:design and preliminary observations

A design for an in situ video camera for plankton observations is presented. A light weight aluminium rig supports a video camera with an opposing stroboscope which produces dark field images of plankton >0.3 mm. In the present configuration the camera can be used down to 100 m and provides video images of 10 to 40x magnification when viewed on a 14 (ca 35 cm) monitor. Video recordings from the field show moderate aggregations of small copepods to the subsurface fluorescence maximum and at times indications of horizontal patchiness. The large copepod Calanus finmarchicus was found to aggregate 5 m below the fluorescence maximum at night. In comparisons with a traditional WP-2 plankton net, the video camera provided similar length frequency distributions while abundance estimates from video recordings were 10 to 20% higher than net estimates. II is concluded that the presented video camera offers excellent observations of the pelagic environment at an affordable cost.

Automatic Plankton Image Recognition

Plankton form the base of the food chain in the ocean and are fundamental to marine ecosystem dynamics. The rapid mapping of plankton abundance together with taxonomic and size composition is very important for ocean environmental research, but difficult or impossible to accomplish using traditional techniques. In this paper, we present a new pattern recognition system to classify large numbers of plankton images detected in real time by the Video Plankton Recorder (VPR), a towed underwater video microscope system. The difficulty of such classification is compounded because: 1) underwater images are typically very noisy, 2) many plankton objects are in partial occlusion, 3) the objects are deformable and 4) images are projection variant, i.e., the images are video records of three-dimensional objects in arbitrary positions and orientations. Our approach combines traditional invariant moment features and Fourier boundary descriptors with gray-scale morphological granulometries to form a feature vector capturing both shape and texture information of plankton images. With an improved learning vector quantization network classifier, we achieve 95% classification accuracy on six plankton taxa taken from nearly 2,000 images. This result is comparable with what a trained biologist can achieve by using conventional manual techniques, making possible for the first time a fully automated, at sea-approach to real-time mapping of plankton populations.

A preliminary study of the distribution of plankton using hologrammetry

The distribution and dynamics of particles in the aquatic environment play an important role in the modelling of bio-geochemical processes. Previous work on the measurement of such particles, which vary in size from tens of micrometres (individual cells) to several centimetres (aggregates such as ‘marine snow’), has mainly used electronic counting or conventional photography coupled with image analysis. Here we report on an initial study of the use of holographic mensuration, otherwise known as hologrammetry, for the enumeration, sizing and spatial distribution determination of plankton. We present results on imaging plankton in water tanks using both in-line and off-axis pulsed-laser holography. In this work, we have recorded in-line holograms in a volume of 2400 cm 3 of water with a resolution of better than 20 μm and off-axis holograms in a volume of 36000 cm 3 with a resolution of 140 μm. In both cases, identifiable images of plankton were obtained and precise spatial coordinates determined from the in-line holograms.

Visualization and quantification of plankton and detritus using digital confocal microscopy

The application of digital confocal microscopy to studies of plankton and detritus is described. Plankton cells are fluorescently stained using DAPl and proflavin according to established procedures, while detritus is uniquely stained with DAPI and propidium iodide using a recently published companion protocol. Stacks of digitized images of plankton and detritus are acquired in 3 dimens~ons (3D) using an integrating color charge-coupled device (CCD) mounted atop a fluorescence microscope. A desktop computer drives a z-axis motonzed controller to optically section plankton cells and detritus at very fine intervals (as small as 1 pm or less) Out-of-focus haze associated with each optical slice is removed via a nearest-neighbor algorithm in a process termed deconvolution. lmage sets containing these stacks of deconvolved optical slices are subsequently displayed in 3D by volume-rendering software operating aboard a graphics computer. Since the exact x /y / z dimensions of each 3D picture element, or voxel, are known, the volume of plankton or detritus can be calculated. A procedure is described whereby the volume of detritus can be converted to units of carbon and nitrogen. This approach, combined with more traditional 2D image analysis of plankton communities, offers the first opportunity to separately quantify the pool sizes of plankton and detritus in aquatic ecosystems.

The identification, counting, and measurement of phytoplankton by an image-processing system

A new image-processing system has been developed, to identify and make direct counts of microscope images of plankton and micro-organisms, using pattern analysis. Principal-component analysis has been applied in a novel way to process each of the three primary colour components of the biological pattern recorded by a colour television camera. The added dimension of colour is unique, we believe, in this context. Attempts were made to measure the three horns (apical, right, and left horn length) and width of Ceratium furca , using three techniques. From the comparison of the values measured by the electronic system with those measured by two traditional techniques for 17 samples, we conclude that the former system can identify Ceratium furca , based upon measurement ofthree horn lengths, with ±10% accuracy. Thus it appears that this computerized technique has practical application to the identification of certain species of phytoplankton.