Image classification of coral reef components from underwater color video

Soriano, Maricor; Marcos, Silvia; Saloma, Caesar; Quibilan, Miledel Christine C.; Aliño, P. M.

doi:10.1109/oceans.2001.968254

Cited by 29 publications

(16 citation statements)

References 3 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…A frame-by-frame comparison yields a similar result, showing that this automated set of algorithms can perform classification as well as or better than previous automated techniques using SVMs and neural networks that have been successful in classifying general groups of benthic reef organisms (i.e., Soriano et al 2001;Marcos et al 2005;Johnson-Roberson et al 2007;Mehta et al 2007). However, the automated technique here has the added benefit of providing a finer taxonomic classification capable of discriminating reef benthos, sometimes to the species level of the 18 organism and substrate types present in the library.…”

Section: Stokes and Deanementioning

confidence: 54%

“…Binary textural information and histogram comparisons have been used to attempt to identify crown of thorns starfish against a coral background (Clement et al 2005). Support vector machines (SVMs) as textural classifiers (Soriano et al 2001;Mehta et al 2007) or neural network classification schemes (Marcos et al 2005), in addition to color, have been used to identify corals, but their performance has been limited to discriminating general benthic groups (i.e., "algae," "dead coral," "branching corals," "live coral"). Johnson-Roberson et al (2007) used both visual and acoustic data to classify reef images collected from an autonomous underwater vehicle.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Automated processing of coral reef benthic images

Stokes

Deane

2009

Limnology & Ocean Methods

View full text Add to dashboard Cite

We describe an automated computer algorithm for the classification of coral reef benthic organisms and substrates sampled using a typical photographic quadrat survey. The technique compares subsections of a quadrat sample image (blocks) to a library of identified species blocks and computes a distance or probability of identification in a multidimensional hypervolume of discrimination metrics. The discrimination metrics include texture (calculated from a radial sampling of a two-dimensional discrete cosine transform) and three channels of a normalized color space. A standard multivariate classification technique based on the Mahalanobis distance was unsuccessful in discriminating substrata because of the large morphological variation inherent in reef organisms. An alternative classification scheme based on an exhaustive search through an organism reference library yielded classification maps comparable to those obtained by manual analysis. (Marcos et al. 2005), in addition to color, have been used to identify corals, but their performance has been limited to discriminating general benthic groups (i.e., "algae," "dead coral," "branching corals," "live coral"). Johnson-Roberson et al. (2007) used both visual and acoustic data to classify reef images collected from an autonomous underwater vehicle. Gabor wavelet transformations were used as a texture descriptor and a SVM for semiautomatic classification, but limitations in training data sets for the SVM classification restricted the discrimination to four general coral classes.The partial success of the previous research is due in part to the extremely difficult imaging environment of the coral reef, which has large variations in overall illumination, an oftenvarying level of light absorption between clear and turbid water, and a complicated three-dimensional topography. The species to be identified have complex, plastic morphologies and exhibit subtly varying pigmentation that can be common across many different species groups.In this study, we explore the use of normalized color space and discrete cosine transforms as a textural descriptor in a statistical distance-based classification scheme to identify reef benthos from images collected during a photo-quadrat sampling survey typical of ecological studies. By combining a standardized method using a camera framer and high-resolution images with the automated image processing described here, it is much easier to analyze the large number of image frames required to adequately sample the distribution of reef organisms on a large scale. ProcedureImage acquisition-All images were collected in Bonaire, Netherlands Antilles (12°12'25"N, 68°18'25"W) in January 2008, as part of the NOAA Ocean Explorations program, which included a photographic transect survey of the islands' reefs. Detailed descriptions of the leeward reefs of Bonaire can be found in Scatterday (1974) and the works of Kobluk andLysenko (1984) andvan Duyl (1985). To maximize the initial image quality, which would aid later manual or automatic proc...

show abstract

Section: Stokes and Deanementioning

confidence: 54%

Section: Introductionmentioning

confidence: 99%

Automated processing of coral reef benthic images

Stokes

Deane

2009

Limnology & Ocean Methods

View full text Add to dashboard Cite

show abstract

“…proposed a neural network-based coral image classifier using texture and color features. The authors classified corals into 5 benthic categories: dead coral, live coral, dead coral with algae, algae, and abiotics (sand/rock) [2]. For pre-processing, the coral RGB image is first transformed into normalized chromaticity coordinates.…”

Section: Previous Studies On Coral Classificationmentioning

confidence: 99%

A Coral Mapping and Health Assessment System Based on Texture Analysis

Naval

Soriano

Esmero

et al. 2014

Intelligent Information and Database Systems

View full text Add to dashboard Cite

Abstract. Corals have long played an important role in the environment with reefs hosting over four thousand species of marine animals all over the world. However, coral are very delicate in that slight unfavorable changes in environmental conditions cause them harm. With extreme changes in the environment happening ever more frequently, there is a need to efficiently monitor these corals for environmental efforts to keep pace. Manual monitoring of corals is expensive, tedious, and time-consuming. In this paper, an information system called the Coral Mapping and Health Assessment System (CMHAS) is proposed. The system aims to provide a digital repository of information that includes videos and images of corals, which are analyzed using image processing algorithms based on texture features to assess the health status of corals reefs. Evaluation on more than a hundred coral images show promising results with recognition rates as high as 82%.

show abstract

“…Scallops have been detected using a blob detector followed by template matching [29] and by segmentation followed by color-based region agglomeration [164]. Coral reefs have been segmented and classified using features ranging from filter banks [79,135], twodimensional discrete cosine transforms [160], morphological features [84], and local binary patterns [157].…”

Section: Object Detectionmentioning

confidence: 99%

Computational strategies for understanding underwater optical image datasets

Kaeli¹

2013

View full text Add to dashboard Cite

A fundamental problem in autonomous underwater robotics is the high latency between the capture of image data and the time at which operators are able to gain a visual understanding of the survey environment. Typical missions can generate imagery at rates hundreds of times greater than highly compressed images can be transmited acoustically, delaying that understanding until after the vehicle has been recovered and the data analyzed. While automated classification algorithms can lessen the burden on human annotators after a mission, most are too computationally expensive or lack the robustness to run in situ on a vehicle. Fast algorithms designed for mission-time performance could lessen the latency of understanding by producing low-bandwidth semantic maps of the survey area that can then be telemetered back to operators during a mission.This thesis presents a lightweight framework for processing imagery in real time aboard a robotic vehicle. We begin with a review of pre-processing techniques for correcting illumination and attenuation artifacts in underwater images, presenting our own approach based on multi-sensor fusion and a strong physical model. Next, we construct a novel image pyramid structure that can reduce the complexity necessary to compute features across multiple scales by an order of magnitude and recommend features which are fast to compute and invariant to underwater artifacts. Finally, we implement our framework on real underwater datasets and demonstrate how it can be used to select summary images for the purpose of creating low-bandwidth semantic maps capable of being transmitted acoustically.

show abstract

Image classification of coral reef components from underwater color video

Cited by 29 publications

References 3 publications

Automated processing of coral reef benthic images

Automated processing of coral reef benthic images

A Coral Mapping and Health Assessment System Based on Texture Analysis

Computational strategies for understanding underwater optical image datasets

Contact Info

Product

Resources

About