Automatic recognition of biological particles in microscopic images

Ranzato, Marc ' Aurelio; Taylor, Portia E.; House, James; Flagan, Richard C.; LeCun, Yann; Perona, Pietro

doi:10.1016/j.patrec.2006.06.010

Cited by 102 publications

(65 citation statements)

References 11 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…At present, there have been too few studies to make any definitive statement about which approach is best, or most appropriate for certain situations. Nevertheless, there are clear signs that an algorithmic approach to classification can increase the taxonomic resolution of the sporomorph record Mander et al 2013), and we believe that the incorporation of quantitative computational morphometric methods (e.g., Mander et al 2013) within supervised and unsupervised machine-learning frameworks (e.g., De Sá -Otero et al 2004;Zhang et al 2004;Chun et al 2006;Ranzato et al 2007;Costa and Yang 2009;Holt et al 2011;Punyasena et al 2012) is an area that requires further investigation. Certain approaches that have been designed to fit the requirements of a specific clade (e.g., Mander et al 2013 for grass pollen) are unlikely to be suitable for the generation of standard paleoecological counts of thousands of sporomorphs from many different higher taxa through time.…”

Section: Computational Image Analyses and Machine Learningmentioning

confidence: 99%

“…This can allow a wider range of features to be incorporated into the process of classification, including both abstract ) and morphometric (Mander et al 2013) features, and it can prevent researchers from defaulting to conservative classifications in an effort ensure repeatability . The diversity of approaches that have been explored to date (e.g., De Sá -Otero et al 2004;Zhang et al 2004;Chun et al 2006;Ranzato et al 2007;Costa and Yang 2009;Holt et al 2011;Punyasena et al 2012;Mander et al 2013) suggests an emerging distinction between researchers who employ computational image analyses, often with machine learning, in order to reduce the amount of time expert palynologists spend undertaking routine work (e.g., Holt et al 2011), and those whose primary aim is to increase taxonomic resolution with less regard for the amount of time spent doing so (e.g., Mander et al 2013). Exploring this distinction promises exciting avenues of future research and may lead to the development of clade-specific palynology, in which certain combinations of imaging techniques and analytical approaches are tailored to suit different sporomorph groups.…”

Section: Synthesis and Outlookmentioning

confidence: 99%

See 1 more Smart Citation

On the Taxonomic Resolution of Pollen and Spore Records of Earth’s Vegetation

Mander

Punyasena

2014

International Journal of Plant Sciences

View full text Add to dashboard Cite

Premise of research. Pollen and spores (sporomorphs) are a valuable record of plant life and have provided information on subjects ranging from the nature and timing of evolutionary events to the relationship between vegetation and climate. However, sporomorphs can be morphologically similar at the species, genus, or family level. Studies of extinct plant groups in pre-Quaternary time often include dispersed sporomorph taxa whose parent plant is known only to the class level. Consequently, sporomorph records of vegetation suffer from limited taxonomic resolution and typically record information about plant life at a taxonomic rank above species.Methodology. In this article, we review the causes of low taxonomic resolution, highlight examples where this has hampered the study of vegetation, and discuss the strategies researchers have developed to overcome the low taxonomic resolution of the sporomorph record. Based on this review, we offer our views on how greater taxonomic precision might be attained in future work.Pivotal results. Low taxonomic resolution results from a combination of several factors, including inadequate reference collections, the absence of sporomorphs in situ in fossilized reproductive structures, and damage following fossilization. A primary cause is the difficulty of accurately describing the very small morphological differences between species using descriptive terminology, which results in palynologists classifying sporomorphs conservatively at the genus or family level to ensure that classifications are reproducible between samples and between researchers.Conclusions. In our view, the most promising approach to the problem of low taxonomic resolution is a combination of high-resolution imaging and computational image analysis. In particular, we encourage palynologists to explore the utility of microscopy techniques that aim to recover morphological information from below the diffraction limit of light and to employ computational image analyses to consistently quantify small morphological differences between species.

show abstract

Section: Computational Image Analyses and Machine Learningmentioning

confidence: 99%

Section: Synthesis and Outlookmentioning

confidence: 99%

On the Taxonomic Resolution of Pollen and Spore Records of Earth’s Vegetation

Mander

Punyasena

2014

International Journal of Plant Sciences

View full text Add to dashboard Cite

show abstract

“…For each tracked object, a binary mask is obtained with segmentation algorithms described in II.b, allowing features to be extracted. The detected object is then classified using a Gaussian mixture model [13] of feature vectors based on Schmid invariants using a training set obtained with the help of professional annotators at MBARI.…”

Section: E Towards Classificationmentioning

confidence: 99%

“…As a proof of concept test, a benthic image training data base was created with a total of about 6000 frames and 200 events. Grayscale square subimages containing an example of object classes were processed, local jets were used to extract features [13] and the training data was modeled with a mixture of Gaussians [13]. Classification tests were carried out using Matlab using three training classes: Rathbunaster californicus, Parasticopus leukothete, and "other" (containing sub-images representing others events e.g.…”

Section: E Towards Classificationmentioning

confidence: 99%

Detecting, Tracking and Classifying Animals in Underwater Video

et al. 2006

View full text Add to dashboard Cite

-For oceanographic research, remotely operated underwater vehicles (ROVs) and underwater observatories routinely record several hours of video material every day. Manual processing of such large amounts of video has become a major bottleneck for scientific research based on this data. We have developed an automated system that detects, tracks, and classifies objects that are of potential interest for human video annotators. By pre-selecting salient targets for track initiation using a selective attention algorithm, we reduce the complexity of multi-target tracking. Then, if an object is tracked for several frames, a visual event is created and passed to a Bayesian classifier utilizing a Gaussian mixture model to determine the object class of the detected event.

show abstract

“…[1]. Only [4] used a data set from real air samples containing a reasonable number of pollen grains (3686) from 27 species. But even on a reduced data set containing only 8 species and dust particles, the recall was only 64,9% with a precision of 30%.…”

Section: Introductionmentioning

confidence: 99%

3D Invariants with High Robustness to Local Deformations for Automated Pollen Recognition

Ronneberger

Wang

Burkhardt

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Abstract. We present a new technique for the extraction of features from 3D volumetric data sets based on group integration. The features are invariant to translation, rotation and global radial deformations. They are robust to local arbitrary deformations and nonlinear gray value changes, but are still sensitive to fine structures. On a data set of 389 confocally scanned pollen from 26 species we get a precision/recall of 99.2% with a simple 1NN classifier. On volumetric transmitted light data sets of about 180,000 airborne particles, containing about 22,700 pollen grains from 33 species, recorded with a low-cost optic in a fully automated online pollen monitor the mean precision for allergenic pollen is 98.5% (recall: 86.5%) and for the other pollen 97.5% (recall: 83.4%).

show abstract

Automatic recognition of biological particles in microscopic images

Cited by 102 publications

References 11 publications

On the Taxonomic Resolution of Pollen and Spore Records of Earth’s Vegetation

On the Taxonomic Resolution of Pollen and Spore Records of Earth’s Vegetation

Detecting, Tracking and Classifying Animals in Underwater Video

3D Invariants with High Robustness to Local Deformations for Automated Pollen Recognition

Contact Info

Product

Resources

About