Limits...
A tool for developing an automatic insect identification system based on wing outlines.

Yang HP, Ma CS, Wen H, Zhan QB, Wang XL - Sci Rep (2015)

Bottom Line: In five repeated experiments, the mean accuracy for identification of each species ranged from 90% to 98%.The accuracy increased to 99% when the samples were first divided into two groups based on features of their compound eyes.DAIIS can therefore be a useful tool for developing a system of automated insect identification.

View Article: PubMed Central - PubMed

Affiliation: Department of Entomology, China Agricultural University, Beijing, China.

ABSTRACT
For some insect groups, wing outline is an important character for species identification. We have constructed a program as the integral part of an automated system to identify insects based on wing outlines (DAIIS). This program includes two main functions: (1) outline digitization and Elliptic Fourier transformation and (2) classifier model training by pattern recognition of support vector machines and model validation. To demonstrate the utility of this program, a sample of 120 owlflies (Neuroptera: Ascalaphidae) was split into training and validation sets. After training, the sample was sorted into seven species using this tool. In five repeated experiments, the mean accuracy for identification of each species ranged from 90% to 98%. The accuracy increased to 99% when the samples were first divided into two groups based on features of their compound eyes. DAIIS can therefore be a useful tool for developing a system of automated insect identification.

No MeSH data available.


Related in: MedlinePlus

Flowchart of the automated identification system.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4528224&req=5

f5: Flowchart of the automated identification system.

Mentions: The system primarily includes two subsystems (Fig. 5). The first subsystem is the identification model generator, which builds the identification model with SVMs and a training dataset. The second subsystem is the model validator, which calculates the identification accuracy by comparing the model output with test data. The training and test samples were identified by a taxonomist. All data were randomly divided into the training and test datasets at a 1:1 ratio. After a valid identification model is established, users can apply our system to identify an unknown specimen. In this paper, we describe the procedures to implement those functions step by step.


A tool for developing an automatic insect identification system based on wing outlines.

Yang HP, Ma CS, Wen H, Zhan QB, Wang XL - Sci Rep (2015)

Flowchart of the automated identification system.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4528224&req=5

f5: Flowchart of the automated identification system.
Mentions: The system primarily includes two subsystems (Fig. 5). The first subsystem is the identification model generator, which builds the identification model with SVMs and a training dataset. The second subsystem is the model validator, which calculates the identification accuracy by comparing the model output with test data. The training and test samples were identified by a taxonomist. All data were randomly divided into the training and test datasets at a 1:1 ratio. After a valid identification model is established, users can apply our system to identify an unknown specimen. In this paper, we describe the procedures to implement those functions step by step.

Bottom Line: In five repeated experiments, the mean accuracy for identification of each species ranged from 90% to 98%.The accuracy increased to 99% when the samples were first divided into two groups based on features of their compound eyes.DAIIS can therefore be a useful tool for developing a system of automated insect identification.

View Article: PubMed Central - PubMed

Affiliation: Department of Entomology, China Agricultural University, Beijing, China.

ABSTRACT
For some insect groups, wing outline is an important character for species identification. We have constructed a program as the integral part of an automated system to identify insects based on wing outlines (DAIIS). This program includes two main functions: (1) outline digitization and Elliptic Fourier transformation and (2) classifier model training by pattern recognition of support vector machines and model validation. To demonstrate the utility of this program, a sample of 120 owlflies (Neuroptera: Ascalaphidae) was split into training and validation sets. After training, the sample was sorted into seven species using this tool. In five repeated experiments, the mean accuracy for identification of each species ranged from 90% to 98%. The accuracy increased to 99% when the samples were first divided into two groups based on features of their compound eyes. DAIIS can therefore be a useful tool for developing a system of automated insect identification.

No MeSH data available.


Related in: MedlinePlus