Identification of Migratory Insects from their Physical Features using a Decision-Tree Support Vector Machine and its Application to Radar Entomology
| Autor: | Xiaowei Fu, Rui Wang, Teng Long, Cheng Hu, Shaoyang Kong |
|---|---|
| Jazyk: | angličtina |
| Rok vydání: | 2018 |
| Předmět: |
0106 biological sciences
Entomology Insecta Support Vector Machine Computer science Population 0211 other engineering and technologies Decision tree lcsh:Medicine 02 engineering and technology 01 natural sciences Article law.invention law Species identification Animals Radar lcsh:Science education 021101 geological & geomatics engineering education.field_of_study Multidisciplinary business.industry lcsh:R fungi Decision Trees Pattern recognition Support vector machine 010602 entomology Insect migration lcsh:Q Animal Migration Artificial intelligence business |
| Zdroj: | Scientific Reports Scientific Reports, Vol 8, Iss 1, Pp 1-11 (2018) |
| ISSN: | 2045-2322 |
| Popis: | Migration is a key process in the population dynamics of numerous insect species, including many that are pests or vectors of disease. Identification of insect migrants is critically important to studies of insect migration. Radar is an effective means of monitoring nocturnal insect migrants. However, species identification of migrating insects is often unachievable with current radar technology. Special-purpose entomological radar can measure radar cross-sections (RCSs) from which the insect mass, wingbeat frequency and body length-to-width ratio (a measure of morphological form) can be estimated. These features may be valuable for species identification. This paper explores the identification of insect migrants based on the mass, wingbeat frequency and length-to-width ratio, and body length is also introduced to assess the benefit of adding another variable. A total of 23 species of migratory insects captured by a searchlight trap are used to develop a classification model based on decision-tree support vector machine method. The results reveal that the identification accuracy exceeds 80% for all species if the mass, wingbeat frequency and length-to-width ratio are utilized, and the addition of body length is shown to further increase accuracy. It is also shown that improving the precision of the measurements leads to increased identification accuracy. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|