Automatic insect classification with Machine Learning techniques: a comparison of similarity and feature extraction approaches
Resumo
Insects are intimately related to human beings, in both positive and negative ways. For example, insect pests consume and destroy around US$40 billion worth of food each year. In contrast, insects pollinate at least two-thirds of all the food consumed in the world, with bees alone responsible for pollinating one-third of this total. In the last decades, many researchers have developed an arsenal of chemical, biological, mechanical and educational methods of insect control. However, to be effectively used, such methods require knowledge of the spatio-temporal distribution of the insects. Without such knowledge, the use of these techniques becomes costly and inefficient. A sensor for capturing insect information is being developed with the aim of being used as a tool to assist in the control of disease vectors and agricultural pests. The main elements of this sensor are a laser beam and an array of phototransistors. When an insect crosses the laser beam, a variation in the light is caused by partial occlusion of light due to their movements. This variation is stored as a time series and should be used to count and classify insects that cross the sensor. In this paper, we investigate the use of different approaches for time series classification that can be applied to insect recognition by the laser sensor: similarity search and feature extraction. In an experiment that includes nine species of insects, we demonstrate that the feature extraction approach can be more accurate that the similarity search. More specifically, the Support Vector Machine algorithm with RBF kernel trained with mel-cepstral coefficients achieved the best accuracy in the insect recognition task.
Referências
Batista, G. E. A. P. A., Keogh, E. J., Mafra-Neto, A., and Rowton, E. (2011b). SIGKDD demo: sensors and software to allow computational entomology, an emerging application of data mining. In ACM International Conference on Knowledge Discovery and Data Mining, pages 761–764. ACM.
Batista, G. E. A. P. A., Wang, X., and Keogh, E. J. (2011c). A complexity-invariant distance measure for time series. In SIAM International Conference on Data Mining, pages 699–710.
Benedict, M. Q. and Robinson, A. S. (2003). The first releases of transgenic mosquitoes: an argument for the sterile insect technique. Trends in parasitology, 19(8):349–355.
Boll, S. (1979). Suppression of acoustic noise in speech using spectral subtraction. IEEE Transactions on Acoustics, Speech and Signal Processing, 27(2):113–120.
Chaves, V., Travieso, C., Camacho, A., and Alonso, J. (2012). Katydids acoustic classification on verification approach based on mfcc and hmm. In IEEE International Conference on Intelligent Engineering Systems, pages 561–566.
Ganchev, T., Potamitis, I., and Fakotakis, N. (2007). Acoustic monitoring of singing insects. In IEEE International Conference on Acoustics, Speech and Signal Processing, volume 4, pages 721–724.
Hall, M. A. (1999). Correlation-based feature selection for machine learning. PhD thesis, The University of Waikato.
Hsu, C. W., Chang, C. C., and Lin, C. J. (2003). A practical guide to support vector classification. Technical report, Department of Computer Science, National Taiwan University.
Kahn, M. C., Celestin,W., and Jr,W. O. (1945). Recording of sounds produced by certain disease-carrying mosquitoes. Science, 101(2622):335–336.
Kahn, M. C. and Jr, W. O. (1949). The identification of certain west african mosquitoes by sound. The American Journal of Tropical Medicine and Hygiene, 29(5):827–836.
Kevan, P. (1999). Pollinators as bioindicators of the state of the environment: species, activity and diversity. Agriculture, Ecosystems & Environment, 74(1-3):373–393.
Kononenko, I. (1994). Estimating attributes: analysis and extensions of relief. In European Conference on Machine Learning, pages 171–182.
Le-Qing, Z. (2011). Insect Sound Recognition Based on MFCC and PNN. International Conference on Multimedia and Signal Processing, 2:42–46.
Moore, A. (1991). Artificial neural network trained to identify mosquitoes in flight. Journal of Insect Behavior, 4:391–396.
Moore, A. (1998). Development of a data acquisition system for long-term outdoor recording of insect flight activity using a photosensor. In Conference on Aerobiology and Biometeorology. American Meteorological Society.
Moore, A., Miller, J. R., Tabashnik, B. E., and Gage, S. H. (1986). Automated identification of flying insects by analysis of wingbeat frequencies. Journal of Economic Entomology, 79(6):1703–1706.
Park, T. H. (2004). Towards Automatic Musical Instrument Timbre Recognition. PhD thesis, Princeton University.
Patnaik, J. L., Juliusson, L., and Vogt, R. L. (2007). Environmental predictors of human West Nile virus infections, Colorado. Emerging Infectious Diseases, 13(11):1788– 1790.
Pimentel, D. (2009). Environmental and economic costs of the application of pesticides primarily in the United States. In Integrated Pest Management: Innovation- Development Process, pages 89–111.
Potamitis, I., Ganchev, T., and Fakotakis, N. (2007). Automatic acoustic identification of crickets and cicadas. In International Symposium on Signal Processing and Its Applications, pages 1 –4.
Walker, K. (2002). A review of control methods for african malaria vector. Technical Report 108, Bureau for Global Health.
W.H.O. (2009). Dengue: guidelines for diagnosis, treatment, prevention and control. Technical report, World Health Organization.
W.H.O. (2012). The world malaria report. Technical report, World Health Organization.
