Abstract:

Image classification features and examples of statistical results for the data mining approach using a one-versus-one strategy to implement a SVM (support vector machine) multi-class classifier. Data published in: Fefilatyev, S., K. Kramer, L. Hall, D. Goldgof, R. Kasturi, A. Remsen, K. Daly. 2011. Detection of Anomalous Particles from the Deepwater Horizon Oil Spill Using the SIPPER3 Underwater Imaging Platform. Proceedings of International Conference on Data Mining Workshops, p. 741-748. Awarded Data Mining Practice Prize at the IEEE International Conference on Data Mining (ICDM), Vancouver, Canada, December 11-14, 2011. DOI 10.1109/ICDMW.2011.65.

Suggested Citation:

Fefilatyev, Sergiy et al.. 2015. Application of image processing and machine learning techniques to distinguish suspected oil droplets from plankton and other particles for the SIPPER imaging system. Distributed by: GRIIDC, Harte Research Institute, Texas A&M University–Corpus Christi. doi:10.7266/N74X55RS

Publications:

Fefilatyev, S., Kramer, K., Hall, L., Goldgof, D., Kasturi, R., Remsen, A., & Daly, K. (2011). Detection of Anomalous Particles from the Deepwater Horizon Oil Spill Using the SIPPER3 Underwater Imaging Platform. 2011 IEEE 11th International Conference on Data Mining Workshops. doi:10.1109/icdmw.2011.65

Purpose:

The purpose of this study was to investigate a data mining approach to detect suspected oil droplets using SIPPER (Shadow Imaging Particle Profiler and Evaluation Recorder) images.

Data Parameters and Units:

Performance of oil droplet image detection algorithms using different kinds of image features for classifying zooplankton, particles, and oil droplets. Units are counts or % as indicated in the tables and below. Parameters include: Table 1 Category Sub-category Feature count Table 2 Category of classes of particles # classes in category (counts) Table 3 Dataset identifier Particles total (counts) Oil droplets (counts) Table 4 Oil droplet (counts or %) Other (counts or %) Count (counts) Table 5 Oil droplet (counts or %) Other (counts or %) Count (counts) Table 6 Oil droplet (counts or %) Other (counts or %) Count (counts) Table 7 Oil droplet (counts or %) Other (counts or %) Count (counts) Table 8 Oil droplet (counts or %) Other (counts or %) Count (counts) Table 9 Oil droplet (counts or %) Other (counts or %) Count (counts) Table 10 Oil droplet (counts or %) Other (counts or %) Count (counts) Table 11 Oil droplet (counts or %) Other (counts or %) Count (counts)

Methods:

(Fefilatyev et al. 2011) A custom designed software, the Plankton Imaging Classification Extraction System (PICES), was used to quickly extract, classify, manage and analyze the plankton images obtained by the SIPPER. A database management subsystem within PICES allows management of the large amount of data generated by SIPPER. PICES uses a simple algorithm to extract images of separate particles based on foreground-background segmentation and a connected components algorithm. After segmenting the image of a particle, a number of features are calculated/extracted and a feature vector is created. The features are used by a classification algorithm in order to assign a class label to the image. PICES uses a trained support vector machine (SVM) to classify particles in a supervised manner. We used SVM with a one-versus-one strategy for every possible two-class combination by selecting the winning class using voting. The feature selection process consisted of two steps: initial SVM parameter tuning and binary feature selection. The parameters (gamma, C, A) of the SVM were optimized by performing a grid-search with an interval across the training dataset. Using the SVM parameters determined in the first stage of the selection process, binary class feature selection was performed using wrappers. Each combination of features and SVM parameters was evaluated using 5-fold cross validation and the classification accuracy on the training set was used to guide the selection process further. In cases, where the classification accuracy is equal for several evaluated sets, the correctness of probability (CPP) was used to rank the sets.

Instruments:

SIPPER

Provenance and Historical References:

Fefilatyev, S., Kramer, K., Hall, L., Goldgof, D., Kasturi, R., Remsen, A., Daly, K. (2011). Detection of anomalous particles from Deepwater Horizon oil spill using SIPPER3 underwater imaging platform. In IEEE International Conference on Data Mining. Fourth Workshop on Data Mining Case Studies.

Individual Files: