UVM Theses and Dissertations
Format:
Print
Author:
Zurn, Jane Brooks
Title:
Dept./Program:
Electrical and Computer Engineering
Year:
2005
Degree:
MS
Abstract:
This work describes a non-invasive video tracking system for measurement of rodent behavioral activity under near-infrared (NIR) illumination. The novel methods used here allows position tracking of a single rodent in the dark, when rodents are generally most active, or under visible light. The system also improves upon current video tracking methods when a rodent's coat color is of low-contrast to the background. Both real-time and video-based tracking were addressed. Open-field locomotor activity and a learning task (holeboard) were investigated under a daylight condition (illumination with wavelengths of 460.5 nm to 561.1 nm), two different wavelengths of NIR (880 nm and 940 nm), and a night-time condition (less than 5% of 940nm NIR levels).
Tracking results of the camera-based system were not significantly different from data simultaneously recorded using a widely used NIR crossbeam-based tracking system. Experimentally, locomotor travel distance under "dark" light was significantly different from the "visible" light condition, but not from the NIR condition. Holeboard learning results were similar to those of previous studies.
We also manually extracted rodent features and classified three common behaviors (sitting, walking, and rearing) using an inductive algorithm - a decision tree (ID3). In addition, we proposed the use of a time-spatial incremental decision tree (IDSR), with which new behavior instances can be used to update the existing decision tree in an on-line manner. The classification accuracy for the set of new test data was 81.3% for ID3 and 73.0% for ID5R.
Tracking results of the camera-based system were not significantly different from data simultaneously recorded using a widely used NIR crossbeam-based tracking system. Experimentally, locomotor travel distance under "dark" light was significantly different from the "visible" light condition, but not from the NIR condition. Holeboard learning results were similar to those of previous studies.
We also manually extracted rodent features and classified three common behaviors (sitting, walking, and rearing) using an inductive algorithm - a decision tree (ID3). In addition, we proposed the use of a time-spatial incremental decision tree (IDSR), with which new behavior instances can be used to update the existing decision tree in an on-line manner. The classification accuracy for the set of new test data was 81.3% for ID3 and 73.0% for ID5R.