The Smart Vivarium Abstract Problem Our Solution

The Smart Vivarium Abstract Problem Our Solution www.phwiki.com

The Smart Vivarium Abstract Problem Our Solution

Heninger, Sean, Host has reference to this Academic Journal, PHwiki organized this Journal The Smart Vivarium Serge Belongie, Kristin Branson, Keith Jenné, Vincent Rabaud, Phil Richter, Geert Schmid-Schoenbein, John Wesson http://vision.ucsd.edu Abstract UCSD’s vivariums contain thous in addition to s of cages of mice, making close monitoring of individual mice impossible. Both animal care in addition to experimental data collection would be improved by constant monitoring of mice. The goal of the Smart Vivarium Project is to automatically analyze the behavior of mice in a cage from video surveillance. The first step in automated behavior analysis of individual mice involves estimating the position of each mouse in each frame of video. Next, the mouse positions in addition to domain-specific cues are used to label the instantaneous behaviors of each mouse. Finally, the individual mouse behaviors over time are catalogued in addition to analyzed to determine the health of each mouse. Problem UCSD in addition to other research centers’ vivariums contain many thous in addition to s of mice. Because of the huge number of mice, even the simplest monitoring tasks are time-consuming in addition to expensive. Close monitoring of the mice is prohibitively expensive.

Delta University EG www.phwiki.com

This Particular University is Related to this Particular Journal

Figure 1: A room full of racks of cages of mice. Our Solution We are creating a system to automatically monitor mouse behavior from video surveillance. The system uses computer vision in addition to artificial intelligence techniques to track individual mice in addition to identify their behaviors. We there as long as e call our system the Smart Vivarium. System Requirements Our system must be nonintrusive; it must not affect the well-being or behavior of the mice. Our system must be general: it must identify many different types of behaviors. For practical reasons, our system must be easy to add to existing vivariums. As mice are nocturnal, our system must function with little visible light. To our knowledge, ours is the only system that meets these requirements.

System Goals Our system will: Per as long as m basic management tasks. Analyze animal well-being in addition to health. Log reproduction statistics. Management Management tasks include: Census measurement. Supply monitoring. Waste monitoring. Identification of individual mice. Environment monitoring. Census Measurement Census measurement tasks include: Identifying individual mouse properties, e.g. breed in addition to age. Counting the number of mice per cage. Counting the number of cages per room. Figure 2: Example images of three mice in a cage.

Supply Monitoring The state of supplies must be monitored: Food, Water, Bedding, Enrichment, Caging materials Waste Monitoring Metabolic products Ammonia Heat – humidity Feces Identification The identities of each mouse must be tracked.

Environment Monitoring Room Air Light Humidity Light intensity Location in room Equipment Percent of time cage open Basic Well Being When evaluating the health of a mouse, the following factors are pertinent: Body composition. Activities (normal in addition to abnormal) per as long as med. Amount of food in addition to water eaten in addition to drunk. Amount of urination in addition to defecation. Body composition Body composition parameters are: Body mass Lean body mass Body fat Total body water Percent body water Change in total body water per unit time

Normal Activity Monitoring normal activity levels requires measuring: Total number of steps Total distance traveled Amount of time in motion Amount of time climbing Average velocity Grooming Sleeping/Resting Exploratory novel stimuli Abnormal Activity Abnormal behaviors include: Circling Rolling Stumbling Scratching Death Drinking in addition to Eating Time spent at water lick Volume of water intake Frequency of water intake Time spent eating Volume of food intake Frequency of food intake Food preference

Urination in addition to Defecation Volume of urine production Urine specific gravity Frequency of micturation Volume of feces production Frequency of elimination Fecal water content Reproduction The following reproduction activities must be monitored: Breeding Nesting Parturition Lactation/nursing Breeding in addition to Nesting Number of matings Mating time per mating Total mating time Completed breeding Time producing a nest

Parturition in addition to Lactation Time in delivery Total animals per litter Total animals lost per litter Total nursing time Nursing frequency System Description Our system uses video taken from the side of the cage. For night surveillance, we plan on using a near-IR camera. We first estimate the position of each mouse in each frame of video. Next, the mouse positions in addition to domain-specific cues are used to label the instantaneous behaviors of each mouse. Finally, the individual mouse behaviors over time are catalogued in addition to analyzed to determine the health of each mouse. Video sequence Identity Tracking Mouse Positions Behavior Analysis Behavior Identification Behavior Labels Health Reports

Heninger, Sean JOX Roundtable - WJOX-FM, The Host www.phwiki.com

Tracking Identities Tracking mice in a cage is a uniquely challenging tracking task because The mice are virtually identical. The mice have few, if any, trackable features. The motion of the mice is erratic. We thus created a new tracking algorithm to track mouse identities, described in: “Three Brown Mice: See How They Run” to appear in VS-PETS 2003. Preliminary tracking results are below: Figure: Example frames showing the raw image frames from an occlusion event in the top row in addition to the Gaussian parameters estimated by our algorithm. The ellipses correspond to 2 st in addition to ard deviations of the Gaussians. (a) Frames 49, 64, 80, 104. (b) Frames 516, 525, 539, 556. (a) (t,x) raw image data (360 × 776 pixels): a single scanline of the image at every frame. (b) (t,x) predicted image (360 × 776 pixels): membership of points in a scanline of the image at every frame. Figure: Tracking results (t,x) plot of results. The x-axis in these images is time in addition to the y-axis is the x-axis of the original frame. Each column corresponds to the same scanline of a different frame.

Behavior Analysis The next step is to label the behavior of each mouse in each frame. Our current system labels the following behaviors: Sitting St in addition to ing Walking Cleaning It also keeps track of the number of steps taken by each individual mouse. Please view the video demos as long as more tracking in addition to behavior analysis results.

Heninger, Sean Host

Heninger, Sean is from United States and they belong to JOX Roundtable – WJOX-FM, The and they are from  Birmingham, United States got related to this Particular Journal. and Heninger, Sean deal with the subjects like Sports

Journal Ratings by Delta University

This Particular Journal got reviewed and rated by Delta University and short form of this particular Institution is EG and gave this Journal an Excellent Rating.