08-08-2012, 10:03 AM
Detecting, Tracking and Classifying Animals in Underwater Video
[attachment=30476]
Introduction
For oceanographic research, remotely operated underwater vehicles (ROVs) and underwater observatories routinely record several hours of video material every day. Manual processing of such large amounts of video has become a major bottleneck for scientific research based on this data.
An automated system for detecting animals (events) visible in the videos is being developed.
The videos are processed with an attentional selection algorithm that has been shown to work robustly for target detection in a variety of natural scenes.
The candidate locations identified by the attentional selection module are tracked across video frames using linear Kalman filters.
If objects can be tracked successfully over several frames, they are stored as potentially 'interesting‘ events.
Based on low-level properties, interesting events are identified and marked in the video frames.
Interesting events are then processed by a classification module trained to classify specific animal categories.
Moving Object Detection
Background Subtraction
Main goal: detecting moving objects from a video sequence of a fixed camera
Background – static scene
Foreground – moving objects
Approach: detect the moving objects as the difference between the current frame and the image of the scene background.
Places where there are differences are detected and classified as moving objects.
Adaptive Image Enhancement
Homomorphic filtering or histogram equalizations have been used for the enhancement of images with shaded regions and images degraded.
small contrasts in the very high and/or low luminance regions cannot be well detected by the human eye.
Image enhancement which enhances contrasts that were hardly seen in the original image is possible by enhancing the local contrast as well as modifying the local luminance mean for the very high and/or low luminance regions to the level where the human eye can easily detect them.
Adaptive Image Enhancement using Histogram Equalization
Histogram equalization is a common technique for enhancing the appearance of images.
Suppose we have an image which is predominantly dark. Then its histogram would be skewed towards the lower end of the grey scale and all the image detail is compressed into the dark end of the histogram.
If we could `stretch out' the grey levels at the dark end to produce a more uniformly distributed histogram then the image would become much clearer.
Attentional selection and tracking algorithm
Four main processing steps are involved in our video analysis procedure after the video has been captured.
Initially, some generic preprocessing is performed for each frame of the input video stream.
then the vicinity of locations are scanned for the occurrence of animals at which the Kalman Filter trackers predict them.
thirdly, every five frames the image is processed to find salient objects that are not yet tracked,
last step, visual events are classified into “interesting” or “boring” according to low-level properties of the tokens involved.
Motion Detection
In underwater observatories a static background is obtained by calculating an average of the first few consecutive frames.
This static background is then used to segment moving objects by comparing the background from the current frame.
So we find the moving objects after subtracting current frame from background frame
CONCLUSION
We report here our progress in developing a new method for processing video streams from underwater videos automatically.
This technology has potentially significant impact on the work of video annotators by aiding the annotators in looking for noteworthy events in the videos.
Eventually, we hope that the software will be able to perform a number of routine tasks fully automatically.
Most of this work can be done in an unsupervised, fully automated fashion.
[attachment=30476]
Introduction
For oceanographic research, remotely operated underwater vehicles (ROVs) and underwater observatories routinely record several hours of video material every day. Manual processing of such large amounts of video has become a major bottleneck for scientific research based on this data.
An automated system for detecting animals (events) visible in the videos is being developed.
The videos are processed with an attentional selection algorithm that has been shown to work robustly for target detection in a variety of natural scenes.
The candidate locations identified by the attentional selection module are tracked across video frames using linear Kalman filters.
If objects can be tracked successfully over several frames, they are stored as potentially 'interesting‘ events.
Based on low-level properties, interesting events are identified and marked in the video frames.
Interesting events are then processed by a classification module trained to classify specific animal categories.
Moving Object Detection
Background Subtraction
Main goal: detecting moving objects from a video sequence of a fixed camera
Background – static scene
Foreground – moving objects
Approach: detect the moving objects as the difference between the current frame and the image of the scene background.
Places where there are differences are detected and classified as moving objects.
Adaptive Image Enhancement
Homomorphic filtering or histogram equalizations have been used for the enhancement of images with shaded regions and images degraded.
small contrasts in the very high and/or low luminance regions cannot be well detected by the human eye.
Image enhancement which enhances contrasts that were hardly seen in the original image is possible by enhancing the local contrast as well as modifying the local luminance mean for the very high and/or low luminance regions to the level where the human eye can easily detect them.
Adaptive Image Enhancement using Histogram Equalization
Histogram equalization is a common technique for enhancing the appearance of images.
Suppose we have an image which is predominantly dark. Then its histogram would be skewed towards the lower end of the grey scale and all the image detail is compressed into the dark end of the histogram.
If we could `stretch out' the grey levels at the dark end to produce a more uniformly distributed histogram then the image would become much clearer.
Attentional selection and tracking algorithm
Four main processing steps are involved in our video analysis procedure after the video has been captured.
Initially, some generic preprocessing is performed for each frame of the input video stream.
then the vicinity of locations are scanned for the occurrence of animals at which the Kalman Filter trackers predict them.
thirdly, every five frames the image is processed to find salient objects that are not yet tracked,
last step, visual events are classified into “interesting” or “boring” according to low-level properties of the tokens involved.
Motion Detection
In underwater observatories a static background is obtained by calculating an average of the first few consecutive frames.
This static background is then used to segment moving objects by comparing the background from the current frame.
So we find the moving objects after subtracting current frame from background frame
CONCLUSION
We report here our progress in developing a new method for processing video streams from underwater videos automatically.
This technology has potentially significant impact on the work of video annotators by aiding the annotators in looking for noteworthy events in the videos.
Eventually, we hope that the software will be able to perform a number of routine tasks fully automatically.
Most of this work can be done in an unsupervised, fully automated fashion.