Moving Object Detection based on Temporal and Spatial Resolution Mosaic
Transcript of Moving Object Detection based on Temporal and Spatial Resolution Mosaic
Moving Object Detection based on Temporal and Spatial Resolution Mosaic
Mohammed Abdel-Megeed SalemFaculty of Computer and Information Science,
Ain Shams University in [email protected]
Motivation
• Simultaneously accurate moving object detection and low computational complexity.
• Multiresolution analysis simplifies the process, improves its results, and allows real time processing.
• Representing the images in a mosaic of different resolutions
Moving Object Detection based on Temporal and Spatial Resolution Mosaic 01Mohammed A-Megeed Salem ICCES 2011
Outline
• Principles• Introduction• 3D Wavelet-based Algorithm• Mosaic-based Algorithm• Results and Analysis• Discussion and Conclusion
Moving Object Detection based on Temporal and Spatial Resolution Mosaic Mohammed A-Megeed Salem ICCES 2011
Principles• Resolution of analysis relevance of information
– High resolution => edges and objects of interest (local information)
– Low resolution => homogeneous regions and background (global information)
• Dimensionality of analysis dimensionality of information
– Images: 2D analysis => 2D spatial correlation between pixels– Videos: 3D analysis => 3D spatial and temporal features of the
motion
Moving Object Detection based on Temporal and Spatial Resolution Mosaic 02Mohammed A-Megeed Salem ICCES 2011
Outline
• Principles• Introduction• 3D Wavelet-based Algorithm• Mosaic-based Algorithm• Results and Analysis• Discussion and Conclusion
Moving Object Detection based on Temporal and Spatial Resolution Mosaic Mohammed A-Megeed Salem ICCES 2011
Introduction
Video Segmentation
Frame Differencing Optical Flow Background Subtraction
Moving Object Detection based on Temporal and Spatial Resolution Mosaic 03Mohammed A-Megeed Salem ICCES 2011
Outline
• Principles• Introduction• 3D Wavelet-based Algorithm• Mosaic-based Algorithm• Results and Analysis• Discussion and Conclusion
Moving Object Detection based on Temporal and Spatial Resolution Mosaic Mohammed A-Megeed Salem ICCES 2011
3D Wavelet-based Algorithm
Generating Binary Masks
Detection of Motion
Divide the sequence into group of frames
Input image sequence
Combine two subbands Xj = avrg(Dj4, Dj
7)
Apply the 3D wavelet transformd4
d7 x
After thresholding
After smoothing
After dilation
Thresholding
Median filtering
Dilation operation
Many levels analysis
Level of the extracted regions
Level of processing
Extraction of ROI and Active Traffic Area
Moving Object Detection based on Temporal and Spatial Resolution Mosaic 04Mohammed A-Megeed Salem ICCES 2011
3D Wavelet-based Algorithm
• The strong dependence between the resolution levels of the spatial and the temporal analyses
• All parts of the scene are analysed on the same resolution level
Detection of Motion
Input image sequence
D4
D7
XA D4 D2 D6 D1 D5 D3 D7
( )⋅φ ( )⋅ψ
Moving Object Detection based on Temporal and Spatial Resolution Mosaic 05Mohammed A-Megeed Salem ICCES 2011
Solution?• Independent spatial and temporal resolutions
– Low spatial resolution helps to avoid disturbances of the irrelevant motions that take place in background.
– Low temporal resolution means to miss the fast motion.
• Spatial resolution mosaic– Far views need to be processed in high spatial
resolution.– Close views need to be processed in low spatial
resolution.
Moving Object Detection based on Temporal and Spatial Resolution Mosaic 06Mohammed A-Megeed Salem ICCES 2011
Outline
• Principles• Introduction• 3D Wavelet-based Algorithm• Mosaic-based Algorithm• Results and Analysis• Discussion and Conclusion
Moving Object Detection based on Temporal and Spatial Resolution Mosaic Mohammed A-Megeed Salem ICCES 2011
The Mosaic Image
Resolution level
Information relevance
Temporal (3 rd) dimension
Moving Object Detection based on Temporal and Spatial Resolution Mosaic 07Mohammed A-Megeed Salem ICCES 2011
The Mosaic Map
High resolution for fine details
(small objects/far view)
.
.
.
Low resolution for global information
(big objects/close view)
Moving Object Detection based on Temporal and Spatial Resolution Mosaic 08Mohammed A-Megeed Salem ICCES 2011
The Mosaic Map
High resolution for fine details
(small objects/far view)
.
.
.
Low resolution for global information
(big objects/close view)
Moving Object Detection based on Temporal and Spatial Resolution Mosaic 08Mohammed A-Megeed Salem ICCES 2011
The Algorithm
Image(Original
resolution)
Map
Different levels 2D wavelet transform
A H V D
Spatial transformation
Spatial resolution mosaic subbands
Moving Object Detection based on Temporal and Spatial Resolution Mosaic 09Mohammed A-Megeed Salem ICCES 2011
The Algorithm
Image(Original
resolution)
Map
Different levels 2D wavelet transform
A H V D
1D (temporal) wavelet transform
Spatial transformation
Temporal transformation
Temporal-spatial subbands
AA AD HA HD VA VD DA DD
Spatial resolution mosaic subbands
Temporal 1D arrays
Moving Object Detection based on Temporal and Spatial Resolution Mosaic 10Mohammed A-Megeed Salem ICCES 2011
The Algorithm
Image(Original
resolution)
Map
Different levels 2D wavelet transform
A H V D
1D (temporal) wavelet transform
Spatial transformation
Temporal transformation
Temporal-spatial subbands
A D4 D2 D6 D1 D5 D3 D7
AA AD HA HD VA VD DA DD
Spatial resolution mosaic subbands
Temporal 1D arrays
Moving Object Detection based on Temporal and Spatial Resolution Mosaic 11Mohammed A-Megeed Salem ICCES 2011
Outline
• Principles• Introduction• 3D Wavelet-based Algorithm• Mosaic-based Algorithm• Results and Analysis• Discussion and Conclusion
Moving Object Detection based on Temporal and Spatial Resolution Mosaic Mohammed A-Megeed Salem ICCES 2011
Results and Analysis
Moving Object Detection based on Temporal and Spatial Resolution Mosaic 12Mohammed A-Megeed Salem ICCES 2011
Results and Analysis
Moving Object Detection based on Temporal and Spatial Resolution Mosaic 13Mohammed A-Megeed Salem ICCES 2011
Results and Analysis
• False Alarms (FA):– The number of extracted regions that contain
no active moving objects.• Missed Objects (MO):
– The number of active moving objects that are not included in any extracted regions.
• Delayed Detections (DD):– The number of active moving objects that are
delayed for the first time detection.
Moving Object Detection based on Temporal and Spatial Resolution Mosaic 14Mohammed A-Megeed Salem ICCES 2011
Results and Analysis3D based Algorithm Mosaic based Algorithm
Moving Object Detection based on Temporal and Spatial Resolution Mosaic 15Mohammed A-Megeed Salem ICCES 2011
Results and Analysis3D based Algorithm Mosaic based Algorithm
Moving Object Detection based on Temporal and Spatial Resolution Mosaic 16Mohammed A-Megeed Salem ICCES 2011
Outline
• Principles• Introduction• 3D Wavelet-based Algorithm• Mosaic-based Algorithm• Results and Analysis• Discussion and Conclusion
Moving Object Detection based on Temporal and Spatial Resolution Mosaic Mohammed A-Megeed Salem ICCES 2011
Discussion and Conclusion
• Automatic mosaic map
• High temporal resolution
• Time complexity is O(N2)
Moving Object Detection based on Temporal and Spatial Resolution Mosaic 17Mohammed A-Megeed Salem ICCES 2011
Discussion and Conclusion
• An algorithm based on resolution mosaic and 3D wavelet packet is proposed.
• The algorithm is based on analyzing the spatial domain independent of the temporal domain.
• A proper spatial resolution is chosen based on the relevance of the expected information.
• No need to have a compromise between the false alarm rates and the missed object rates.
Moving Object Detection based on Temporal and Spatial Resolution Mosaic 18Mohammed A-Megeed Salem ICCES 2011
ContactsMoving Object Detection based on Temporal and Spatial Resolution MosaicThe 2011 International Conference on Computer Engineering and Systems
November 29-30, December 1, 2011, Cairo, Egypt
Dr. Mohammed Abdel-Megeed M. Salem
Faculty of Computer and Information Sciences,Ain Shams University
Abbassia, Cairo, EgyptTel.: +2 011 1727 1050
Email: [email protected]