Two-dimensional and three-dimensional sliding window-based methods and systems for detecting vehicles
Provided is a method and system for efficient localization in still images. According to one exemplary method, a sliding window-based 2-D (Dimensional) space search is performed to detect a parked vehicle in a video frame acquired from a fixed parking occupancy video camera including a field of view associated with a parking region.
1. A computer implemented method of detecting a vehicle in a video frame, the video frame acquired from a fixed parking occupancy video camera including a field of view associated with a vehicle parking region, the method comprising:
a) capturing a video frame from the fixed parking occupancy video camera, the video frame including a ROI (Region of Interest) oriented by an orientation angle relative to an orientation of an image plane associated with the captured video frame, the ROI including one or more parking spaces of the vehicle parking region;
b) rotating the ROI by the orientation angle of the ROI; and
c) performing a sliding window-based 2-D (Dimensional) space search for a vehicle within the rotated ROI, the sliding window-based 2-D space search extracting one or more features associated with each of a plurality of windows and accessing a classifier to classify each window as including a vehicle or not including a vehicle.
2. The computer-implemented method of detecting a vehicle according to claim 1 , further comprising:
d) suppressing one or more windows including a common vehicle to eliminate overlapping windows detecting the common vehicle.
3. The computer implemented method of detecting a vehicle according to claim 1 , the method comprising:
performing steps a)-c) for a plurality of captured video frames.
4. The computer implemented method of detecting a vehicle according to claim 1 , wherein the ROI is defined by a longitudinal axis, and the orientation angle is defined by the longitudinal axis of the ROI relative to the orientation of an axis associated with the image plane.
5. The computer implemented method of detecting a vehicle according to claim 1 , wherein step c) performs the sliding window-based 2-D space search along an x-axis associated with the ROI, at various window widths, while the windows maintain a fixed height.
6. The computer implemented method of detecting a vehicle according to claim 1 , wherein step c) performs the sliding window-based 2-D space search along an x-axis associated with the ROI, at various window widths, while a height of the ROI changes relative to a street curb associated with the ROI.
7. The computer implemented method of detecting a vehicle according to claim 1 , wherein step c) performs a sliding window-based 3-D space search for a vehicle within the rotated ROI, the sliding window-based 3-D space search performing the search along an x-axis associated with the ROI, at various window widths and various window heights.
8. The computer implemented method of detecting a vehicle according to claim 1 , wherein step b) rotates the captured video frame including the ROI, and the ROI is cropped from the captured video frame.
9. The computer implemented method of detecting a vehicle according to claim 1 , wherein the ROI is defined as one of a polygon, an ellipse, and a parallelogram shape.
10. The computer implemented method of detecting a vehicle according to claim 1 , wherein the classifier is trained with video frames acquired from the fixed parking occupancy video camera.
11. The computer implemented method of detecting a vehicle according to claim 10 , wherein the classifier is trained with positive samples of video including a detected vehicle and negative samples of video not including a detected vehicle.
12. The computer implemented method of detecting a vehicle according to claim 11 , wherein positive and negative samples are extracted from the rotated ROI for training the classifier.
13. The computer implemented method of detecting a vehicle according to claim 1 , wherein the parking region includes one or more of street parking, spot parking and parallel parking.
14. A vehicle detection system associated with a vehicle parking region, the vehicle detection system comprising:
a parking occupancy video camera directed towards the vehicle parking region; and
a controller operatively associated with the parking occupancy video camera, the controller configured to execute computer instructions to perform a process of detecting a vehicle in a video frame including:
a) capturing a video frame from the parking occupancy video camera, the video frame including a ROI (Region of Interest) orientated by an orientation angle, relative to an orientation of an image plane associated with the captured video frame, the ROI including one or more parking spaces of the vehicle parking region;
b) rotating the ROI by the orientation angle of the ROI; and
c) performing a sliding window-based 2-D (Dimensional) space search for a vehicle within the rotated ROI, the sliding window-based 2-D space search extracting one or more features associated with each of a plurality of windows and accessing a classifier to classify each window as including a vehicle or not including a vehicle.
15. The vehicle detection system according to claim 14 , the process further comprising:
d) suppressing one or more windows including a common vehicle to eliminate overlapping windows detecting the common vehicle.
16. The vehicle detection system according to claim 14 , the process comprising:
performing steps a)-c) for a plurality of captured video frames.
17. The vehicle detection system according to claim 14 , wherein the ROI is defined by a longitudinal axis, and the orientation angle is defined by the longitudinal axis of the ROI relative to the orientation of an axis associated with the image plane.
18. The vehicle detection system according to claim 14 , wherein step c) performs the sliding window-based 2-D space search along an x-axis associated with the ROI, at various window widths, while the windows maintain a fixed height.
19. The vehicle detection system according to claim 14 , wherein step c) performs the sliding window-based 2-D space search along an x-axis associated with the ROI, at various window widths, while a height of the ROI changes relative to a street curb associated with the ROI.
20. The vehicle detection system according to claim 14 , wherein step c) performs a sliding window-based 3-D space search for a vehicle within the rotated ROI, the sliding window-based 3-D space search performing along an x-axis associated with the ROI, at various window widths and various window heights.
21. The vehicle detection system according to claim 14 , wherein step b) rotates the captured video frame including the ROI, and the ROI is cropped from the captured video frame.
22. The vehicle detection system according to claim 14 , wherein the ROI is defined as one of a polygon, an ellipse, and a parallelogram shape.
23. The vehicle detection system according to claim 14 , wherein the classifier is trained with video frames acquired from the fixed parking occupancy video camera.
24. The vehicle detection system according to claim 23 , wherein the classifier is trained with positive samples of video including a detected vehicle and negative samples of video not including a detected vehicle.
25. The vehicle detection system according to claim 14 , wherein the parking region includes one or more of street parking, spot parking and parallel parking.
26. A computer implemented method of determining parking occupancy associated with a vehicle parking region comprising:
a) capturing a video frame from a parking occupancy video camera directed towards the vehicle parking region;
b) rotating the captured video frame by a predetermined orientation angle of a longitudinal axis associated with a ROI (Region of Interest) associated with the captured video frame, relative to an orientation of an axis associated with an image plane associated with the captured video frame, the ROI including one or more parking spaces of the parking region; and
c) performing a sliding window-based 2-D (Dimensional) space search for a vehicle within the ROI associated with the rotated captured video frame, the sliding window-based 2-D space search extracting one or more features associated with each of a plurality of windows and accessing a classifier to classify each window as including a vehicle or not including a vehicle.
27. The computer implemented method of determining parking occupancy according to claim 26 , wherein step c) performs non-maxima suppression.
28. The computer implemented method of determining parking occupancy according to claim 26 , wherein
steps a)-c) are performed for a plurality of captured video frames; and
step c) performs the sliding window-based 2-D space search along an x-axis associated with the ROI at various window widths, and the classifier is trained with video frames acquired from the fixed parking occupancy camera, the video frames including positive samples of video including a detected vehicle and negative samples of video not including a detected vehicle.
29. The computer implemented method of determining parking occupancy according to claim 28 , wherein step c) extracts one or more feature vectors associated with each of the plurality of windows, and the classifier is a SVM (Support Vector Machine).
30. A computer implemented method of detecting a vehicle in a video frame, the video frame acquired from a fixed parking occupancy video camera including a field of view associated with a vehicle parking region, the method comprising:
a) capturing a video frame from the fixed parking occupancy video camera, the video frame including a ROI (Region of Interest) oriented by an orientation angle relative to an orientation of an image plane associated with the captured video frame, the ROI including one or more parking spaces of the vehicle parking region; and
b) performing a sliding window-based 2-D (Dimensional) Space Search for a vehicle with the ROI, the sliding window-based 2-D space search extracting one or more features associated with each of a plurality of windows and accessing a classifier to classify each window as including a vehicle or not including a vehicle,
wherein step b) performs the sliding window-based 2-D space search along an orientation axis of the ROI associated with the orientation angle of the ROI relative to the orientation of the image plane.