Method for available parking distance estimation via vehicle side detection
View Patent ↗A computer-implemented method, system, and computer-readable medium is disclosed for determining an estimated available parking distance for a vehicle via vehicle side detection in one or more image frames from an operational video. The operational video can be acquired from a fixed parking occupancy video camera and can include a field of view associated with a parking region. The method can include obtaining operational video from a fixed parking occupancy video camera; detecting, within a region of interest (ROI) of the one or more image frames from the operational video, a side of one or more vehicles parked in a parking region facing a traffic lane using a trained classifier that is trained to detect the side of the one or more vehicles; and determining an estimated available parking distance based on the side of the one or more vehicles that are detected.
1. A computer-implemented method for determining an estimated available parking distance for a vehicle via vehicle side detection in one or more image frames from an operational video, the operational video acquired from a fixed parking occupancy video camera comprising a field of view associated with a parking region, the method comprising:
obtaining operational video from a fixed parking occupancy video camera;
detecting, within a region of interest (ROI) of the one or more image frames from the operational video, a side of one or more vehicles parked in a parking region facing a traffic lane using a trained classifier that is trained to detect the side of the one or more vehicles; and
determining an estimated available parking distance based on the side of the one or more vehicles that are detected and a look-up table (LUT), wherein the LUT stores various image pixel locations and their corresponding real-world coordinates of an outer boundary of the parking region.
2. The computer-implemented method according to claim 1 , wherein the estimated available parking distance is between adjacently parked vehicles, a front portion of a parked vehicle and a parking lane boundary, or a rear portion of a parked vehicle and a parking lane boundary.
3. The computer-implemented method according to claim 1 , wherein the fixed parking occupancy video camera comprises a field of view, wherein an image plane of the fixed occupancy video camera is oblique or parallel to a street direction on which the one or more vehicles are parked.
4. The computer-implemented method according to claim 1 , further comprising training the classifier, wherein training the classifier comprises:
obtaining additional video from the fixed parking occupancy video camera;
extracting one or more positive samples and one or more negative samples from the additional video that was obtained;
extracting a set of features from the one or more positive samples and a set of features from the one or more negative samples; and
training the classifier via machine learning techniques.
5. The computer-implemented method according to claim 1 , wherein detecting the side of the one or more vehicles further comprises performing a sliding window-based space search for the side of the vehicle within video that was obtained.
6. The computer-implemented method according to claim 5 , wherein the sliding window-based space search comprises extracting one or more features associated with each of a plurality of windows and accessing an operatively associated classifier to classify each window as including the side of the vehicle or not including the side of the vehicle.
7. The computer-implemented method according to claim 2 , wherein the distance between two vehicles is estimated based on the pixel locations of the corners of two vehicles that are facing the traffic lane between the two vehicles in the ROI.
8. The computer-implemented method according to claim 4 , wherein the set of features comprise one or more local image descriptors.
9. The computer-implemented method according to claim 8 , wherein the one or more local image descriptors comprise any, or combinations of, a color attribute-based features, texture-based features, scale invariant features, a bag of visual words, and Fisher vectors.
10. The computer-implemented method according to claim 4 , wherein the one or more positive samples and the one or more negative samples are extracted from a carved out region of the video using a hierarchical training approach.
11. The computer-implemented method according to claim 4 , wherein the one or more positive samples comprise images of sides of different vehicles parked along a street facing traffic lane.
12. The computer-implemented method according to claim 4 , wherein the one or more negative samples comprises any, or combinations of, a background image from the parking region, a rear portion of the parked vehicles in the video, a middle portion of the parked vehicles, and a front portion of the parked vehicles in the video.
13. The computer-implemented method according to claim 1 , wherein the classifier comprises a linear classifier, wherein the linear classifier is a linear Support Vector Machine (SVM).
14. The computer-implemented method according to claim 1 , wherein the classifier comprises a non-linear classifier.
15. The computer-implemented method according to claim 1 , wherein the classifier is trained using features extracted from a set of training samples and one or more labels of the training samples.
16. The computer-implemented method according to claim 1 , wherein the classifier is trained using a machine learning process.
17. A vehicle detection system associated with a parking region, the vehicle detection system comprising:
a fixed parking occupancy video camera including a field of view associated with a parking region; and
a controller operatively associated with the fixed parking occupancy video camera, the controller configured to execute computer instructions to perform a process of determining an estimated distance between vehicles from one or more image frames in operational video including:
obtaining operational video from a fixed parking occupancy video camera;
detecting, within a region of interest (ROI) of the one or more image frames from the operational video, a side of one or more vehicles parked in a parking region facing a traffic lane using a trained classifier that is trained to detect the side of the one or more vehicles; and
determining an estimated parking distance based on the side of the one or more vehicles that are detected, wherein the estimated available parking distance is between adjacently parked vehicles, a front portion of a parked vehicle and a parking lane boundary, or a rear portion of a parked vehicle and a parking lane boundary and the distance between two vehicles is estimated based on the pixel locations of the corners of two vehicles that are facing the traffic lane between the two vehicles in the ROI.
18. A non-transitory computer readable storage medium comprising instructions that cause one or more processors to perform a method comprising:
obtaining operational video from a fixed parking occupancy video camera detecting, within a region of interest (ROI) of the one or more image frames from the operational video, a side of one or more vehicles parked in a parking region facing a traffic lane using a trained classifier that is trained to detect the side of the one or more vehicles; and
determining an estimated parking distance based on the side of the one or more vehicles that are detected,
wherein the estimated available parking distance is between adjacently parked vehicles, a front portion of a parked vehicle and a parking lane boundary, or a rear portion of a parked vehicle and a parking lane boundary and the distance between two vehicles is estimated based on the pixel locations of the corners of two vehicles that are facing the traffic lane between the two vehicles in the ROI.