80 likes | 100 Views
This project focuses on detecting available parking spaces in parking lots using computer vision methods. Addressing challenges such as projective distortion, occlusions, illumination changes, noise, and foreground-background separation. Solutions include geometric rectification, data collection, PCA, EM-PCA, robust PCA, parking classification, line detection, noise mitigation, and occupancy calculation.
E N D
Parking Spot Recognition from Video Footage Shani Achtenberg Liron Silfin Advisor: Oren Freifeld
GOALS • Automatic detection of free spaces in a given parking lot • Become familiar with computer vision and image processing methods
CHALLENGES • Projective distortion due to camera geometry • Occlusions due to viewing angle • Illumination changes • Noise • Separation of foreground from background
OUR SOLUTION • Geometric rectification using homography estimation
OUR SOLUTION • Data collection • PCA • EM-PCA • Robust PCA
OUR SOLUTION • Parking classification • Line detection via RANSAC
OUR SOLUTION • Mitigating pixel-level noise using super pixels of the results • Calculate occupancy percentage of each parking spot • Graphic representation