Parking Space DetectionNicholas TrueDepartment of Computer ScienceUniversity of California, San DiegoCalifornia, CA [email protected] enjoys circling parking lots looking for non-existant parking spaces.Thus, I propose a computer vision based ”Parking Space” detection project tohelp tackling this common problem. My project would involve placing a cameraon EBU1 to look down on parking lot P503 to detect available parking spaces.1 QualificationsI have developed sign detection software for mobile robots while interning at SPAWAR UnmannedRobotics Laboratory, San Diego. My sign detection algorithm used a combinations of edge-featuredetection as well as Adaboost-based text detection.I also developed sign reading software at SPAWAR. My solution used IBM’s old Tessaract OCRsoftware, increasing its detection rate through a combination of pre and post processing.Lastly, I am currently taking CSE 166.2 Performance Evaluation2.1 Primary Goals• The parking space detection system should be able to detect unobstucted, empty parkingspaces in good daylight conditions. Conversely, they system should have a very low falsedetection rate under the same conditions.• The system should be able to detect occupied parking spaces under good daylight condi-tions when the occupying vehicles are common cars, trucks, or vans.2.2 Secondary Goals• Detecting non-free parking spaces where the parking spaces have been occupied by unusualvehicles, such as construction vehicles and motorcycles, or cars parkedin a non-standard orillegal way (covering two spaces, partially cropping a lineetc).• Make the system robust enough to function at dusk/dawn and night as well as in the day.3 Milestones• (Week 1) Set up camera and auto generate a test set of the parking (over a 12+ hour periodof time). Sort the test set if necessary.• (week 2-4) Research and develop an algorithm for tackling parking space detection.• (Week 5-8) From the algorithm, create a preliminary program which can determine howmany parking spots are available in the current field of view.• (Week 9) If there is time, extend the algrithm to determine what types of spaces are avail-able, and determine if a car is parked illegally (double parked, covering two spaces, etc).• (Week 10) Write up final presentation.4 SoftwareAs of yet, I have not found any non-commercial software to extend upon. Thus, I intend to writemost of the necessary software myself in either Matlab or OpenCV.5 Questions• How does one deal with variances in lighting (from clouds, and shadows, etc).• How effectively/easily could this project be extended to work with real time video footage.• Could this project be made commercially viable?6 Related Papers[1] Ekinci, M. & Gedikli, E. (2005) Silhouette Based Human Motion Detection and Analysis for Real timeAutomated Video Surveillance, Turkish Journal of Electrical Engineering & Computer Sciences, Vol. 13, NO.22005.[2] Fujiyoshi, H. & Kanade, T. (2004) Layered Detection for Multiple Overlapping Objects, IEICE Trans. Inf.& Syst. Vol. E87-D, NO.12 2004.[3] Schneiderman, H., Kanade, T. (2000) A Statistical Method for 3D Object Detection Applied to Faces andCars, IEEE Converence on Computer Vision and Pattern Recognition, 2000.[4] Yilmaz, A., Li, X. Shah, M. (2004) Contour-Based Object Tracking with Occlusion Handling in VideoAcquired Using Mobile Cameras, IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 26No. 11, 2004.[5] Cucchiara, R., Costantino, G., Piccardi M., Prati, A. (2003) Detecting Moving Objects Ghosts and Shadowsin Video Streams, IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 25 No. 10, 2003.[6] Stauffer, C., Grimson E. (2000) Learning Patters of Activity Using Real-Time Tracking, IEEE Transactionson Pattern Analysis and Machine Intelligence, Vol. 22 No. 8, 2000.[7] Nadimi, S., Bhanu B. (2004) Physical Models for Moving Shadow and Object Detection in Video, IEEETransactions on Pattern Analysis and Machine Intelligence, Vol. 26 No. 8,
View Full Document
Unlocking...