Parking lot occupancy measurement using image processing
For a detailed guide on how to use this software, read "https://github.com/haidark/N02062147/blob/master/docs/index.html"
Goal of the project:
-Given static video footage of an aerial view of a parking lot; Determine the number and locations of unoccupied parking spots Determine the number and locations of occupied parking spots
Approach:
- Given static video footage of an aerial view of a parking lot.
- Determine each parking spots location Use user-labeled ROI's to localize each parking spot (A GUI will be provided to get this information easily and save it)
- Then, use one of the following technique to decide whether a parking spot is occupied or not
- Template matching - localized image subtraction
- Canny Edge detection
- Histogram of Oreinted Gradients
- Once the ROI is converted to one of these three forms, compare it to the parking spot when it is unoccupied
- Finally, report the status of each spot
Progress:
9/3/13 - Project start
9/6/13 - Researched related work
9/10/13 - Formulated basic approach
10/3/13 - Create Github Repository
11/3/13 - Created basic framework (mid-way)
11/12/13 - Created documentation
11/14/13 - Created ROI extraction function - getROI.py generates file ROIs.txt containing user specified ROIS
11/16/13 - Create ROI comparison function - loadROI.py get templates specified by ROIs.txt gets a video of parking lot compares template to ROI in current frame of video determines state of parkinglot only done for 1 so far currently using simple image subtraction TO DO: use Canny edge or HOG or SIFT
11/23/13 - changed from simple image subtraction to extracting features with canny edge Then features are compared between template and new image implemented percentage change thresholding for comparison commented code TO DO: get better ROIs 12/4/13 - Finished up project, added command line arguments and OS compatibility changes implemented percent change thresholding