Video Speedometer

Title: VideoSpeedo Task Description Version: 1.0 2015/06/22 Task: Determine speed of automobiles down a road based on video camera capture files, and report with a picture summary of vehicle movement and speeds stamped on picture. Language: Python Installed system: MacOS 10.7-10.10 and Ubuntu 14.04-15.04 Run method: This program will be run from cron periodically. Inputs: Directory structure containing videos: Video files will be in Quicktime Movie format. Videos will be 15fps. Files will be 1920x1080. Files are often large - 45 to 90 megabytes in some cases. Directory structure will look like: TowerCam <directory> | | -> 2015-08-01 <directory> 2015-08-01 15-43-22 MD [login to view URL] 2015-08-01 17-31-01 MD [login to view URL] |-> 2015-08-02 <directory> 2015-08-02 20-01-38 MD [login to view URL] etc. etc. Method: The script must parse through all videos in the directory structure (multiple levels) determine if a speed check has been done (by looking at output directory and comparing filenames) and if not, it should perform a speed check of any vehicles seen on the road in the videos. Summary pictures should be inserted into a configurable directory which contains result images. There are several objects in the field of view that are "known" size and relative location to each other. The program should use one or more of these as references for determining framing, zoom level, and scale. These objects will be highlighted and their dimensions/distances as a separate attachment. They will always be in the frame of the video, though their exact placement is variable due to camera pan/tilt/zoom inconsistencies. It will almost certainly be the case that during the day as light proceeds across the frame that all of these objects may have periods of time where they are difficult to see or are invisible, so using all of them may be required for accurate measurement. Output: JPG image containing a 800x449 image split down the middle left to right. This picture will contain the first and last frames of the video capture, with the left/right images depending on which way the motion was detected as passing through the frame. In an example of a vehicle passing from left to right: the left half of the picture will be the left half of a frame X.Y seconds into the video, and the right half of the JPG will be the right half of a frame Z.A second into the video. (where X.Y is something like 1.5 and Z.A is 2.0 - variable based on experimentation.) A white 4 point line should vertically bisect the image to cue the viewer that there are two frames shown. [snip] Due to 4k limitations, full description cannot be shown. Please refer to [login to view URL] for full description.