Scanning the streets might not need an expensive setup.
Do you remember the 8-camera video stitcher project?
Eight Raspberry Pi and corresponding camera are connected on a local network and synchronized to capture 360° panoramic images.
What if you let it grow taller and attach a bike to it?
You get a vehicle to compete with Google Street View!
Time to go on the street…
From the pictures of the video above, here is a little prototype website to roam a few meters on the street.
Try it here:
- One way: http://vjordan.info/project/street_view/view1.html
- Other way: http://vjordan.info/project/street_view/view2.html
Another idea is to scan the street into one lengthy image, that people would simply scroll left or right.
If we stitch together the frames from the camera perpendicular to the street with stock OpenCV stitcher tool, we get this:
Oups! It did not work as expected, but with some modifications of the stitching process to force picture alignment, we would be going toward a potential solution.
- Add a GPS on top to register a coordinate with each picture
For drawing map and address information
- Add a magnetometer to register the orientation of the camera
For drawing a street overlay on the picture and to align the forward/backward arrows with the actual street when the picture is panned around.
- Add a spinning wheel sensor to register precise length between each picture
So that the forward/backward step is a consistent length (e.g., 10 meters) and does not depends on the speed of the bike
- Add some AI post-processing to blur car plates and people’s faces.