- get video from IP camera or webcam
- control wifi car to move
- find direction when move under vehicles by edge detection
- stitch images to generate a big picture of vehicle bottom(not achieved yet)
- opencv3 and qt5
- run a TCPsocket in a thread to communicate with remote control IP(format IP:port)
- convert image format of opencv to Qimage
- set IP address of IP camera or webcam and control address
- set some param for test
- find direction by edge of the vehicle at two sides.
- mainly depend on two features: large gradient and high brightness around the wanted edges
- steps:
- find gradient threshold using otsu method
- find edge by canny using the gradient threshold
- find brightness threshold around edge using otsu method
- find edge by binary image using the threshold of brightness
- combine both edge
- remove unwanted edge
- gradient angle greater than 30 degree(because we believe direction will not change more than 30)
- gradient direction greater than 0 in the Top half, smaller than 0 in the bottom half(because we know background is white and vehicle bottom is black)
- repeat above steps for one time, just change first step a little bit: find gradient threshold using otsu method around edge(this step will enhance the detected edge because we use more accurate edges)
- find lines, using hough method
- find two clusters of lines, one is the top half, the other is the bottom half.
- predict lines by one line in top half and one line in bottom half.