This project was about automatic detection of the traffic signs in real time video footage. This used colour-based method to segment and detect the road signs (HSV – Hue saturation and value). The video footages were collected using a camera mounted to a motor bicycle in the A12 highway at north central province, Sri Lanka.
The algorithm of the application can be stated as follows. Program separated each image frame of the video footage and applied the denoise techniques. Then applied thresholding by setting the threshold values manually which was experimentally determined using a separate program to the video footages. The resulted image was denoised using dilation, morphological closing and erosion techniques. Next, the canny edge detection method detected the objects from the resulted image. Finally, purification of the detected unwanted contours was achieved using the contour area.
This study was able to identify the road signs in the road with a good range of accuracy. Meanwhile there were several drawbacks which was needed to be addressed in further studies of this domain. Those drawbacks can be stated as follows. Some unwanted objects were detected, short detection distance and some colours were unable to detect.
Figure1: Selected road sign in the video |
Figure2: Thresholded frame from video |
Figure3: Selected road sign in the video |
Figure4: Thresholded frame from video |
Supervisor:
Mr. K.A.S.H Kulathilake