| Paper Details: | Downloads: 3656 |
| Serial Number: | P1151738585
|
| Title: | GreedyHaarSpiker: An Algorithm for In Situ Detection of Highway
Lane Boundaries with 1D Haar Wavelet Spikes
|
| Authors: | Vladimir A. Kulyukin
|
| Abstract: | An algorithm is presented for in situ vision-based detection of highway lane boundaries on a raspberry pi computer coupled to a raspberry pi camera. The raspberry pi unit is placed inside a Jeep Wrangler, next to the windshield, and is powered through a 12V-to-5V
car charger. The algorithm, called \textit{GreedyHaarSpiker}, is based on the detection of 1D Haar Wavelet spikes in 1D Ordered Haar Wavelet Transforms of image rows. To obtain experimental video data for daytime driving, the author drove his Jeep with the installed
raspberry pi unit on a sunny day in September 2016 (run 1) and a cloudy day with light rain in November 2016 (run 2) at a speed of 55-60 miles per hour on Route 30, a two-lane Northern Utah highway. To obtain data video data of driving on snowy roads and night driving, the author drove his Jeep on the same highway and at the same speed on a day after a heavy snowfall (run 3) in January 2017 and on the same day after sunset (run 4). Each run was approximately
35 miles long. Each video was partitioned into frames and a sample of 360 x 240 PNG consecutive frames was selected from each captured video. The performance of the algorithm was tested in
situ on a raspberry pi 3 model B ARMv8 1GB RAM computer on each of the four frame samples. The algorithm is implemented in Python 2.7.9 with OpenCV 3.0. The current implementation
processes 20 frames per second.
|
| Keywords: | Computer Vision, Wavelets, Lane Detection, Autonomous Vehicles
|
| Journal/Conference: | International Journal of Graphics, Vision and Image Processing
|
| Volume: | 17
|
| Issue: | 2
|
| Submission Date: | 9/19/2017 12:00:00 AM
|
| Review Date: | 10/24/2017 12:00:00 AM
|
| Publishing Date: | 11/17/2017 12:00:00 AM
|
| Article Downloads: | 3656
|
| Download: |
|