Lane Detection For ADAS
Applications
ECE 551 – IMAGE PROCESSING
Surya Teja Swamy
California State University Northridge
Objective
 Automotive industry is one of the growing industries and most of the
automotive companies are working on Autonomous Vehicle.
 Lane detection is one of the important aspects of Advanced Driver Assistance
Systems (ADAS), which increases the scope of autonomous driving.
 In ADAS applications, Lane detection can be achieved from various sensor
inputs, one such sensor that can be used is camera sensor.
 When we use the camera output for the Lane Detection, then most of the
image processing techniques that we covered in class can be used.
Stages
 Pre-Processing
 Edge Detection
 ROI (Region of Interest)
 Hough Transformation
Pre-Processing
 Gaussian Filtering
 Masking
Edge Detection
 After applying the masks, to detect the edges of the lane markings, I’m using
canny edge detection technique.
Region of Interest
 To define ROI, either use an ROI variables mat file or select the ROI using
figure tools in matlab.
Hough Transform
 Hough Transformation is a line-fitting algorithm to fit the lane line markings.
Room for Improvement
 In this project, I’m applying lane detection algo for frame-by-frame data instead
I can read a video file and read the frames and can apply the same algo.
 Here I’m detecting the lanes, for the next step I can predict the direction and
provide the turn directions.
 Also there will be several corner case scenarios, so this algorithm shall work in
such cases as well, it requires rigorous training and testing of the data.
 During Intensity thresholding there are hard-coded values, to improve efficiency
I’m trying to automate the threshold selection from image data.
Conclusion
 In order to Conclude, able to detect the lanes by applying different image
processing techniques in different scenarios.
Reference
 https://www.mathworks.com/help/gpucoder/ug/lane-detection-using-houghl
ines.html
 https://www.mathworks.com/help/hdlcoder/ug/lane-detection.html
 https://matlabhelper.com/blog/matlab/lane-detection-system/

Image_Processing_Lanes_Detection_project

  • 1.
    Lane Detection ForADAS Applications ECE 551 – IMAGE PROCESSING Surya Teja Swamy California State University Northridge
  • 2.
    Objective  Automotive industryis one of the growing industries and most of the automotive companies are working on Autonomous Vehicle.  Lane detection is one of the important aspects of Advanced Driver Assistance Systems (ADAS), which increases the scope of autonomous driving.  In ADAS applications, Lane detection can be achieved from various sensor inputs, one such sensor that can be used is camera sensor.  When we use the camera output for the Lane Detection, then most of the image processing techniques that we covered in class can be used.
  • 3.
    Stages  Pre-Processing  EdgeDetection  ROI (Region of Interest)  Hough Transformation
  • 4.
  • 5.
    Edge Detection  Afterapplying the masks, to detect the edges of the lane markings, I’m using canny edge detection technique.
  • 6.
    Region of Interest To define ROI, either use an ROI variables mat file or select the ROI using figure tools in matlab.
  • 7.
    Hough Transform  HoughTransformation is a line-fitting algorithm to fit the lane line markings.
  • 9.
    Room for Improvement In this project, I’m applying lane detection algo for frame-by-frame data instead I can read a video file and read the frames and can apply the same algo.  Here I’m detecting the lanes, for the next step I can predict the direction and provide the turn directions.  Also there will be several corner case scenarios, so this algorithm shall work in such cases as well, it requires rigorous training and testing of the data.  During Intensity thresholding there are hard-coded values, to improve efficiency I’m trying to automate the threshold selection from image data. Conclusion  In order to Conclude, able to detect the lanes by applying different image processing techniques in different scenarios.
  • 10.