Computer Vision for
Autonomous Vehicles
2020.07.22
Seowoo Han
Automatous Vehicles (AV)
Problems to solve
using computer
vision
1. Lane Line Detection
2. Object & Road
Signs/Light
Detection
3. Steering Angle
Computation
Automatous Vehicle levels
Janai, Joel, et al. "Computer vision for autonomous vehicles: Problems, datasets and state of
the art." Arxiv (2017): arXiv-1704.
Camera for Automatous Vehicles @Tesla
Multimodal Sensor
https://www.tesla.com/ko_KR/autopilot?redirect=no
Automatous Vehicles of Tesla, 2016
https://www.tesla.com/ko_KR/autopilot?redirect=no
Lane Line Detection
(A) RGB → Grayscale,
(Gaussian) smoothing
filter
(B) Sobel operator,
threshold
(C) Apply a mask to keep
the lane in the center of
the images
(D) Hough transform
Gandikota, Rohit. "Computer Vision for Autonomous Vehicles." arXiv preprint arXiv:1812.02542 (2018).
Lane Line Detection
(A) RGB → Grayscale,
(Gaussian) smoothing
filter
(B) Sobel operator,
threshold
(C) Apply a mask to keep
the lane in the center of
the images
(D) Hough transform
Gandikota, Rohit. "Computer Vision for Autonomous Vehicles." arXiv preprint arXiv:1812.02542 (2018).
Road Sign Identification
CNN을 이용한 road sign identification
(Basic problem)
3개의 convolution layers와 하나의
pooling layer의 결과를 fully-connected
layer로 연결 후, output 얻음
정확도: 99%~100%
Gandikota, Rohit. "Computer Vision for Autonomous Vehicles." arXiv preprint arXiv:1812.02542 (2018).
Object Detection paper list
Image Recognition vs. Detection
Recognition: object가 어떤 것인지 구분
Detection: object의 존재의 유무만 판단
https://github.com/hoya012/deep_learning_object_detection
2D Object Recognition(YOLO v3(You Only Look Once v3))
Sliding window
Grid
Yolo v3 result →
Object Tracking
Object tracking flow chart
과속 카메라-현재 차량의 속도 파악
Parekh, Himani S., Darshak G. Thakore, and Udesang K. Jaliya. "A survey on object detection
and tracking methods." International Journal of Innovative Research in Computer and
Communication Engineering 2.2 (2014): 2970-2979.
Object Tracking
Object tracking methods
1. Point Tracking
움직이는 객체의 특징점을 찾는 방법이다.
2. Kernel Tracking
일정 영역 내부에 있는 움직임을 찾는 방법이다.
3. Silhouette Tracking
복잡한 형태를 단순화(실루엣) 시켜 움직임을 찾는 방
법이다.
Parekh, Himani S., Darshak G. Thakore, and Udesang K. Jaliya. "A survey on object detection
and tracking methods." International Journal of Innovative Research in Computer and
Communication Engineering 2.2 (2014): 2970-2979.
Steering Angle Computation
R. Rochan M., A. Alagammai K. and S. J., "Computer Vision Based Novel Steering Angle Calculation
for Autonomous Vehicles," 2018 Second IEEE International Conference on Robotic Computing (IRC),
Laguna Hills, CA, 2018, pp. 143-146, doi: 10.1109/IRC.2018.00029. https://github.com/lhzlhz/PilotNet

Computer vision for autonomous vehicles

  • 1.
    Computer Vision for AutonomousVehicles 2020.07.22 Seowoo Han
  • 2.
    Automatous Vehicles (AV) Problemsto solve using computer vision 1. Lane Line Detection 2. Object & Road Signs/Light Detection 3. Steering Angle Computation
  • 3.
    Automatous Vehicle levels Janai,Joel, et al. "Computer vision for autonomous vehicles: Problems, datasets and state of the art." Arxiv (2017): arXiv-1704.
  • 4.
    Camera for AutomatousVehicles @Tesla Multimodal Sensor https://www.tesla.com/ko_KR/autopilot?redirect=no
  • 5.
    Automatous Vehicles ofTesla, 2016 https://www.tesla.com/ko_KR/autopilot?redirect=no
  • 6.
    Lane Line Detection (A)RGB → Grayscale, (Gaussian) smoothing filter (B) Sobel operator, threshold (C) Apply a mask to keep the lane in the center of the images (D) Hough transform Gandikota, Rohit. "Computer Vision for Autonomous Vehicles." arXiv preprint arXiv:1812.02542 (2018).
  • 7.
    Lane Line Detection (A)RGB → Grayscale, (Gaussian) smoothing filter (B) Sobel operator, threshold (C) Apply a mask to keep the lane in the center of the images (D) Hough transform Gandikota, Rohit. "Computer Vision for Autonomous Vehicles." arXiv preprint arXiv:1812.02542 (2018).
  • 8.
    Road Sign Identification CNN을이용한 road sign identification (Basic problem) 3개의 convolution layers와 하나의 pooling layer의 결과를 fully-connected layer로 연결 후, output 얻음 정확도: 99%~100% Gandikota, Rohit. "Computer Vision for Autonomous Vehicles." arXiv preprint arXiv:1812.02542 (2018).
  • 9.
    Object Detection paperlist Image Recognition vs. Detection Recognition: object가 어떤 것인지 구분 Detection: object의 존재의 유무만 판단 https://github.com/hoya012/deep_learning_object_detection
  • 10.
    2D Object Recognition(YOLOv3(You Only Look Once v3)) Sliding window Grid Yolo v3 result →
  • 11.
    Object Tracking Object trackingflow chart 과속 카메라-현재 차량의 속도 파악 Parekh, Himani S., Darshak G. Thakore, and Udesang K. Jaliya. "A survey on object detection and tracking methods." International Journal of Innovative Research in Computer and Communication Engineering 2.2 (2014): 2970-2979.
  • 12.
    Object Tracking Object trackingmethods 1. Point Tracking 움직이는 객체의 특징점을 찾는 방법이다. 2. Kernel Tracking 일정 영역 내부에 있는 움직임을 찾는 방법이다. 3. Silhouette Tracking 복잡한 형태를 단순화(실루엣) 시켜 움직임을 찾는 방 법이다. Parekh, Himani S., Darshak G. Thakore, and Udesang K. Jaliya. "A survey on object detection and tracking methods." International Journal of Innovative Research in Computer and Communication Engineering 2.2 (2014): 2970-2979.
  • 13.
    Steering Angle Computation R.Rochan M., A. Alagammai K. and S. J., "Computer Vision Based Novel Steering Angle Calculation for Autonomous Vehicles," 2018 Second IEEE International Conference on Robotic Computing (IRC), Laguna Hills, CA, 2018, pp. 143-146, doi: 10.1109/IRC.2018.00029. https://github.com/lhzlhz/PilotNet