Skip to content

Latest commit

 

History

History
46 lines (29 loc) · 934 Bytes

README.md

File metadata and controls

46 lines (29 loc) · 934 Bytes

Collision-Avoidance-System

A Vision Based Sense and Avoid System for Small UAV's.

Input:

Border and Filter:

Video:


Files :

main.py -> Video Processing and performing ops in order.

src/horizon.py -> Horizon class for predicting horizon and also keeping track of previousily detected horizons to calculate EMWA.

src/detect.py -> Detect small objects by comparing with previous frames and also by removing False Positives.

src/validation.py -> Used to find the predicted bbox iou with the truth. Run the file to look at the ground truth values.


TODO :

  • Border and CMO filter
  • Obstacle Detection
  • Tracking obstacle

Tried :

  • CNN
  • YOLO
  • LK
  • HMM
  • Temporal Filtering
  • Adaptive Thresholding

Get the dataset at : https://engineering.purdue.edu/~bouman/UAV_Dataset/