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/