Cubicle Detection

Cubicle Detection package handles stereo matching for range estimation, obstacle detection, object classification and tracking, using Darknet as backend.

It contains three main modules:

• Stereo matching and disparity image generation

• Obstacle detection and Slope estimation (Class ObstaclesDetection)

• Object classification and tracking (Namespace darknet_ros)

Stereo matching is performed using GPU for real time performance and other modules run in CPU.

Features

• Obstacle and object detection range 5m - 45m

• Slope detection range 5m - 30m

• Multiple object classification

• Multiple object tracking

Cubicle Detection Contents

• Camera Calibration
• How to Train Object Detection Net
  • Labeling
  • Optimize anchors
  • Training
  • Stop Training
  • After Training
• 1. Stereo Matching
  • 1.1. Camera Model
  • 1.2. 3D Estimation
  • 1.3. Image Rectification
  • 1.4. Stereo Matching using SGM
• 2. Obstacle Detection and Slope Estimation
  • 2.1. Road Extraction
  • 2.2. Positive Obstacle Detection
  • 2.3. Negative Obstacle Detection
  • 2.4. Road Slope Estimation
  • 2.5. Road Pitch Estimation
• 3. Multiple Object Classification and Tracking
  • 3.1. Multiple Object Classification
  • 3.2. Multiple Object Tracking

API REFERENCE

• Class Hierarchy
• File Hierarchy
• Full API
  • Namespaces
  • Classes and Structs
  • Functions
  • Variables
  • Defines
  • Typedefs