Visual Odometry

Visual Odometry with feature extraction and matching techniques to estimate camera motion.

Project Overview

Visual odometry is a technique that allows a robot or vehicle to estimate its position and orientation by analyzing sequences of images captured by a camera onboard. This project implements a basic visual odometry pipeline using feature extraction and matching algorithms.

Dependencies

The script relies on the following Python libraries:

• numpy
• cv2 (OpenCV)
• matplotlib
• m2bk.py (custom module containing dataset handling functions)
• features_algo.py (custom module containing feature extraction and matching functions)

Make sure these libraries are installed before running the script.

Script Breakdown

The core functionalities are implemented in two separate Python files:

• m2bk.py: This file defines the DatasetHandler class responsible for:

  • Loading and storing the image dataset and depth maps.
  • Reading camera intrinsic parameters.
  • Visualizing camera movement based on matched features.
  • Visualizing the estimated camera trajectory.

• features_algo.py: This file contains functions for:

  • Feature extraction using ORB algorithm.
  • Feature visualization.
  • Feature matching using FLANN matcher.
  • Filtering good matches based on distance ratio.
  • Estimating camera motion using Essential Matrix Decomposition (or PnP with RANSAC) for subsequent image pairs.
  • Building the complete camera trajectory across the entire image sequence.

Running the Script

1. Ensure you have all the required libraries installed (refer to Dependencies section).
2. Place the script (along with m2bk.py and features_algo.py) in the same directory containing the image dataset (rgb) and depth map dataset (depth).
3. Run the script using the following command in your terminal:

python visual_odometry.py

Script Output

The script generates the following outputs:

• Matched feature visualization: This visualization displays corresponding features (points) identified in consecutive image frames.
• Camera trajectory: The script estimates and visualizes the 3D trajectory of the camera throughout the image sequence.

Note

This is a basic implementation of visual odometry. More sophisticated techniques can be applied to improve the accuracy and robustness of the results, such as:

• Using more advanced feature detectors and descriptors (e.g., SIFT, SURF).
• Bundle adjustment for global optimization of camera poses and 3D structure.
• Incorporating loop closure detection to handle revisiting locations.

This project provides a starting point for exploring visual odometry concepts and experimenting with feature-based motion estimation techniques.