Project Introduction

This project aims to provide a concise and efficient binocular vision processing framework for the field of autonomous driving. It generates disparity maps through stereo matching of left and right im

Background and Objectives

Background

In the autonomous driving (ADAS) system, environmental perception is one of the key technologies. Binocular vision has the advantages of low cost and simple structure, but it still needs efficient and accurate algorithm support to extract obstacles and road information.

Project Objectives

1. Provide a binocular vision processing framework that can be learned and experimented with.

2. Demonstrate how to use tools such as OpenCV to achieve stereo matching, ground detection and obstacle detection.

3. Explore the application of columnar pixel (Stixel) technology in autonomous driving scenarios.

Project Highlights

1. Modular design Functional modules are relatively independent and easy to expand and debug.

2. Algorithm combination Use V-Disparity to detect the ground area and combine columnar pixels to quickly extract foreground objects.

3. Multiple datasets support Compatible with KITTI dataset by default, and can be extended to other binocular datasets.

Applicable scenarios

• Autonomous driving: Provides basic functions for perception modules.

• Computer vision learning: Can be used as an example project for binocular vision and environmental perception.

• Algorithm verification: Quickly evaluate the performance of binocular vision algorithms in different scenarios.