🚗 Surround View Autonomous Parking System

An autonomous parking system using Surround View Moniter (bird's eye view) and perform YOLO-based parking spot detection at Gazebo simulation.

---

📌 Overview

This project implements a full autonomous parking pipeline using a surround view camera system. Parking spots are detected using YOLOv8 and a pose estimation model, and the vehicle autonomously navigates into the target spot.

Key highlights:

• YOLOv8 pose estimation for parking spot recognition
• Modular ROS2 node architecture (camera_perception_pkg, decision_making_pkg, serial_communication_pkg)
• Supports at Gazebo simulation
• Includes calibration tools and data analysis utilities

---

🛠️ Tech Stack

Category	Tools / Libraries
Middleware	ROS 2 Humble
Simulation	Gazebo
Vision / AI	YOLOv8 (Ultralytics 8.2.69), OpenCV
Language	Python 3, Shell, CMake
Communication	pyserial (serial to motor controller)
Other	pynput, transformers, HuggingFace Hub

---

⚙️ Installation

Prerequisites

• Ubuntu 22.04
• ROS2 Humble
• Python 3.10+

Setup

1. Clone the repository:

   git clone https://github.com/yunss01/surround_view_parking.git
   cd surround_view_parking

2. Run the installation script:

   chmod +x install.sh
   ./install.sh

   This will install:

   • Python dependencies: opencv-python, pyserial, ultralytics, pynput, transformers, huggingface_hub
   • ROS 2 Gazebo packages: gazebo-ros-pkgs, xacro, gazebo-ros, gazebo-msgs, gazebo-plugins
   • Gazebo simulation model assets

3. Build the ROS 2 workspace:

   source /opt/ros/humble/setup.bash
   colcon build --symlink-install
   source ./install/local_setup.bash

---

🚀 Usage

ros2 launch simulation_pkg mission_sim.launch.py

---

🤖 Pre-trained Models

File	Description
best_pose_rev03.pt	YOLOv8 pose estimation model for parking spot orientation

---

🔬 Data Analysis & Utilities

Folder	Description
DL model/	Scripts for training and evaluating deep learning models
data_analyze/	Data exploration and visualization tools
image stdev extract/	Utilities for extracting image standard deviation (used for data quality filtering)
npy/	Preprocessed NumPy arrays for model training/testing

---

📦 ROS2 Package Architecture

surround_view_parking
├── camera_perception_pkg
│   ├── image_publisher_node          → Reads and publishes raw fisheye camera feeds
│   ├── surroundview_stitching_node   → Undistorts, warps, and stitches 4 cameras into BEV
│   ├── yolov8_node                   → Runs YOLO detection/pose on the stitched image
│   └── parking_spot_extractor_node   → Extracts valid parking slot candidates
│
├── decision_making_pkg
│   ├── parking_planner_node          → Selects target slot and computes parking path
│   └── motion_planner_node           → Generates velocity/steering commands
│
├── serial_communication_pkg
│   └── serial_sender_node            → Transmits control commands to vehicle via serial
│
└── simulation_pkg
    └── launch/                        → Launch files (mission_sim.launch.py)
     

---