A vision based top-view transformation model for a vehicle parking assistant

Area of Science:

• Computer Vision
• Robotics
• Automotive Engineering

Background:

• Traditional image coordinate transformation often requires complex camera calibration, including interior and exterior orientation parameters.
• Existing systems for vehicle assistance may lack a clear, accurate representation of the surrounding environment, especially from a top-down perspective.
• The need for efficient and accessible image transformation methods in automotive applications is growing.

Purpose of the Study:

• To propose a novel Top-View Transformation Model for transforming perspective images into bird's eye view.
• To develop a method for estimating transformation parameters without requiring camera interior or exterior orientation.
• To enhance the functionality of car parking assistant systems by providing drivers with improved environmental awareness.

Main Methods:

• Development of the Top-View Transformation Model for image coordinate transformation.
• Implementation of a fitting parameters searching algorithm to estimate necessary transformation parameters.
• Optimization of processing time by utilizing a pre-calculated transformation matrix for video sequences.

Main Results:

• The proposed model successfully transforms perspective images into accurate bird's eye views.
• The system effectively estimates transformation parameters without camera calibration data.
• Experimental results demonstrate a clearer and more accurate bird's eye view for vehicle drivers, improving parking assistance.

Conclusions:

• The Top-View Transformation Model offers an efficient and parameter-free solution for perspective-to-bird's eye view transformation.
• The method significantly enhances the visual feedback provided by car parking assistant systems.
• The approach is suitable for real-time applications on embedded platforms due to optimized processing.