3D Self-Vehicle Image Generation with Pre-Computed Lighting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional image generation devices in vehicles face computational challenges in performing high-level lighting processing due to limited CPU resources, leading to delayed drawing processing.
Innovation Solution
An image generation device comprising an image acquisition unit, storage unit, bird's eye image generation unit, 3D image generation unit, and image synthesis unit, which acquires and processes images from cameras, generates a three-dimensional self-vehicle image using pre-computed self-vehicle data with quasi-lighting effects, reducing computational load while maintaining image quality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If lighting processing is performed to improve image appearance, then image quality is improved, but computational load increases beyond vehicle CPU capabilities
Solution Approach 1:
Lighting processing is performed in advance during self-vehicle data generation, not in real-time during image synthesis. The 3D model of the self-vehicle is pre-rendered with lighting effects (shadow, highlight, reflection) stored as self-vehicle data. During actual image generation, the system only needs to superimpose this pre-processed data, dramatically reducing computational load while maintaining image quality.
Solution Approach 2:
The complex lighting processing function is extracted from the real-time image synthesis process and separated into a preliminary data preparation stage. By extracting the computationally intensive lighting calculations from the time-critical rendering path, the system can use simpler, faster methods during actual image generation while still achieving realistic lighting effects through the pre-computed self-vehicle data.
2Manufacturing precision
If real-time lighting processing is performed, then image realism is improved, but drawing processing time exceeds available time
Solution Approach 1:
All time-consuming lighting calculations are performed beforehand during self-vehicle data generation. The 3D model includes pre-computed shadow maps, highlight regions, and reflection characteristics. During real-time operation, the system only performs simple superimposition of these pre-calculated lighting elements, reducing drawing processing time to well within available time limits while preserving image realism.
3Manufacturing precision
If complex lighting processing is used, then appearance quality is improved, but device complexity increases
Solution Approach 1:
The complex lighting processing algorithms are extracted from the vehicle's real-time image generation system and relocated to an offline data preparation process. The complex calculations are performed once during self-vehicle data generation, and the results are stored. The real-time system then uses simple superimposition operations, dramatically reducing processing complexity while maintaining appearance quality through the pre-computed lighting data.
Solution Approach 2:
Instead of performing complex lighting calculations in real-time, the system creates a copy of the self-vehicle 3D model with pre-applied lighting effects. This copied data (self-vehicle data including shadow and highlight information) is then reused multiple times during image synthesis without requiring recalculation, reducing both processing complexity and computational load.
Data Source
AI summary
An image generation device includes: an image acquisition unit which acquires images taken by cameras installed in a vehicle; a storage unit which stores self vehicle data indicating the vehicle having such gloss that a certain region is higher in brightness than the other region; a bird's eye image generation unit which generates a neighborhood image that is an image of the vehicle and a neighborhood of the vehicle as viewed from a virtual point of view based on the acquired images; a 3D image generation unit which generates a three-dimensional self-vehicle image being a three-dimensional image of the vehicle; and an image synthesis unit which generates a synthesized image in which the self-vehicle image is superimposed on the neighborhood image, and the 3D image generation unit generates the three-dimensional self-vehicle image based on the self vehicle data that is read out from the storage unit.