Multi-Camera Image Merging for Faster Autonomous Driving Recognition

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing autonomous driving technologies face challenges with time delays, synchronization issues, and high processing loads when processing multiple images from various sensors, and there are difficulties in accurately identifying and merging objects across different camera views.

Innovation Solution

A data construction and learning system that rearranges and merges images from multiple cameras, using a single deep learning network to recognize objects and road situations, eliminating the need for radar or LiDAR, and enabling high-speed, accurate recognition across varying field of views.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If multiple images from various sensors are processed independently in parallel or sequentially, then comprehensive object recognition is achieved, but time delay, synchronization issues, and high processing load occur

Engineering Contradiction:
Improveobject recognition accuracyVSAvoidprocessing time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The patent merges multiple images from different cameras into a single composite image that contains all relevant visual information. This consolidation allows the system to process one unified image rather than multiple separate images, significantly reducing processing time and eliminating synchronization issues while maintaining comprehensive object recognition capability across all camera views

Inventive Principle:
Principle #5Merging (Combining)

2Measurement precision

If multiple images from various sensors are processed independently in parallel or sequentially, then comprehensive object recognition is achieved, but high processing load occurs

Engineering Contradiction:
Improveobject recognition accuracyVSAvoidprocessing load
Core Design Contradiction:
Measurement precisionVSPower

Solution Approach 1:

By combining multiple camera images into a single composite image, the system reduces the total computational load. Instead of running separate processing pipelines for each camera, the merged image approach requires only one processing pass, significantly reducing power consumption and computational resources while maintaining the ability to recognize objects across all original camera fields of view

Inventive Principle:
Principle #5Merging (Combining)

3Measurement precision

If individual image recognition and merging is performed, then object identification is attempted, but difficulty in identifying and merging the same object occurs

Engineering Contradiction:
Improveobject identification accuracyVSAvoidmerging complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

Instead of the conventional approach of recognizing objects in each image separately and then merging the results, this patent inverts the process by first merging the images and then performing single-pass object recognition on the combined image. This eliminates the complexity of matching and merging objects across multiple images, as object identification is performed on the unified merged image where all objects are already in their correct relative positions

Inventive Principle:
Principle #13The other way round (Inversion)

Data Source

PatentUS12482268B2Data construction and learning system and method based on method of splitting and arranging multiple images
Publication Date: 2025.11.25 ELECTRONICS & TELECOMM RES INST
  • US12482268B2 patent drawing
  • US12482268B2 patent drawing
  • US12482268B2 patent drawing

AI summary

The present disclosure relates to a data construction and learning system and method based on a method of splitting and arranging multiple images. The data construction and learning system based on a method of splitting and arranging multiple images includes an input unit configured to receive images captured by a plurality of cameras disposed in a vehicle, a memory in which a program for merging the images into a single image and estimating information on a road situation and an object has been stored, and a processor configured to execute the program. The processor merges and recognizes, as one situation, road situations and objects redundantly included in the images.