Autonomous Vehicle Object Recognition With Route-Based Database Search
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for object recognition in autonomous vehicles are inefficient due to processor-intensive searches in large databases, which can lead to dangerous delays in identifying objects at high speeds.
Innovation Solution
A system that downloads a pre-loaded road object database associated with specific routes, uses a streaming search method, and if necessary, compares sensor data to an image feature database or an Internet-coupled database, allowing for rapid identification of objects by prioritizing searches within a smaller database and leveraging cloud resources for unknown objects.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a large database of feature point descriptors is used for object recognition, then the system can identify a wide variety of objects, but the processing time increases significantly
Solution Approach 1:
The patent segments the large database into multiple smaller databases organized by geographic location. Each database contains feature point descriptors for objects in a specific area, allowing the system to search only the relevant subset of data corresponding to the vehicle's current location, thereby reducing processing time while maintaining comprehensive object recognition coverage
Solution Approach 2:
The patent pre-organizes feature point descriptors into geographically-based databases before the autonomous vehicle operation. By having the data structured and ready in advance according to location, the system eliminates the need to process the entire database during real-time operation, significantly reducing processing time while maintaining versatility
2Measurement precision
If feature points are extracted from images and compared to a large database, then comprehensive object identification is achieved, but the method becomes inefficient and time-consuming
Solution Approach 1:
The patent divides the comprehensive database into location-specific segments, allowing the system to maintain high object identification accuracy by searching only the relevant geographic subset, thereby improving processing efficiency without sacrificing completeness
Solution Approach 2:
The patent applies local quality by making the database search scope dependent on the vehicle's current location. Each location has its own optimized subset of feature point descriptors, ensuring that the system searches with appropriate data for that specific area, improving both accuracy and efficiency
Data Source
AI summary
A system for object recognition in an autonomous vehicle includes non-transitory memory storage for storing instructions and an image feature database. One or more processors are in communication with the memory. The one or more processors execute the instructions to download a road object database associated with routes between a departure point and a destination. Sensor data is received from a plurality of sensors of the autonomous vehicle and compared to data in the road object database. If the sensor data is not found in the road object database, the sensor data is compared to data in the image feature database. If the sensor data is not found in the image feature database, it is transmitted over a wireless channel to an Internet-coupled database. The sensor data is interpreted based on at least one of the road object database, the image feature database, or the Internet-coupled database.