Vehicle Location Estimation Using Weighted Specific-Point Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing location determination and map construction methods using GPS/DR technology face accuracy limitations due to error accumulation and require costly RTK corrections, while image recognition methods need reference points and incur data costs, especially in environments with performance limitations.
Innovation Solution
A vehicle apparatus equipped with a camera and processor that recognizes specific points via image analysis, assigns weights to recognized information, and estimates location using vehicle sensors, communication signals, and image recognition models to enhance accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If GPS/DR technology is used for location determination and map construction, then satellite data and sensor data can be utilized, but accuracy decreases due to error accumulation and limitations in GPS/DR
Solution Approach 1:
The patent combines multiple location determination technologies (GPS, DR, image recognition) into a hybrid system that leverages the strengths of each method while compensating for their individual weaknesses, thereby improving overall location accuracy and reliability
Solution Approach 2:
The patent introduces image recognition as an intermediary method to validate and correct location data from GPS and DR systems, particularly in environments where GPS signals are weak or unavailable, thus reducing error accumulation
2Measurement precision
If RTK correction is performed to improve GPS accuracy, then location precision reaches several tens of centimeters, but cost burden increases
Solution Approach 1:
The patent uses standard GPS receivers without expensive RTK correction hardware, accepting lower GPS accuracy in exchange for significantly reduced system cost, while compensating through alternative methods like image recognition and DR calibration
3Measurement precision
If image recognition is used to estimate location, then reference points are required, but this creates limitations in estimating location and constructing maps
Solution Approach 1:
The patent makes the image recognition system universal by enabling it to work with or without pre-defined reference points, allowing the same system to function in both supervised (with reference points) and unsupervised (without reference points) modes
Solution Approach 2:
The patent performs preliminary map construction and reference point identification in advance using image recognition, creating a database that can then be used for faster and more accurate location estimation during actual vehicle operation
4Measurement precision
If image recognition is used in environments with performance limitations, then recognition performance deteriorates, but data costs may continuously incur if information is sent to a server
Solution Approach 1:
The patent segments the image recognition process into two parts: simple, computationally lightweight recognition tasks are performed locally on the vehicle terminal, while only necessary or complex recognition results are transmitted to the server, reducing data costs and improving response time
Data Source
AI summary
A vehicle apparatus includes a camera and a processor, and is configured to estimate a location of a vehicle and construct a map via image recognition. The processor obtains an image by using the camera if the vehicle enters a specific point, recognizes information related to the specific point in the image, assigns a weight to the recognized information according to a predetermined criterion, and estimates the location of the vehicle based on the recognized information to which the weight is applied.