Vehicle Target Detection via Dual-Reference Correction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In the field of intelligent driving, target detection in vehicle environments faces challenges due to sheltered samples in road test data, leading to incomplete features and low accuracy in recognizing road participants and obstacles, resulting in mis-recognition or missing detection.
Innovation Solution
A method involving two pre-trained detectors, one for a reference target and another for the target to be detected, with a comparison and correction process to improve recognition accuracy, using confidence coefficients and weighted averages to verify and correct the recognition results.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a single detector is used for target detection in vehicle running, then the detection process is simple and fast, but the recognition accuracy of sheltered targets is low leading to mis-recognition or missing detection
Solution Approach 1:
The patent combines two detectors with different detection capabilities into a unified detection system. The first detector specializes in detecting reference targets (such as lane guide signs) while the second detector targets to-be-detected objects (such as road guide arrows). By merging their detection results through a correction mechanism, the system achieves higher recognition accuracy for sheltered targets than either detector could achieve alone, resolving the contradiction between accuracy and complexity.
Solution Approach 2:
The patent introduces a correction module as an intermediary between the two detectors. This mediator compares the detection results from both detectors, identifies discrepancies, and corrects the second detector's results using information from the first detector. This intermediary mechanism enables the system to achieve high recognition accuracy for sheltered targets while maintaining a relatively simple overall structure.
2Reliability
If traditional single-detector methods are used, then the system is simple to implement, but sheltered samples with large areas of occlusion result in incomplete features and poor detection performance
Solution Approach 1:
The patent merges the detection capabilities of two specialized detectors to improve reliability. The first detector provides reliable detection of reference targets that are typically visible and well-defined, while the second detector attempts to detect to-be-detected targets that may be partially occluded. By combining their results, the system achieves higher reliability for detecting sheltered targets without requiring a fundamentally complex algorithmic approach.
Solution Approach 2:
The correction module implements a feedback mechanism where the detection results from the first detector are used to verify and correct the results from the second detector. This feedback loop ensures that detection decisions are based on multiple sources of information, significantly improving detection reliability for sheltered targets while maintaining algorithmic simplicity through straightforward comparison and correction logic.
Data Source
Figure 1~2
Figure 3
Figure 4
AI summary
Disclosed are a target detection method and apparatus in running of vehicles, a device and a medium. The method includes: acquiring a picture to be detected of a target vehicle in a running process; inputting the picture to be detected into a pre-trained first detector, so that the first detector outputs a first recognition result of a reference target contained in the picture to be detected; inputting the picture to be detected into a pre-trained second detector, so that the second detector outputs a second recognition result of at least one target to be detected contained in the picture to be detected, each target to be detected and the reference target having a correspondence relationship; comparing the first recognition result with the second recognition result; and correcting the second recognition result according to a comparison result to obtain a target recognition result. The technical solutions of the present application can improve the accuracy of recognition of a sheltered target.