Perception Data Screening for Negative Obstacle Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current perception systems for autonomous vehicles lack comprehensive detection methods for negative samples, leading to incomplete evaluation and potential misrecognition of obstacles like water mist or snow, which can affect driving safety.
Innovation Solution
A perception data detection method that includes acquiring labeled and perception data, performing rough filtering based on negative samples to identify matching obstacles, and determining a detection result using the number of target obstacles to ensure comprehensive detection of negative samples.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If detection is performed only based on positive sample data matching, then the detection process is simple, but the detection comprehensiveness deteriorates because negative sample recognition is not reflected
Solution Approach 1:
The detection process is segmented into two independent parts: positive sample detection and negative sample detection. Positive samples are detected through traditional matching methods, while negative samples are detected through a separate process involving rough filtering based on negative sample data and precise filtering through matching with the first obstacle data. This segmentation allows each type of sample to be processed with appropriate methods, improving overall detection comprehensiveness without excessive complexity.
Solution Approach 2:
Rough filtering is performed as a preliminary action before precise matching. The rough filtering step uses negative sample data to pre-screen and eliminate obviously unrelated perception data, reducing the search space for subsequent precise matching. This preliminary action improves detection efficiency and comprehensiveness by ensuring that potential negative samples are not missed while maintaining a manageable processing complexity.
2Ease of operation
If traditional matching based on positive samples is used, then the detection method is straightforward, but it cannot reflect the recognition of negative samples leading to incomplete detection
Solution Approach 1:
The detection methodology is segmented into distinct pathways for positive and negative samples. Positive samples continue to use traditional matching methods, while negative samples are processed through a dedicated pipeline that includes rough filtering using negative sample data and precise filtering through matching with first obstacle data. This ensures negative sample information is captured without complicating the overall detection framework.
Solution Approach 2:
The second obstacle data acts as an intermediary between the raw perception data and the final detection result for negative samples. The rough filtering process uses negative sample data to identify potential matches, and then precise filtering uses matching with first obstacle data to confirm them. This intermediary process ensures negative sample information is properly captured and processed.
3Device complexity
If only positive samples are considered in detection, then the evaluation is simpler, but the accuracy of obstacle recognition deteriorates due to misrecognition of negative samples like water mist or snow
Solution Approach 1:
The evaluation process is segmented to separately assess positive and negative sample recognition. Positive samples are evaluated through traditional matching accuracy metrics, while negative samples are evaluated through a separate process that compares rough filtering results with precise filtering results. This segmentation allows accurate evaluation of both sample types without requiring a single complex evaluation framework.
Solution Approach 2:
Rough filtering serves as a preliminary evaluation step that identifies potential negative samples before precise filtering confirms them. This two-stage approach allows the system to efficiently screen for negative samples and then accurately verify them, improving overall obstacle recognition accuracy while maintaining manageable evaluation complexity.
Data Source
AI summary
The present disclosure discloses a perception data detection method and apparatus. The specific implementation scheme is: acquiring labeled data and perception data, where the labeled data includes a labeled position and a labeled type of at least one first obstacle, and the perception data includes a perception position and a perception type of at least one second obstacle; performing, according to a negative sample in the labeled data, rough filtering on the second obstacle in the perception data to obtain a third obstacle remaining after the rough filtering, where the negative sample is a non-physical obstacle; determining a matching relationship between the third obstacle and the negative sample, and determining the negative sample having the matching relationship with the third obstacle as a target obstacle; and determining, according to the number of the target obstacle and the number of the first obstacle, a detection result of the perception data.