Static Object Region Detection Accuracy in Autonomous Vehicle Sensors
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face challenges in recognizing objects due to varying driving environments and unique image characteristics from different sensors, leading to potential failures in object detection.
Innovation Solution
A method involving an electronic device that obtains images from sensors, detects static object regions using an object detection model, determines the accuracy of these regions, and decides whether to include the images in a training dataset. The method also determines a ground truth static object region based on space-occupancy information, allowing for adaptive learning and improved object detection performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If an object detection model is applied to detect static object regions in images captured by sensors with unique characteristics, then the detection process can be automated, but the accuracy and reliability of object detection deteriorate due to varying driving environments and sensor characteristics
Solution Approach 1:
The patent changes the parameters of the object detection model by collecting training images specific to each driving environment and sensor characteristics, then retraining the model to adapt to these parameter variations. This resolves the contradiction by maintaining automation while improving reliability through environment-specific parameter optimization.
Solution Approach 2:
The patent implements a dynamic approach where the object detection model is continuously adapted to different driving environments through automated collection of training images and retraining processes. This dynamic adaptation allows the system to maintain high reliability across varying conditions while preserving automation.
2Reliability
If the object detection model is re-trained in the current driving environment to stabilize the recognizer, then the reliability of object detection is improved, but the time and resources required for training increase
Solution Approach 1:
The patent performs preliminary actions by automatically collecting training images from the current driving environment before retraining is needed. This preparation reduces the overall training time and resources required when actual retraining occurs, as the data collection phase is already completed.
Solution Approach 2:
The system performs self-service by automatically collecting training images and identifying when retraining is necessary without requiring external intervention. This automation of the training process reduces the time and resource investment needed while maintaining recognizer stability.
3Adaptability or versatility
If images are collected as part of a training dataset from diverse driving environments, then the adaptability of the object detection model is improved, but the quantity of data to be processed increases
Solution Approach 1:
The patent applies local quality by collecting training images specific to each driving environment and sensor type, rather than using a generic large dataset. This approach improves adaptability to local conditions while avoiding the need to process excessively large quantities of data from all possible environments.
Solution Approach 2:
The patent segments the training data collection process by environment and sensor type, processing and training on smaller, targeted datasets for each segment rather than processing one large heterogeneous dataset. This reduces the overall data processing burden while maintaining adaptability.
Data Source
AI summary
A method performed by an electronic device includes: obtaining an image captured by a sensor; detecting, in the image, a static object region corresponding to a static object of the image, wherein the detecting is performed by applying an object detection model to the image; determining whether to collect the image as part of a training dataset based on an accuracy level of the detected static object region; and determining a ground truth static object region for the static object of the image from space-occupancy information of the static object with respect to the image collected as part of the training dataset.