Machine Learning Device for Accurate Distance Image Generation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current machine learning devices face challenges in generating highly accurate distance images for vehicle external environment recognition, particularly due to issues like mirror reflections and inconsistencies between image sensors and Lidar devices, which affect the accuracy of distance value detection and processing.
Innovation Solution
A machine learning device comprising a road surface detection processor, a distance value selector, and a learning processor that detects road surfaces and selects relevant distance values from a distance image, using these values to generate a learning model for improved accuracy in distance image generation, thereby reducing noise and enhancing the accuracy of distance value processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If all distance values from the distance image are used for learning processing, then the learning model can be trained with comprehensive data, but the accuracy is reduced due to inclusion of erroneous distance values from mirror reflections and sensor inconsistencies
Solution Approach 1:
The patent extracts and removes erroneous distance values from the distance image before using them for learning processing. The road surface detection processor identifies distance values corresponding to road surfaces, and the distance value selector processor selectively extracts only those distance values that are likely to be accurate, excluding values affected by mirror reflections or sensor inconsistencies. This extraction principle resolves the contradiction by filtering out harmful data while retaining sufficient training data for the learning model.
Solution Approach 2:
The patent introduces an intermediary processing stage between the distance image generation and the learning model training. The road surface detection processor and distance value selector processor act as intermediaries that filter and select appropriate distance values. This intermediary layer resolves the contradiction by mediating between the raw distance data (which contains errors) and the learning model (which requires accurate data), allowing comprehensive data usage while maintaining high accuracy.
2Reliability
If distance values from regions with mirror reflections or sensor inconsistencies are included in learning processing, then more training data is available, but the reliability of the learning model decreases due to erroneous distance values
Solution Approach 1:
The patent applies preliminary action by performing road surface detection and distance value selection before the learning processing stage. The road surface detection processor identifies road surface regions in advance, and the distance value selector processor pre-selects reliable distance values from these regions before they are used for training the learning model. This preliminary filtering resolves the contradiction by preventing erroneous data from entering the learning process, thereby maintaining reliability without requiring complex post-processing or validation mechanisms.
3Measurement precision
If selective processing of distance values is performed to improve accuracy, then the learning model quality is enhanced, but the processing complexity increases due to additional detection and selection steps
Solution Approach 1:
The patent merges the road surface detection function and the distance value selection function into an integrated processing pipeline. The road surface detection processor and distance value selector processor work together as a combined system that simultaneously performs detection and selective extraction. This merging resolves the contradiction by consolidating multiple functions into a unified processing architecture, reducing overall system complexity while maintaining the ability to selectively process accurate distance values for enhanced learning model quality.
Data Source
AI summary
A machine learning device according to an embodiment of the disclosure includes: a road surface detection processor configured to detect, on the basis of a first captured image and a first distance image depending on the first captured image, a road surface included in the first captured image; a distance value selector configured to select one or more distance values to be processed, from among distance values included in the first distance image, on the basis of a processing result of the road surface detection processor; and a learning processor configured to generate a learning model to be supplied with a second captured image and to output a second distance image depending on the second captured image, by carrying out machine learning processing on the basis of the first captured image and the one or more distance values.