Autonomous Vehicle Depth Validation Using Photometric Loss
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current autonomous vehicles rely on artificial neural networks for depth estimation, which lack reliable certification methods for depth information, posing a safety risk.
Innovation Solution
A method and system using a photometric loss function to determine the quality of depth information in real-time, enabling reliable validation of depth data without human input, and allowing for adaptive vehicle control based on depth information quality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If artificial neural networks are used for depth estimation in autonomous vehicles, then depth information can be obtained from sensor data, but reliable certification of the depth information becomes difficult
Solution Approach 1:
The patent implements a feedback mechanism where the quality of depth information is continuously assessed using a photometric loss function. The system projects the initial image using depth information to create a first projected image, compares it with the actual first image captured by the sensor, and uses the photometric loss function to determine quality metrics. This feedback loop enables reliable certification of depth information by providing quantitative quality measures based on image consistency.
Solution Approach 2:
The patent introduces an intermediary quality assessment mechanism that mediates between the neural network's depth estimation and the vehicle's monitoring system. The photometric loss function acts as an intermediary that evaluates the reliability of depth information by comparing projected images with actual sensor images, providing a bridge that enables certification without requiring direct human intervention or complex additional sensors.
2Reliability
If depth information quality assessment is performed using photometric loss function with image projection, then reliable and deterministic quality determination is achieved, but computational complexity increases
Solution Approach 1:
The patent applies preliminary action by capturing an initial image and determining depth information before the vehicle moves significantly. This initial depth information is then used to project the initial image to a first projected image at a later time point. By performing this projection and quality assessment in advance and in a predetermined manner, the system achieves reliable quality determination while managing computational complexity through efficient use of pre-captured data.
Solution Approach 2:
The patent uses copying by creating a first projected image from the initial image using the determined depth information. This projected image is a synthetic copy that represents what the initial image would look like at the first time point based on the depth data. By comparing this copied projected image with the actual first image captured by the sensor, the system can assess depth quality without requiring complex real-time processing during vehicle operation.
3Adaptability or versatility
If autonomous vehicles use neural network-based depth estimation, then distance measurement capability is provided, but operational safety is compromised due to lack of validation
Solution Approach 1:
The patent implements a feedback-based validation system where the quality of depth information is continuously monitored using photometric loss functions. The system projects initial images using depth data, compares projected images with actual sensor images, and uses the loss function values to determine quality metrics. This feedback mechanism provides operational safety by enabling the vehicle to validate depth information reliability and adjust operations accordingly.
Solution Approach 2:
The patent enables the autonomous vehicle to perform self-validation of its depth information without requiring external human intervention or additional complex validation systems. The vehicle uses its own captured images and the photometric loss function to automatically assess the quality and reliability of its neural network's depth estimation, providing self-service safety validation that maintains operational versatility while ensuring reliability.
Data Source
Figure 1
Figure 2
AI summary
A method for monitoring an autonomously moving vehicle (10) comprises the following steps, which are performed during movement of the autonomously moving vehicle (10): capturing an initial image at an initial time and a first image at a first time using at least one sensor (13), determining depth information for the initial image using an artificial neural network, determining a first projected image for the first time from the initial image and the depth information, determining the quality of the depth information using a photometric loss function based on the first image and the first projected image, and monitoring the autonomously moving vehicle (10) taking into account the quality of the depth information. Furthermore, a system for monitoring an autonomously moving vehicle (10) is provided.