Vehicle Resume Device Using Low-Level Feature Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current adaptive cruise control (ACC) systems have a high false negative rate, missing relevant objects and a high false positive rate, which can lead to unsafe automatic resumption of vehicle movement, especially when objects like small children or pedestrians are partially occluded or close to the vehicle.
Innovation Solution
A device and method that utilize low-level features, which cannot be characterized in a single image, to determine the presence of objects in the scene, using artificial intelligence and machine learning algorithms to analyze these features and track potential objects across multiple images, thereby reducing false negatives and positives by considering information previously not usable in ADAS systems.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If current ACC systems use traditional object detection methods, then the system is simple to operate, but the false negative rate is high and relevant objects are missed
Solution Approach 1:
The patent segments object detection into multiple levels: traditional high-level object recognition and new low-level feature analysis. The system processes images through multiple stages, extracting both traditional object characteristics and additional low-level features such as edge patterns, texture information, and spatial relationships. This segmentation allows the system to maintain traditional simple detection while adding more sophisticated analysis layers that reduce false negatives without completely redesigning the entire system.
Solution Approach 2:
The patent adds another dimension to object detection by incorporating low-level feature analysis alongside traditional high-level object recognition. Instead of relying solely on object classification, the system analyzes additional dimensional information including edge orientations, gradient patterns, and spatial relationships between features. This multi-dimensional approach enables the system to detect objects that traditional single-dimension methods miss, thereby reducing false negatives while managing system complexity through structured processing.
2Reliability
If current ACC systems use traditional object detection methods, then the device complexity is low, but the false positive rate is high leading to unsafe automatic resumption
Solution Approach 1:
The patent implements feedback mechanisms where the system continuously monitors low-level features and adjusts its confidence in object detection accordingly. When low-level features indicate potential objects that traditional methods miss, the system generates feedback signals to trigger further verification. This feedback loop allows the system to reduce false positives by cross-validating detections through multiple analysis layers, thereby improving safety without requiring complete system redesign.
Solution Approach 2:
The patent performs preliminary analysis of low-level features before making final object detection decisions. By pre-processing images to extract edge patterns, gradients, and spatial relationships, the system prepares additional verification information in advance. This preliminary action allows the system to cross-check traditional object detections against low-level feature evidence, reducing false positives that could lead to unsafe automatic resumption while managing complexity through staged processing.
3Reliability
If the system only uses high-level object features, then the processing is fast and simple, but partially occluded or close objects cannot be detected
Solution Approach 1:
The patent segments the feature extraction process into high-level object recognition and low-level feature analysis. For occluded or close objects where traditional high-level features may be insufficient, the system separately processes low-level features such as visible edge patterns, texture information, and spatial relationships. This segmentation allows the system to compensate for missing high-level information by relying on persistent low-level features that remain detectable even when objects are partially occluded or very close to the vehicle.
Solution Approach 2:
The patent adds low-level feature analysis as another dimension to object detection, enabling the system to detect objects that traditional high-level feature methods miss. By incorporating edge orientations, gradient patterns, and spatial relationships as additional dimensional information, the system can identify partially occluded or close objects through their geometric and textural characteristics even when complete object recognition is not possible. This multi-dimensional approach improves detection reliability without requiring a complete system overhaul.
Data Source
Figure 1~2
Figure 3a~4
Figure 5
AI summary
The present invention relates to a device for enabling a vehicle to automatically resume moving. It is described to provide a processing unit (30) with at least one image, the at least one image relating to a scene external to a vehicle. The processing unit processes the at least one image to extract low level features, a low level feature being defined such that an object in the scene cannot be characterised on the basis of at least one low level feature in a single image. The processing unit determines if at least a part of an object is present in a part of the scene, the determination comprising an analysis of the low level features. An output unit (40) outputs information that "there is something present" in the scene.