Road Dividing Object Detection via Image Feature Sequences
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional road dividing object detection methods struggle to accurately detect road dividing objects in complex road conditions, particularly when multiple or discontinuous objects are present, as they are typically designed to detect specific features rather than generalizing across various types of road dividing objects.
Innovation Solution
A method and device that convert a three-dimensional road scene into a time series-based two-dimensional image feature sequence, allowing for the extraction of sequence characteristics to detect both continuous and discontinuous road dividing objects without restricting the detection to specific objects, enabling detection in unknown or complicated road conditions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional detection methods focus on single specific features (fence, wall, building), then detection accuracy for that specific object type is improved, but the system cannot detect discontinuous or varied road dividing objects in complex conditions
Solution Approach 1:
The patent applies universality by creating a detection system that can identify multiple types of road dividing objects (fences, walls, buildings, pillars, discontinuous objects) using a unified approach. The system extracts feature quantities from sequences of images and detects periodic changes that characterize all road dividing objects regardless of their specific type, making the detection method versatile and adaptable to various road conditions without requiring separate detection algorithms for each object type.
2Device complexity
If the detection system is designed for specific road dividing objects, then the device complexity is reduced, but it fails to detect road dividing objects in complicated road conditions
Solution Approach 1:
The patent applies parameter changes by shifting from detecting static, object-specific features to detecting dynamic temporal patterns in feature quantities extracted from image sequences. The system monitors how feature quantities change over time as the vehicle moves, identifying periodic variations that indicate the presence of road dividing objects. This parameter transformation allows a relatively simple detection structure to reliably identify diverse objects by focusing on their temporal behavior rather than their specific visual characteristics.
3Speed
If conventional methods use static image analysis, then processing speed is improved, but sequence characteristics of road dividing objects cannot be extracted
Solution Approach 1:
The patent applies periodic action by exploiting the periodic nature of feature quantity changes as the vehicle passes road dividing objects. The system extracts feature quantities from sequences of images captured at regular intervals and analyzes periodic variations in these features. This approach enables the system to capture temporal sequence characteristics while maintaining efficient processing by focusing on periodic patterns rather than analyzing every pixel in every image, thus preserving important temporal information without excessive computational overhead.
Data Source
Figure 1
Figure 2
Figure 3A~4C
AI summary
Disclosed is a method for detecting a road dividing object in a target scene. The method includes steps of obtaining an input image of the target scene; selecting an image unit from the input image; extracting an image feature within the image unit in each of continuous input images of the target scene so as to generate an image feature sequence; and detecting the road dividing object based on the image feature sequence.