Saliency-Based Road Target Extraction in Night Vision Infrared Images
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing visual attention models for night road scenes have low accuracy in extracting salient targets due to influence from non-salient regions and background lights, and fail to clearly outline prominent targets in night vision infrared images.
Innovation Solution
A road target extraction method using the Graph-Based Visual Saliency (GBVS) model combined with spectral scale space analysis in the hypercomplex frequency domain, followed by fusion of global and local cues to enhance feature extraction and improve saliency image clarity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional object detection algorithms are used in complex road scenes, then comprehensive feature analysis is achieved, but data processing capacity requirements increase and processing efficiency decreases
Solution Approach 1:
The patent segments the complex detection task into two stages: first using a visual saliency model to segment and extract potential target regions from the entire image, then applying traditional detection algorithms only to these segmented regions. This segmentation approach maintains detection accuracy while significantly reducing the data processing burden by limiting subsequent analysis to only the most promising regions identified by the saliency model.
Solution Approach 2:
The visual saliency model performs preliminary action by pre-processing the image to identify and highlight potential target regions before the main detection algorithm is applied. This preliminary filtering step prepares the data by organizing it into salient regions, which then allows the subsequent detection algorithm to work more efficiently on pre-organized data rather than processing the entire image from scratch.
2Measurement precision
If existing visual attention models are used for night road scenes, then processing speed is maintained, but detection accuracy decreases due to influence from non-salient regions and background lights
Solution Approach 1:
The patent applies the taking out principle by extracting and isolating the salient target regions from the complex night road scene using the visual saliency model. This extraction process separates the important target information from the harmful background elements and non-salient regions, allowing the detection algorithm to focus solely on the extracted salient regions and ignore the distracting background lights and irrelevant areas.
Solution Approach 2:
The patent implements local quality by applying different processing strategies to different regions of the image. The visual saliency model identifies regions with different salient qualities, and the detection algorithm is then applied selectively to high-salience regions with appropriate local adjustments. This allows the system to adapt to local characteristics of different road scenes, improving accuracy in specific regions without being affected by global background conditions.
3Measurement precision
If low-level feature fusion is used for saliency extraction, then processing simplicity is maintained, but extraction accuracy decreases by ignoring high-level features
Solution Approach 1:
The patent merges multiple feature levels by combining low-level visual features (color, intensity, texture) processed through the visual saliency model with high-level semantic information from the detection algorithm applied to salient regions. This merging of features at different levels of abstraction achieves more accurate saliency extraction and target detection while managing complexity through the staged processing approach, where each stage contributes complementary information.
Data Source
AI summary
The present invention belongs to the field of machine vision. A saliency-based method for extracting a road target from a night vision infrared image is disclosed. The method combines saliencies of the time domain and the frequency domain, global contrast and local contrast, and low level features and high level features, and energy radiation is also considered to be a saliency factor; thus, the object of processing is an infrared image and not the usual natural image. The extraction of a salient region is performed on the raw natural image on the basis of energy radiation, and the obtained extraction result if the salient region is more accurate and thorough, and the contour of a target in the salient region is clearer.