研究目的
To overcome the difficulties of segmenting human faces in thermal infrared images due to low contrast and limited information by proposing an improved circular shortest path method.
研究成果
The proposed improved circular shortest path method effectively segments thermal infrared faces by enhancing gradient information, applying shape constraints, and using penalty terms to handle local minima and noise, outperforming existing methods in both visual and quantitative evaluations.
研究不足
The method may be affected by very sparse gradient information and local minima regions in thermal images, and the computational time increases with higher acquisition scales.
1：Experimental Design and Method Selection:
The study uses an improved circular shortest path (CSP) algorithm for thermal face segmentation, incorporating gradient-based cost functions, shape constraints, and penalty terms to enhance accuracy and robustness.
2：Sample Selection and Data Sources:
A dataset of 2000 thermal infrared images from 50 people with diverse features, hairstyles, facial expressions, and head postures, captured with an infrared camera (wave spectrum
3：5-14μm) at distances of 1-5 meters. List of Experimental Equipment and Materials:
Infrared camera with specified wave spectrum, computer for image processing.
4：Experimental Procedures and Operational Workflow:
Images are transformed from polar to Cartesian coordinates, cost functions are applied (including gradient enhancement, shape constraints, and penalty terms), and the shortest path is computed using dynamic programming to extract face contours.
5：Data Analysis Methods:
Quantitative evaluation using Miss-classification Error (ME), Relative foreground Area Error (RAE), and False Alarm Rate (FAR); visual comparison with other segmentation methods.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
