Human Detection Using Random Color Similarity Feature and Random Ferns Classifier.

Area of Science:

• Computer Vision
• Machine Learning

Background:

• Color-based features are underutilized in vision-based human detection due to significant intra-class variations.
• Existing methods often rely on features less sensitive to color variations.

Purpose of the Study:

• To propose a novel color-based feature for human detection.
• To evaluate the efficacy of random ferns classifier for this task.
• To improve human detection accuracy and efficiency.

Main Methods:

• Development of a novel random color similarity (RCS) feature.
• Integration of histogram of oriented gradient based local binary feature (HOG-LBF).
• Utilizing a random ferns classifier for its speed and performance.

Main Results:

• The proposed RCS feature effectively characterizes human appearances.
• The random ferns classifier demonstrated faster training and testing than SVM without performance loss.
• The combined approach achieved competitive human detection results on public datasets.

Conclusions:

• The novel RCS feature offers a viable solution for color-based human detection.
• Random ferns classifier is an efficient and effective choice for human detection tasks.
• The proposed method shows promise for real-world vision-based human detection applications.