Year: 2025
Type: conference-paper
Abstract: Aiming at the problem that traditional pedestrian behavior recognition methods heavily rely on manual feature selection and extraction, which leads to difficulty in feature extraction and low recognition accuracy, a pedestrian behavior recognition model based on the combination of Convolutional Neural Network (CNN) and Support Vector machine (SVM) is proposed. The model trains data using a CNN network with a Softmax layer until convergence, enabling automatic feature extraction for pedestrian activity data. The Softmax layer is then replaced by SVM to optimize classification performance. Experimental results show that the proposed CNN-SVM model achieves a recognition accuracy of 97.77%, outperforming comparison models and demonstrating superior pedestrian activity recognition performance.
Authors: Jun Lu, Qinglei Han, Wanhang Chu, Min Kan, Yunsheng Yang
Institution: Wuhan Donghu University
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W4413812400