Pedestrian detection in infrared image using HOG and Autoencoder

Tianbiao Chen; Hao Zhang; Wenjie Shi; Yu Zhang

doi:10.1117/12.2295804

15 November 2017 Pedestrian detection in infrared image using HOG and Autoencoder

Tianbiao Chen, Hao Zhang, Wenjie Shi, Yu Zhang

Proceedings Volume 10605, LIDAR Imaging Detection and Target Recognition 2017; 106053Q (2017) https://doi.org/10.1117/12.2295804
Event: LIDAR Imaging Detection and Target Recognition 2017, 2017, Changchun, China

Abstract

In order to guarantee the safety of driving at night, vehicle-mounted night vision system was used to detect pedestrian in front of cars and send alarm to prevent the potential dangerous. To decrease the false positive rate (FPR) and increase the true positive rate (TPR), a pedestrian detection method based on HOG and Autoencoder (HOG+Autoencoder) was presented. Firstly, the HOG features of input images were computed and encoded by Autoencoder. Then the encoded features were classified by Softmax. In the process of training, Autoencoder was trained unsupervised. Softmax was trained with supervision. Autoencoder and Softmax were stacked into a model and fine-tuned by labeled images. Experiment was conducted to compare the detection performance between HOG and HOG+Autoencoder, using images collected by vehicle-mounted infrared camera. There were 80000 images for training set and 20000 for the testing set, with a rate of 1:3 between positive and negative images. The result shows that when TPR is 95%, FPR of HOG+Autoencoder is 0.4%, while the FPR of HOG is 5% with the same TPR.

Citation Download Citation

Tianbiao Chen, Hao Zhang, Wenjie Shi, and Yu Zhang "Pedestrian detection in infrared image using HOG and Autoencoder", Proc. SPIE 10605, LIDAR Imaging Detection and Target Recognition 2017, 106053Q (15 November 2017); https://doi.org/10.1117/12.2295804

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available