Bright Channel Prior Attention for Multispectral Pedestrian Detection

Chenhang Cui,Jinyu Xie,Yechenhao Yang

2023-05-22

Abstract:Multispectral methods have gained considerable attention due to their promising performance across various fields. However, most existing methods cannot effectively utilize information from two modalities while optimizing time efficiency. These methods often prioritize accuracy or time efficiency, leaving room for improvement in their performance. To this end, we propose a new method bright channel prior attention for enhancing pedestrian detection in low-light conditions by integrating image enhancement and detection within a unified framework. The method uses the V-channel of the HSV image of the thermal image as an attention map to trigger the unsupervised auto-encoder for visible light images, which gradually emphasizes pedestrian features across layers. Moreover, we utilize unsupervised bright channel prior algorithms to address light compensation in low light images. The proposed method includes a self-attention enhancement module and a detection module, which work together to improve object detection. An initial illumination map is estimated using the BCP, guiding the learning of the self-attention map from the enhancement network to obtain more informative representation focused on pedestrians. The extensive experiments show effectiveness of the proposed method is demonstrated through.

Computer Vision and Pattern Recognition,Machine Learning

What problem does this paper attempt to address?

This paper attempts to address the issue of multispectral pedestrian detection under low-light conditions. Specifically, although existing multispectral methods perform well in various fields, they still have shortcomings in optimizing time efficiency while effectively utilizing information from both modalities. These methods often prioritize either accuracy or time efficiency, resulting in limited performance improvement.

To tackle these issues, the authors propose a new method called "Bright Channel Prior Attention," which integrates image enhancement and detection within a unified framework to improve pedestrian detection performance under low-light conditions. The main contributions include:

1. **Bright Channel Prior Attention**: Using the V channel (brightness) in the HSV color space of thermal images as an attention map, it triggers an unsupervised autoencoder for visible light images, gradually emphasizing pedestrian features.
2. **Advanced Network and Unsupervised Bright Channel Prior Algorithm**: Used for light compensation in low-light images.
3. **Unified Enhancement and Detection Framework**: Combines image enhancement and object detection within one framework, allowing the enhancement effects to benefit the detection task.

Through these innovations, this method aims to improve the accuracy and robustness of pedestrian detection under low-light conditions, thereby providing better technical support for applications such as autonomous driving and surveillance.