Related Experiment Video
Updated: Jun 27, 2026
03:31
End-To-End Deep Neural Network for Salient Object Detection in Complex Environments
Published on: December 15, 2023
Improved CenterNet-Based Multimodal Object Detection for Low-Light and Complex Environments
Zhigang Yao1, Hengxin Xu1, Huazhong Zhang1,2
1College of Aviation and Electronics and Electrical, Civil Aviation Flight University of China, Guanghan 618307, China.
Sensors (Basel, Switzerland)
|June 26, 2026
Summary
This study introduces an improved CenterNet object detection method using fused and infrared images. The novel approach enhances multimodal fusion and localization accuracy for low-light conditions, achieving a 3.51% mAP@0.5 improvement.
Area of Science:
- Computer Vision
- Machine Learning
- Artificial Intelligence
Background:
- Object detection in low-light and complex backgrounds faces challenges with detail representation, cross-modal fusion, and localization accuracy.
- Existing methods struggle to effectively integrate information from multiple image modalities under adverse conditions.
Purpose of the Study:
- To propose an improved CenterNet-based multimodal object detection method for enhanced performance in low-light and complex environments.
- To address limitations in detail representation, cross-modal fusion, and localization accuracy.
Main Methods:
- A dual-source input using fused and infrared images with infrared wavelet priors for enhanced texture and structure.
- A Feature Fusion Attention (FFA) module for improved cross-modal feature interaction.
- A Heatmap-Guided Detection Head (HGDH) for explicit enhancement of target regions and a two-stack Hourglass backbone for multi-scale feature extraction.
Main Results:
- The proposed method achieved a 3.51% improvement in mean Average Precision at 0.5 IoU (mAP@0.5) on the constructed RH-25 dataset compared to the baseline.
- Ablation and comparative experiments validated the effectiveness of the proposed modules.
- Supplementary experiments on the MFAD dataset demonstrated cross-dataset adaptability.
Conclusions:
- The developed multimodal object detection method significantly improves detection performance in low-light and complex environments.
- The integration of infrared wavelet priors, FFA module, and HGDH contributes to better feature representation and localization.
- The method shows promise for real-world applications requiring robust object detection under challenging visual conditions.
Related Concept Videos
02:16Light Acquisition
In order to produce glucose, plants need to capture sufficient light energy. Many modern plants have evolved leaves specialized for light acquisition. Leaves can be only millimeters in width or tens of meters wide, depending on the environment. Due to competition for sunlight, evolution has driven the evolution of increasingly larger leaves and taller plants, to avoid shading by their neighbors with contaminant elaboration of root architecture and mechanisms to transport water and nutrients.
01:15Depth Perception and Spatial Vision
Depth perception is the ability to perceive objects three-dimensionally. It relies on two types of cues: binocular and monocular. Binocular cues depend on the combination of images from both eyes and how the eyes work together. Since the eyes are in slightly different positions, each eye captures a slightly different image. This disparity between images, known as binocular disparity, helps the brain interpret depth. When the brain compares these images, it determines the distance to an object.