Technical gazette, Vol. 32 No. 4, 2025.
Original scientific paper
https://doi.org/10.17559/TV-20241219002204
Research on Human Body Detection Algorithm in Car Cabin Based on RGB-IR Dual Light Fusion
Siyu Chen
; School of Advanced Manufacturing, Nanchang University, Nanchang 330031, China
Juhua Huang
; School of Advanced Manufacturing, Nanchang University, Nanchang 330031, China
Yinyin Liu
; School of Economics and Management, Nanchang University, Nanchang, 330031, China
*
Fengping Xu
; Yicheng Automotive Technology (Shanghai) Co., Shanghai, 201800, China
* Corresponding author.
Abstract
A strategy of RGB-IR dual optical fusion is proposed, that is, each component of multiple color Spaces is detected individually, several channels with the best human detection performance are counted, and the prospects of these channels are fused to obtain the final human detection result. In addition, to address the challenge of insufficient contour prediction accuracy in the semantic segmentation task of RGB-IR dual-modal images, we propose an innovative multi-scale contour enhancement dual-modal semantic segmentation method, and introduce a novel location and channel attention mechanism module, which can effectively promote cross-scale feature fusion. Thus, the contour prediction ability of various scales can be accurately improved. Through the detection of human body in car cabin based on infrared and visible light mode fusion, in order to verify the effectiveness of the proposed modal fusion detection algorithm for the actual traffic scene, the practical application of the model is realized through the fusion detection algorithm. The experimental results show that the detection results of this algorithm are better than those of human body based on single color space in complex scenes, and the algorithm can effectively deal with dynamic background problems.
Keywords
dual light fusion; human body detection in the car compartment; mode fusion; semantic segmentation; RGB-IR
Hrčak ID:
332863
URI
Publication date:
29.6.2025.
Visits: 650 *