山区高速公路隧道出入口视觉融合技术研究北大核心CSCD

针对红外非可见光与可见光视觉在成像过程中不同的感光特性,面向隧道典型的“黑洞”和“白洞”问题,从自动驾驶车辆视角研究光照环境突变条件下的视觉辨识以及融合感知技术。分别选取低照度车辆进入隧道以及弱光线条件下车辆驶离隧道两种情形,利用局部能量、卷积稀疏表示算法(CSR)对两种图像进行融合实验,结合MI、SF、AG、QAB/F、SSIM、PSNR六种评价指标进行评价。实验结果表明,在隧道入口处图像CSR-E算法对比Curvelet、NSCT、NSCT-T、SR-C&L、SF-Energy-Q五种算法,边缘信息传递因子(QAB/F)提高了14.14%,隧道出口处图像运行平均时间减少1.17 ms,结构相似性(SSIM)提高了3.38%,所提出的红外非可见光与可见光视觉融合成像方法弥补单一传感器针对特定场景表达的不全面,实现对场景全面清晰准确的表达,有效解决了源图像的边缘信息丢失,增强图像的光谱信息。

Research on visual fusion technology for entrance and exitof mountain expressway tunnel

Aiming at the different photosensitive characteristics of infrared non visible light and visible light vision in the imaging process,and facing the typical "black hole" and "white hole" problem in tunnels,the visual identification and fusion perception technology under the sudden illumination environment is studied from the perspective of autonomous vehicles. Two scenarios are selected:low illumination vehicles entering the tunnel and vehicles leaving the tunnel under low light conditions. Local energy and convolution sparse representation algorithm (CSR) are used to fuse the two images,and six evaluation indexes including MI,SF,AG,Q AB/F,SSIM and PSNR are used to evaluate the images. The experimental results show that CSR E algorithm for images at the tunnel entrance improves the edge information transfer factor (Q AB/F) by 14.14%,the average running time of image at the tunnel exit is reduced by 1.17ms and the structural similarity (SSIM) is improved by 3.38%compared with the five algorithms of Curvelet,NSCT,NSCT T,SR C&L and SF Energy Q. The proposed infrared non visible and visible vision fusion imaging method made up for the incomplete representation of a specific scene by a single sensor,achievesa comprehensive,clear and accurate representation of the scene,effectively solves the loss of edge information in the source image,and enhances the spectral information of the image.