面向路灯照明自适应调节的雾天能见度分类

Visibility Classification in Fog for Adaptive Regulation of Street Lighting

在智慧城市研究领域中,基于能见度分类网络进行路灯照明的智能调节,是近年来有效减少雾天发生交通安全事故的重要研究分支。但是,现有相关方法研究普遍存在准确率低和结果滞后等问题。基于此,提出了一种面向路灯照明自适应调节的雾天能见度分类方法。首先,通过分析多色彩空间的像素级信息,从而实现对图像的日夜判别;其次,基于此通过融合堆叠式注意力分割的残差网络,对雾天图像进行能见度精细化分类;最后,通过雾天能见度分类结果引导路灯自适应调节亮度、色温。实验验证表明,所提方法在自建数据集中分类准确率达到了80.56%,相较于ResNet提升了1.39%,能够有效引导路灯照明的自适应调节,为雾天车辆行驶提供了安全保障。

In recent smart city research,intelligent regulation of street lighting based on visibility classification network is an impor-tant research branch to reduce traffic safety accidents effectively in foggy days.However,existing research on related methods generally has problems such as low accuracy and lagging results.To this end,this paper proposes a fog visibility classification method for adaptive adjustment of street lighting.The method firstly analyzes pixel-level information in multi-color space to discriminate the day and night of the image.Secondly,based on this,the fog visibility classification is refined by fusing residual network with stacked attention segmenta-tion.Finally,fog visibility classification results are used to guide street light to adjust the brightness and color temperature adaptively.Experimental validation shows that the proposed method achieves 80.56%classification accuracy in self-built dataset,which is 1.39%better than ResNet.And this method can effectively guide the adaptive adjustment of street lighting,which provides a safety guarantee for vehicle driving in foggy days.