基于多特征融合的防爆自动导引车避障方法研究
2021-03-24万伟刘子龙孙帅张莹
万伟 刘子龙 孙帅 张莹



摘 要:为解决在工厂环境中防爆自动导引车(Automated Guided Vehicle,AGV)难以在不同光照条件下实现行人检测的问题,本文提出将“结构光+双目视觉相机”的图像采集方案应用于爆炸性危险环境。针对传统的基于RGB(代表红、绿、蓝三个通道的颜色)图像的HOG-LBP(Histogram of Oriented Gradient-Local Binary Pattern)方法受光照影响较大且对特征利用率低的情况,设计了基于RGB-D(RGB-Depth)图像的HOG-CLBP(Histogram of Oriented Gradient-Completed Local Binary Pattern)框架,并提出了一种改进的多特征融合方法,通过支持向量机分类实现实时行人目标检测。实验验证该方法可以有效解决在不同光照条件下实现对行人目标的检测,从而实现避障功能。
关键词:双目视觉;行人检测;避障;特征融合
中图分类号:TP29 文献标识码:A
Abstract: This paper proposes an approach to applying image acquisition with "structured light + binocular vision camera" to an explosively dangerous environment since it is difficult for explosion-proof Automatic Guided Vehicle (AGV) to detect pedestrians under different lighting conditions in factory environment. Traditional HOG-LBP (Histogram of Oriented Gradient-Local Binary Pattern) method based on RBG (red, blue, green) channels of images is greatly affected by illumination and has low utilization rate. Therefore, a HOG-CLBP (Histogram of Oriented Gradient-Completed Local Binary Pattern) framework based on RGB-D (RGB-Depth) images is designed, and an improved multi-feature fusion method is proposed to realize real-time pedestrian target detection through support vector machine classification. Experiments verify that this method can effectively solve the problems of pedestrian targets under different lighting conditions to achieve the obstacle avoidance function.
Keywords: binocular vision; pedestrian detection; obstacle avoidance; feature fusion
1 引言(Introduction)
防爆AGV是指依據GB3836系列国家标准[1]设计,经相关防爆检验检测机构认证,可以应用于爆炸性危险环境的装备有自动导引装置的多功能移载运输车。随着2019年10月1日防爆电器正式被纳入CCC(China Compulsory Certification)认证管理范围,自2020年10月1日起,未取得3C证书和认证标志的防爆产品被禁止出厂、销售、进口,或在其他经营活动中使用,因此防爆AGV具有极为广阔的市场前景。
与通用型AGV结构相同,防爆AGV主要通过隔爆型、增安型和本质安全型等[2]防爆方式解决AGV小车的控制单元、驱动单元等部分的电气防爆问题。由于应用环境的特殊性,通用型AGV常用的激光雷达等避障方案在爆炸性危险环境中较难解决防爆安全问题。由于AGV运行轨迹通常较为固定,且在爆炸性环境中对于物料堆放有着严格的规定,因此本文中需要躲避的障碍对象主要为出现在小车前进路线上的行人。……
