基于深度信息的照明目标识别方法  

Method for Illumination Target Recognition Based on Depth Information

作　　者：白旭 李宏坤[1] 陈星[1] BAI Xu;LI Hongkun;CHEN Xing(School of Mechanical Engineering,Dalian University of Technology,Dalian 116033,China)

机构地区：[1]大连理工大学机械工程学院,大连116033

出　　处：《自动化与仪表》2024年第6期137-141,共5页Automation & Instrumentation

摘　　要：照明用电存在的能源浪费问题越来越受到关注,图像传感方法逐渐应用在照明控制领域中。基于单目图像的照明传感方法容易受到光照和纹理等因素的干扰。针对以上问题,该文提出了结合深度信息进行背景分离,提取场景中潜在的照明目标,并利用图像矩作为描述子进行目标识别的方法。搭建了双目实验平台进行室内场景测试,结果表明,该方法可以有效分离、识别并定位出场景中的照明目标,相对单目背景分离方法得到的分离轮廓更完整,对背景与目标灰度差异不敏感,对于大面积、高精度照明控制场景具备一定的应用价值。The issue of energy waste in lighting has been increasingly receiving attention,and image sensing methods have been gradually applied in the field of lighting control.However,lighting sensing methods based on monocular images were susceptible to interference from factors such as lighting and texture.To address these issues,this paper proposed a method that combined depth information for background separation,extracted potential lighting targets in the scene,and utilized image moments as descriptors for target recognition.A binocular experimental platform was constructed for indoor scene testing.The results showed that this method could effectively separate,recognize,and locate lighting targets in the scene.Compared to monocular background separation methods,the separation contours obtained were more complete and less sensitive to grayscale differences between the background and targets.Therefore,this method had certain application value for large-area and high-precision lighting control scenarios.

关 键 词：照明控制 双目视觉 背景分离 图像矩 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]