摘要
在结构光测量系统中需要对采集到的激光条纹图像进行有效处理。该文提出了一种基于差图像的阈值分割方法,应用该方法对采集到的激光条纹图像进行阈值分割从而得到二值激光条纹图像,利用神经网络对二值激光条纹图像进行细化处理,处理过的细化条纹图像满足0.5%的精度要求,图像二值化处理时间不到1s,细化过程的时间小于2s,相对传统的图像处理时间有所增加,已经成功地应用在物体三维重建时特征参数匹配的预处理中。
The laser stripe image gathered in the structured light measurement system is needed to be processed effectively. A threshold segmentation method based on the difference image is put forward, which is used to segment the laser stripe image and a two-value image is obtained, then neural network is applied to process the two-value laser stripe image to get the thinning stripe. The processed thinning stripe satisfies the requirement that the precision is 0.05%, the two-value processing time is no more than one second and that of the thinning is less than two seconds. The time increased compared to traditional methods and the result has already been applied to the pre-processing of feature parameter matching in the object 3D reconstruction successfully.
出处
《计算机工程》
CAS
CSCD
北大核心
2007年第16期162-163,共2页
Computer Engineering
基金
国家自然科学基金资助项目(60572030)
教育部博士学科点专项科研基金资助项目(20050214006)
黑龙江省教育厅海外学者资助项目(1055HZ027)
哈尔滨市重点科技攻关计划资助项目(2005AA1CG152)
关键词
激光条纹
阈值分割
神经网络
细化
laser stripe
threshold segmentation
neural network
thinning
