水下双目视觉系统中的目标分割和目标定位

发布时间：2019-01-14 10:55

【摘要】：针对典型灰度化和阈值方法用于水下目标分割时存在的目标分割不完整等问题,提出了基于灰度对比度增强的两步目标分割方法:第一步基于彩色矢量阈值进行目标粗分割;第二步分析局部对比度和区域内均匀性,动态确定不同通道的权值和反向处理,进而增强灰度对比度,完成目标精分割.针对典型标定、立体匹配和目标位置测量方法不适用于本研究环境的问题,提出了基于目标尺寸估计的目标定位方法,首先完成在线外参标定和目标圆心匹配,然后利用双目位置测量数据估计目标的尺寸,最后采用单目位置测量得到定位数据.结果表明:该分割方法能够准确分割目标,分割准确率和错分类比率均优于典型方法,定位方法能够得到目标三维位置数据,研制的双目视觉系统能够有效配合水下运载器-机械手系统完成自主作业任务.
[Abstract]:In order to solve the problem of incomplete target segmentation in typical grayscale and threshold method, a two-step object segmentation method based on gray contrast enhancement is proposed: the first step is based on color vector threshold; The second step is to analyze the local contrast and the homogeneity in the region, determine the weights and the reverse processing of different channels dynamically, and then enhance the gray contrast to complete the fine segmentation of the target. Aiming at the problem that the typical calibration, stereo matching and target position measurement methods are not suitable for this research environment, a target location method based on target size estimation is proposed. Firstly, the on-line external parameter calibration and the target center matching are completed. Then the size of the target is estimated by the binocular position measurement data, and the location data is obtained by the monocular position measurement. The results show that the segmentation method can segment the target accurately, the segmentation accuracy and the error classification ratio are better than the typical method, and the location method can get the 3D position data of the target. The binocular vision system can effectively cooperate with the underwater launch vehicle-manipulator system to complete the autonomous task.
【作者单位】：哈尔滨工程大学机电工程学院;
【基金】：国防基础科研基金资助项目(B2420133003)
【分类号】：TP391.41

【相似文献】