In-Situ Soil Characterization by Vision Cone Penetrometer (V-CPT).

摘要建议：CMS 97-00128 PI：Roman D.Hryciw，密歇根大学“用视锥渗透仪进行土壤原位表征”本研究计划的目标是发展用于土壤原位表征的计算机视觉方法。利用新近研制的视锥透过仪(V-CPT)获得了连续的地下图像。从数字获取的图像将确定土壤的粒度分布。为了完成这项任务，将研究许多可能的VISION方法。它们包括边缘检测方案，如Hough变换以及主动轮廓绘制、线性扫描和纹理分析。对土壤的目视观察将提供对土壤类型和地层的连续和明确的评估。因此，利用拟议的亚表面视觉探测器进行现场观测将是非常可取的。研究计划包括继续改进V-CPT硬件，包括最先进的数字视频记录；开发前面描述的视觉技术；以及在密歇根大学校园和西北大学国家岩土试验现场进行现场测试。****

Abstract Proposal: CMS 97-00128 PI: Roman D. Hryciw, University of Michigan "In-Situ Soil Characterization by Vision Cone Penetrometer" The objective of this research program is the development of computer vision methods for in-situ soil characterization. Continuous subsurface images are obtained by a recently developed Vision-Cone Penetrometer (V-CPT). From the digitally acquired images the grain size distribution of the soil will be determined. A number of possible vision approaches will be investigated to accomplish this task. They include edge detection schemes such as Hough transforms along with active contouring, linear scanning, and textural analysis. Visual observation of the soil would provide a continuous and unambiguous assessment of soil type and stratigraphy. As such, in-situ observation by the proposed Subsurface Vision Probe would be highly desirable. The research program includes continued improvement of V-CPT hardware by inclusion of state-of-the-art digital video recording; development of the previously described vision techniques; and field testing both on the University of Michigan campus and the Northwestern University National Geotechnical Experimentation Site. ****