Construction and rehabilitation of highways, tunnels, and bridges require detailed information about subsurface stratigraphy. This study presents development of a new method for characterizing subsurface soil in situ using computer vision. Hardware and software systems are integrated to obtain the grain-size distribution (GSD) of subsurface soils continuously with depth and to identify small-scale subsurface anomalies. Research is being conducted in three phases. The first phase consists of measuring the GSD of detached cohesionless soil specimens in the laboratory from digital images obtained with a computer vision system (CVS). The second phase uses the CVS to develop image processing and analysis techniques to classify soil assemblies in the laboratory and identify subsurface anomalies by simulating the manner in which images will be acquired in situ. A texture analysis approach has been developed that can detect changes in stratigraphy. The technique has been successful in identifying different types of dry, uniformly graded soils. Finally, a subsurface vision probe is being designed and constructed that will capture video images at three different levels of magnification continuously with depth.