We present a workflow for rapid delineation and microtopographic characterization of ice wedge polygons within high-resolution digital elevation models. The workflow, which is extensible to other forms of remotely sensed imagery, incorporates a convolutional neural network to detect pixels representing troughs. A watershed transformation is then used to segment imagery into discrete polygons. Regions of non-polygonal terrain are partitioned out using a simple post-processing procedure. Results from training and validation sites at Barrow and Prudhoe Bay, Alaska demonstrate robust performance in diverse tundra landscapes. The methodology permits fast, spatially extensive measurements of polygonal microtopography and trough network geometry.