This paper introduces a novel technique to assess the alignment quality of heliostats in operation. This technique combines theoretical imaging with aerial vision of an object heliostat reflecting the back of a neighbor target heliostat. An accurate optical model is developed to generate the theoretical images that allows to detect alignment errors by superposition on actual photographs. The resulting code is grounded in geometric transformations between systems of coordinates, and makes use of the pinhole camera model and the geometric determination of reflection points in concave mirrors. The optical model is experimentally validated with real heliostats at the National Solar Thermal Test Facility (NSTTF). In terms of pixel shifts in the target-reflected image, a sensitivity analysis of the technique is performed. The influence of several geometric factors is researched: camera position, camera focal length, heliostat-to-camera distance, and curvature of the facets. Preliminary analysis shows that canting errors as low as 0.25 mrad can be detected by facet framing.