Wide-area space surveillance sensors are the backbone to cataloging of Earth orbiting objects. Their core capability should be to efficiently detect as many space objects as possible over a large space domain. As such, the question of how to quantitively evaluate the object detection performance of the sensors is critical. The evaluation is traditionally performed by means of infield static tests and out-field calibration satellite tests. However, this simplified method is flawed in terms of its representativeness in spatial-temporal coverage and object types, because space objects vary greatly in orbit type, size, and shape, and thus the evaluation results may be overoptimistic. This paper proposes a practically implementable procedure to quickly and reliably evaluate the object detection performance of space surveillance sensors in which a catalog containing a vast number of on-orbit objects is used as a reference. It first constructs a unified model to estimate the size of objects from its radar cross section (RCS) data, then it presents a hierarchy scheme to efficiently compute object visibility, and finally, it makes the sensor performance evaluation through a data point matching technique. Experiments with two simulated sensors demonstrate that the realized performance is always inferior to the designed one, and in some cases the difference is significant and concerning. The presented approach could be routinely applied to evaluate the performance of any operational surveillance sensors and provide insight on how the sensor performance could be improved through refined design, manufacture, and operation.