A system for the thermal imaging of sonically or ultrasonically excited subsurface defects in a structure. The system includes a hand-held sound source, a thermal imaging camera and a control unit. The sound source emits pulses of sound energy into the structure, and the camera generates images of defects in the structure that are heated by the sound energy. The control unit controls the operation of the sound source on the camera for timing purposes. The sound source includes a transducer that is positioned against the structure at a desirable location. The source further includes a pair of legs that are also positioned against the structure to define a plane in combination with the transducer. The length of each leg is adjustable relative to the length of the transducer so that the gun can be used against irregular surfaces. The legs include a rubber tip to further prevent the transducer from slipping on the structure. In an alternate embodiment, the gun includes three transducers that define a plane and act to stabilize the gun against movement.