Aiming　at　quantitative　identification　of　defects　in　structures,　the　scattering　characteristics　of　defects　in　structures　are　studied.　The　scattering　simulation　model　of　elliptic　defect　is　established　and　the　effect　of　defect　deflection　angle　and　size　on　the　de-correlation　of　scattering　signal　is　studied.　The　results　show　that　the　de-correlation　coefficient　matrix　of　ultrasonic　signal　has　a　great　correlation　with　the　defect　geometry　state　(such　as　size　and　angle),　but　it　is　not　a　simple　linear　correlation.　On　this　basis,　combining　the　parametric　finite　element　analysis　and　Bayesian　theory,　a　defect　inversion　method　based　on　ultrasonic　scattering　signal　de-correlation　analysis　is　developed,　and　the　numerical　simulation　and　detection　experiments　of　defect　inversion　are　carried　out.　The　results　show　that　the　developed　defect　inversion　method　can　well　realize　the　inversion　of　the　position　and　size　of　defect　in　aluminum　block　and　plate.　The　inversion　error　of　the　open　defect　size　(depth)　of　the　aluminum　block　is　about　0.2　times　of　the　excitation　wavelength,　and　the　position　error　is　about　0.6　times　of　the　excitation　wavelength.　The　inversion　error　of　the　circular　defect　size　of　aluminum　plate　is　about　0.04　times　of　the　excitation　wavelength,　and　the　position　error　is　about　0.4　times　of　the　excitation　wavelength.　The　research　work　has　made　a　beneficial　exploration　for　the　quantitative　identification　and　evaluation　of　defects　in　structures.　©　2021　Acta　Acustica.