With　the　development　of　integrated　circuit　packaging　technology,　ultrasonic　scanning　detection　technology　is　more　and　more　widely　used　in　the　semiconductor　industry.　From　chips　to　various　MEMS　electronic　packaging　devices,　they　can　be　used　to　detect　their　internal　defects.　This　paper　introduces　the　basic　principle　of　ultrasonic　scanning　microscope　system　used　to　detect　multilayer　structure　and　the　key　technology　of　defect　judgment.　This　includes　a　phase　judgment　algorithm　based　on　the　amplitude　variation　of　A-scan　waveform　and　the　principle　of　phase　inversion.　By　marking　layered　defects,　the　algorithm　can　generate　scanned　images　with　real　defect　shapes,　which　improves　the　efficiency　of　manual　analysis.　In　this　paper,　several　multilayer　devices　and　the　corresponding　A-scan　waveforms　are　used　to　verify　the　algorithm.　This　paper　also　includes　the　statistical　method　of　defects　based　on　blob　analysis.　This　paper　analyzes　the　defects　according　to　the　characteristic　parameters　of　the　connection　area,　and　summarizes　a　set　of　defect　statistical　methods,　which　can　obtain　the　target　defect　information　statistics　and　determine　the　number,　size,　area　and　other　information　of　defects.　In　addition,　this　paper　also　proposes　an　image　interpolation　method　suitable　for　ultrasonic　scanning.　Experiments　show　that　this　method　can　improve　the　scanning　efficiency　about　twice　while　ensuring　the　image　clarity.　This　paper　provides　a　theoretical　basis　for　the　detection　method　of　multilayer　chip,　and　also　provides　a　reference　for　the　actual　production　practice.　©　2022　IEEE.