Solder　balls　used　in　electronic　packaging　should　have　high　dimensional　and　shape　accuracy　requirements　to　ensure　the　connection　quality　of　solder　balls　during　soldering.　However,　the　existing　solder　ball　detection　methods　were　hard　to　measure　the　diameter　and　roundness　in　an　integrated　and　fast　manner.　Therefore,　a　new　method　and　device　for　measuring　the　diameter　and　roundness　of　solder　balls　was　proposed　based　on　machine　vision　technology,　including　detection　device　and　software　design.　The　experimental　device　could　detect　the　parameters　of　at　least　40　solder　balls　at　a　time.　The　solder　ball　image　was　preprocessed　in　the　software　system　to　improve　the　quality　of　the　solder　ball　image,　and　then　the　Harris　operator　successfully　detected　the　concave　point　on　the　bonded　solder　ball　image　and　realized　the　solder　ball　separation　based　on　the　shortest　path　matching　criterion.　Compared　with　the　scanning　electron　microscope　(SEM)　measurement　value,　the　diameter　and　roundness　of　solder　balls　measured　by　the　detection　system　were　smaller　than　3%.