Defect inspection of the surface foreign matter and the link ring crack of polyethylene terephthalate（ PET） caps without support rings are researched. Use hue saturation value（ HSV） color space transformation

to overcome shortcomings of gray scale of caps surface and the foreign matter features are not obvious uneven im original red green blue（ RGB） image. Saturation image is used to effectively segment foreign matter region. Due to the crack gap is small and gray intensity is discontinuous

a dynamic threshold algorithm is employed. The algorithm uses mean filtering to estimate the background

compared the inspected image with the background image

and set dynamic threshold as a result

the crack region is segmented. Experiments and practical application demonstrate that the accuracy of detecting system rate up to 99 % and the detection speed is up to 12000 bottles per hour

which can satisfy online detection requirements of the production line.