The work aims to propose a glue dispensing quality detection scheme based on HALCON with respect to the problems of high detection cost, slow speed and low precision in the process of existing glue dispensing quality detection. By means of image de-noising, template analysis, image segmentation and morphological processing, the glue contour of the sample to be tested was obtained. The difference set with the qualified glue's centerline was operated to check for the broken glue and tell if wave glue existed according to the width of glue contour. The experimental results showed that the scheme could quickly, accurately and effectively determine the qualified or unqualified products, and detect such problems as wave glue and broken glue. The detection accuracy rate was 99.6% and the average detection time was about 475 ms. The detection accuracy and detection speed of the scheme meet the detection requirements of industrial production line, and can be applied in the actual production.