| The current prevalence of bottled liquid foreign matters detection methodmainly comprises artificial detection method and high performance liquidchromatography (HPLC) method. Using artificial detection method is complicated.It wastes much time and manpower. And HPLC method has high cost. Thesemethods can’t meet the requirements of automated production. The machine visiontechnology, owing to the advantages of non-contact, fast speed, and high detectionprecision, has been valued by more and more manufacturers. In this paper, wedesigned a set of on-line detection system for bottled liquid foreign mattersdetection based on machine vision technology, its main contents were as follows:(1)We formulate image acquisition program, select a light source and lightingfixtures for the system and build the experimental platform of image acquisition.(2)Based on a detailed analysis of acquired image, we choose Non-localaveraging algorithm to eliminate noise. We research a new image edge detectionalgorithm based on NSCT and morphology. Compared with the traditionalalgorithm, the experimental results show that this algorithm can detect the completeand continuous image edge, and has a strong anti-noise performance.(3)We adopt gray image threshold segmentation method for the single imageforeign matters detection and adopt optical flow method, classical frame differencemethod for sequence images foreign matters detection. By analysing the comparedexperiment, this paper uses a fuzzy support vector machine method to test thequality of the liquid and further improves the ability of processing complexproblem and anti-noise ability.(4)The image processing algorithms based on PC were realized onSEED-VPM6467T successfully，and the real-time detection of dynamic images wasdiscussed preliminary. The software design，focusing on the demand of inspectionand the characteristic of DSP, was accomplished and a detailed description of theprogram is provided. Finally, we design friendly user operator interface. |
PDF Full Text Request