Abstract:Due to the nonconformity of coating formula and low-level automation of seedcoating machines, seed-coating technologies in China exhibit low coating-success rates, detection accuracy and efficiency. To solve these problems, an recognition and detection system for pelleted coating seeds was designed to recognize the coating seeds with spherical shape. Firstly, a vision shooting platform was built, and the captured image was transferred to the recognition control system for image pre-processing. Secondly, according to the characteristics of different types of coated seeds after image processing, a recognition and detection algorithm was proposed. According to the difference of image area ratio between damaged coated seeds and other coated seeds, the recognition of damaged coated seeds was realized by advanced morphological processing. The identification of multiple seeds and qualified seeds was realized according to the difference of the pixel values of multiple seeds and qualified seeds. Finally, the total number of seeds, the number of qualified seeds, the number of multiple seeds and the number of damaged seeds were detected, and the qualified rate of coating was calculated. The experiment was carried out on red clover seeds. The results showed that the time of single image acquisition, processing and recognition was about three seconds. The accuracy of using advanced morphological treatment to identify damaged coated seeds was 98.8%. When the test samples was 200, the success rate of the total number recognition algorithm was 99.1%, the relative error rate of qualified coating seeds and multiple coating seeds was 1.18% and 3.36% respectively. All these results suggested that the developed recognition and detection system realized the functions of shooting, image processing, detection and recognition, as well as output and storage of results. Therefore, the developed recognition and detection system can be used to fulfil non-destructive testing of coating seeds.