Abstract: Conventional vigor tests couldnt meet current demand for the development of automatic breeding because of their subjectivity， high cost and difficult operations. In order to improve the automatic level and quality of seed vigor detection， chili seed images were used to study the detection method. The innovative gray transformation was used to process seed\|color images. This method could protect seed information enough. Combining self\|developed system characteristics， threshold processing was used to remove noise， which could reduce the time of system processing seed images. For marking every seed， the paper used cvFloodFill algorithm to fill seeds into different gray values. Then the seed germination index and average seed radicle length could be acquired by self\|developed software. Finally we combined system hardware and software perfectly. Through processing different seed images， the result showed the system algorithm had certain advantages in the processing of such images. For assessing precision of system through calculation of chili seed vigor index， the results showed that the accuracy of seed vigor index detection was greatly high， surpass 92%. So this system could provide research foundation for online production of vegetable seeds.

Key words: chili seed, algorithms, image processing, vigor assessment, seed vigor index, accuracy