Abstract: Abstract: The seeds cannot fully sprout owing to the seed quality, sowing precision and environmental differences. In order to get the accurate information of the null plug holes and the boundary of plug holes and provide the basis for automatic seedling supplement device, a automatic seedling supplementing test-bed was developed with seedling shortage detecting system and seedling supplementing system as core components. The seedling shortage detection system was composed of a hardware system for image processing, which include a CMOS industrial camera, a controller and a computer, and a software system programmed in MATLAB. Color images (RGB images) of Arabidopsis plug seedlings with the age of 25 and 35d was acquired with the automatic seedling supplementing test-bed. The grayscale images of seedlings and plug holes are obtained by graying the color images applying different linear transformations to three color components of R, G and B. Applying Otsu algorithm, binary images of the grayscale images of seedlings were obtained by threshold segmentation. Then the morphological corrosion operation and expansion operation of the binary images were carried out by the disk 2 × 2 type structure operator. By marking single connected domain, analyzing the characteristics of connected domain and removing the isolated area, the noise in the binary images were removed and the characteristic images of the seedlings were extracted from the background effectively. The feature images of plug tray were acquired by removing feature images of plug seedlings from that of the plug tray according to the extracted seedling information and de-noising. And then statistics on the peak value, peak width and standardizing structure of plug tray were made according to the row and column pixel of binary images of plug trays, so that the edges of plug holes were determined accurately. The statistics of the pixel of the seedling image in each plug hole were made to determine whether the hole was short of seedlings according to the feature images of the seedlings, the position, and the edge information of the plug holes. The results showed that the statistics value of the seedlings image pixel of the plug holes with the Arabidopsis plug seedlings of 25 and 35 d were 1 895 to 4 572, and 3 386 to 8 710, respectively, while the statistics value of the seedlings image pixel of the plug holes without seedlings was 0. There were significant differences in the statistics value of the seedling image pixels between the plug holes with seedlings and the null plug holes. The testing results of the missing plug tray hole were the same as the actual situation, and the statistical value of seedling pixels corresponds to the projection area of stem and leaf. According to the threshold value of seedlings at different growth stages, the undeveloped seedlings can be marked for removing, so the seedling early stage was the super time for the detection of the null plug hole and the determination of the undeveloped seedlings. The accuracy rate of judging the null plug holes and the seedling holes with the Arabidopsis plug seedlings of 25 and 35 d were all 100% applying the detection device. The accurate determination of position of the null plug hole provides basis for automatic supplementing system taking out the substrate without seedlings, removing dysplasia seedlings and supplementing the healthy seedlings with same age.