Efficient Determination Of Relevant Indicators In An Image-based Rice Seed Germination Test

Seed vigor is an important index to reflect seed quality,and seed vigor detection is related to the yield and product quality.At present,standard germination test is still the main method for determination of rice seed vigor.Determination of seed germination rate is one of the important means of seed vigor detection.The traditional method of manual determination of bud length and root length is complicated,time-consuming and laborious,and susceptible to subjective judgment.Based on image processing technology,this thesis provides an efficient and convenient method to measure the indexes of seed germination test.This thesis mainly studies the following contents:(1)Rice seed germination image pretreatment and image segmentation.After the image is collected,the image is preprocessed,suitable grainization processing algorithm,image morphology denoising,edge detection and other algorithms are selected,and the adaptive Otsu method based on the improved manual experience is used to segmentation the rice seed image.(2)Seed region localization of rice seed image.By RGB and HSV color space conversion method,analyzes the space,through analyzing three rice sprout pictures and HSV component analysis,puts forward the decision tree identification method to determine seed position is segmented,and to highlight and extracting seed area boundary Canny edge detection operator is adopted,positioning preserved the coordinates of the seed.(3)Isolation and measurement of rice image bud length and root length.Based on depth-first search and skeleton extraction,a method to obtain the longest path of skeleton image based on DFS(Depth First Search)was proposed.The bud and root were separated by the center location information of seed region,and the buds and roots were automatically detected and measured respectively.Combined with the measured data of 100 rice seed seedlings,the experimental results show that the average percentage error of bud length and root length is 1.48% and 1.57%respectively.Finally,a visual program was designed using Python development language,and related programs were written.The measuring system of rice bud and root length was designed,which realized the automatic segmentation,recognition and measurement of the target area,and the processing time of an image was about 2s on average,which met the requirements of use.The experimental results show that the method proposed in this thesis based on image processing technology is feasible and can effectively improve the accuracy and efficiency of measuring the sprout and root length in the germination test.