| According to the information released by the Bureau of Statistics of China on December 12,2022,the total grain output of China in 2022 has reached 686.53 million tons,and among the main grain crops such as wheat,soybeans,corn and rice,the output share of corn is the highest,so the stable increase of corn output will play a positive role in ensuring the national grain and oil output.At the same time,a lot of automation is used in the process of corn planting,which on the one hand increases the grain output,but on the other hand raises new problems for the safe planting work.With the development of agricultural science and technology,precision agriculture has been regarded as an important production mode in the process of corn production in order to improve the yield of corn.Compared with the traditional production mode,the identification and positioning of seedlings through precision agriculture in the traditional process of sowing,weeding and fertilization can effectively reduce the harm of operating equipment or pesticides on corn seedlings.The study of image processing technology based on computer vision can provide theoretical support for maize seedling feature recognition and stem position marking,which can promote agricultural production to become more safe and efficient.The research content is introduced in the following four parts.The image samples of maize seedlings were collected.In the process of maize growth,stems and leaves in the seedling stage overlapped less,which was conducive to the feature extraction of the stem part,further obtaining the position information of maize stems and marking,which could provide reference for the field operation requiring the protection of maize seedlings.In this study,the maize seedlings at 4～6 leaf stage were taken as the object of observation,and color images of maize seedlings were obtained by CMOS camera.The pretreatment method of maize seedling image is studied.According to maize image sample,the image is processed by gray level,image enhancement and filtering.After comparing the effect evaluation of various methods,the histogram equalization combined with median filtering can reduce the influence of uneven illumination in maize seedling image to the greatest extent,which lays a foundation for the subsequent image segmentation.Maize seedling image segmentation was studied to realize maize seedling recognition.LBP technology was used to extract texture information from preprocessed images.Kmeans clustering algorithm was used to segment images based on texture feature data,and Otsu technology was used for binary processing of gray-scale images after segmentation.Based on the binary images obtained,morphological processing was carried out to further eliminate the noise,and the binary values that could clearly identify the maize seedling body were obtained.The position markers of maize seedling stems were studied.The connected domain was constructed and the maximum connected domain was selected as the body region of maize seedling.The pixel values of this region were accumulated by projection method,so as to obtain the maximum point and further obtain the pixel coordinates of the center point of maize stem roots.In this study,due to the randomness of the growth of maize seedling leaves,there will be two interference cases: single leaf and double leaf.Through the projection method limiting the value range of x and taking the maximum value of y axis as the y axis coordinates of pixel points within the diameter range of the stem,two-dimensional coordinate data of the root of the stem was finally obtained.In this study,200 maize seedling images were tested,and 20 groups of test results were selected and compared with manual labeling results.The experimental results showed that the mean value of horizontal error was 7.55,the mean value of standard deviation was 1.04%,the mean value of column error was 10.85,the mean value of standard deviation was 2.26%,and the mean value of actual error was 12.33.Through numerical analysis,this study was able to obtain more accurate position coordinates of maize stem roots as the result of maize seedling stem position markers.In order to facilitate the application of this method in practice,an application system of seedling stalk position labeling was designed.The imported seedling image was used as the data source,and automatic stalk position labeling was realized through three main functional modules:image preprocessing,image segmentation and corn seedling stalk position labeling.To sum up,an image-based maize stalk location marking method was proposed and an application system was designed in order to locate maize stalks during field operations.The study of this method can provide reference for the field operation and the development of relevant detection equipment which need to protect maize seedlings. |
PDF Full Text Request