Purpose In order to meet the requirements of fast, accurate and automatic thinning of tobacco seedlings, we put a forward automatic tray seedling thinning algorithm based on machine vision. It can avoid the shortcomings of the low efficiency, arbitrariness of the traditional human eye observation.
Methods K-means clustering was used to image segmentation of tobacco seedlings in Lab color space, according to the matrix row sum method, draw pixel coordinates between the two peaks, locate its the region position, divide the plug into 128 cells, convert the target area to a binary image. Research indicated that each kind of shape feature such as roundness, aspect ratio and rectangularity had different values and could be used as the separating parameters by comparing each feature respectively for the three ingredients, which are the single, multi-plant and holes. The use of seedling plants area and perimeter, in different growth period to set a suitable threshold to achieve the purpose of automatic seedling.
Results The simulation data and analysis showed the roundness of 1.256 6, aspect ratio of 1.571 4, rectangular of 0.716 5 the best difference between the effect. Taking the area distribution in the 111-243 (pixels), the circumference of the distribution in the 16-33 (pixels) can be determined as strong seedlings. We developed a tobacco seedling automatic thinning software system based on machine vision on the MATLAB R2015a environment.
Conclusion The result showed that the correct identification rate of tobacco seedling had reached more than 97.04%, the hole position had reached 100%, for the thinning position and sound seedling average rate were respectively 94.76% and 89.58%, this method provided a theoretical basis and technical support for the further development of the automatic seedling thinning machine based on machine vision.
DownLoad: