Research on extraction of cotton seedling plant number based on UAV RGB image
Rapid,accurate and large-scale acquisition of cotton seedlings number plays a vital role in the decision-making of early cotton breeding and the realization of precise management of cotton fields.Aiming at the interference of mulching film and other disturbances in the field of cotton seedlings,which can easily affect the extraction of plant numbers,this paper proposes a method based on the over-red index and ultra-green index,combined with image processing methods for counting the number of cotton plants in UAV RGB images.This article uses the UAV images of cotton fields in Awati County,Aksu Prefecture,Xinjiang to conduct research,and preprocesses the collected data,and calculates the Excess green index(ExG)and Excess red index(ExR),Otsu threshold segmentation and other processing,and then the processed binary image noise and misclassified images generated by the mulch film are selected to use Majority analysis processing for denoising,among which,the denoising effect of 3×3、5×5、7×7、9×9 transformation kernels with different sizes in the Majority analysis is compared and analyzed,and finally the the number of cotton seedlings is extracted.The experiment shows that the extraction effect of cotton plant number is the best under the redness index 9×9 conversion kernel treatment,and the accuracy rate of statistical plant number reaches 97.84%.The statistical accuracy rate of the super green index for the number of cotton plants after different sizes of transformed kernels is above 95%,and the accuracy rate of cotton plant numbers extracted based on the 5×5 transformed kernels reaches 98.86%.The results show that this method can not only improve the accuracy of counting the number of cotton plants,but also provide technical support for early breeding and precise management of cotton fields.
cotton breedingcotton field precise managementnumber of cotton plantsExcess red indexExcess green indexthreshold segmentation
NETL
NSTL
万方数据
