Abstract
<abstract> <bold>Abstract.</bold> In order to provide useful information for flower thinning and fruit production forecasting of apple trees, it is valuable to accurately identify the number of flowers in blossom. This article aims at exploring the method of calculating the flowering amount on each apple tree based on aerial multispectral images, and an algorithm to estimate the number of apple tree flowers was proposed. Experiment field is located in an apple orchard in Changping district of Beijing China. The multispectral camera was mounted under the unmanned aircraft, then the multi-spectral images of 60 apple trees with different flowering level were acquired from air overlooking view. In order to distinguish the flower, leaf, and soil separately from the aerial images, the spectral characteristics of three target elements in the multispectral image were analyzed. The reflectance at three channels (R/G/NIR) and NDVI were selected as classification independent variables. With the help of C-Support Vector Classification, the objects in each image were classified into three categories. The parameters of the model were adjusted until the accuracies of calibration and validation reached high. The SVM classification accuracy rate is above 86.67%. After counting the number of pixels in each category classified using the classification model, the mosaic ratio was calculated and it can be used as the indicator of the amount of flowering for flower thinning. Finally, the possible solutions to improve the estimation methods were discussed. This study proposed a simple and effective method based on aerial multispectral images to estimate the number of flowers of apple tree in blossom, and it could contribute to flower thinning and fruit production forecasting in the future.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call