Pear Flower Cluster Quantification Using RGB Drone Imagery

Abstract

High quality fruit production requires the regulation of the crop load on fruit trees by reducing the number of flowers and fruitlets early in the growing season, if the bearing is too high. Several automated flower cluster quantification methods based on proximal and remote imagery methods have been proposed to estimate flower cluster numbers, but their overall performance is still far from satisfactory. For other methods, the performance of the method to estimate flower clusters within a tree is unknown since they were only tested on images from one perspective. One of the main reported bottlenecks is the presence of occluded flowers due to limitations of the top-view perspective of the platform-sensor combinations. In order to tackle this problem, the multi-view perspective from the Red–Green–Blue (RGB) colored dense point clouds retrieved from drone imagery are compared and evaluated against the field-based flower cluster number per tree. Experimental results obtained on a dataset of two pear tree orchards (N = 144) demonstrate that our 3D object-based method, a combination of pixel-based classification with the stochastic gradient boosting algorithm and density-based clustering (DBSCAN), significantly outperforms the state-of-the-art in flower cluster estimations from the 2D top-view (R2 = 0.53), with R2 > 0.7 and RRMSE < 15%.

Technologies for Forecasting Tree Fruit Load and Harvest Timing—From Ground, Sky and Time

“…Another issue for use of a UAV is the visibility of flowers or fruit in-tree when viewed from above, as opposed to a side of tree view. A 3D object-based method using multi view perspectives from drone imagery outperformed a 2D top-view (R 2 > 0.7 compared to 0.53, against field-based counts) for the estimation of pear flower cluster number per tree [45]. Apolo-Apolo et al [46] report that a 3D reconstructed image from an aerial view accounted for only 27% of fruit on the 19 apple trees assessed, with an R 2 of 0.80, MAE of 129 and RMSE of 131 fruit per tree achieved for a linear regression of machine vision estimated fruit counts against hand harvest counts.…”

Status of Phenological Research Using Sentinel-2 Data: A Review

“…Several studies have used spectral mixture analyses to identify photosynthetic and non-photosynthetic fractions [76][77][78] and vegetation indices to estimate concentration of plant pigments [79][80][81]. However, similar efforts are lacking to discriminate flowers from other parts of the plant and existing techniques have been applied mostly to high-resolution aerial or hyperspectral data [82][83][84][85][86]. Notably, the EBI developed by Chen et al (2019) [73] was found to perform poorly in discriminating non-white flowers, and it may be that other band combinations are more suitable for different coloured petals.…”

Suitability of resampled multispectral datasets for mapping flowering plants in the Kenyan savannah

“…multispectral images with fewer broad bands for optimal discrimination of functional flowering groups [29,[37][38][39]. However, whereas broadband multispectral data of lower spectral and medium spatial resolution such as the Landsat series have become popular in landscape mapping, they could mask out specific spectral features of functional flowering groups, resulting in very low mapping accuracy.…”

“…Specifically, the Hughes effect and the high redundancy rates of some bands in models developed using hyperspectral data impede landscape classification [ 33 – 36 ]. In this regard, it is paramount to explore the utility of multispectral images with fewer broad bands for optimal discrimination of functional flowering groups [ 29 , 37 – 39 ]. However, whereas broadband multispectral data of lower spectral and medium spatial resolution such as the Landsat series have become popular in landscape mapping, they could mask out specific spectral features of functional flowering groups, resulting in very low mapping accuracy.…”