<i>A performance comparison of RGB, NIR, and depth images in immature citrus detection using deep learning algorithms for yield prediction</i>

Abstract

Yield forecasting is important for farm management. In this study, red, green, and blue (RGB), nearinfrared (NIR), and depth sensors were implemented in an outdoor machine vision system to determine the number of immature citrus in tree canopies in a citrus grove. The main objective was to compare the performances of three image data types for citrus yield forecasting. The performance comparison was conducted with two machine vision algorithm steps: 1) circular object detection for potential fruit areas and 2) classification of citrus fruit from the background. For circular object detection, circular Hough transform was used in the RGB and NIR images. For the depth images, CHOI's Circle Estimation ('CHOICE') algorithm was developed using depth divergence and vorticity to find circular objects in the depth images. The classification process was conducted using AlexNet, a deep learning algorithm for all three image types. The implementation of a convolutional neural network allowed the machine vision algorithms to remain bias-free process during feature generation and selection. NIR images performed best with 96% true positive rate for both the circular object detection and classification. A machine vision system using this image type will produce a more objective yield prediction with a higher accuracy than other types.

An Extensive Review of Mobile Agricultural Robotics for Field Operations: Focus on Cotton Harvesting

“…The normal digital camera may be used if the target on the field can be visually identified. For example, to identify green citrus or green bell pepper in a population of green plants may require using an alternative sensor, or the method of detection may be complicated by involving advanced methods of machine learning [2,[34][35][36][37][38]. Images may suffer from illumination changes, motion change, cluttering, temperature swings, camera motion, wind-induced movements, deformation, and scene complexity.…”

“…Images may suffer from illumination changes, motion change, cluttering, temperature swings, camera motion, wind-induced movements, deformation, and scene complexity. Hence, some image refinement algorithms may be required to enhance the images [35,38]. Then, object recognition or feature extraction using pattern recognition and other machine vision algorithms can be performed.…”

“…Then, object recognition or feature extraction using pattern recognition and other machine vision algorithms can be performed. There are several methods of image rectification and enhancement that have been reported; image smoothing and segmentation [2,36,39], morphological operations and filters [2,34,38], a fast bilateral filtering based Retinex [35], illumination normalization [2,35], image color space-changing [40], and normalized co-occurrence matrix and gray level co-occurrence matrix [41]. Feature extraction can be achieved using classical image processing techniques or advanced techniques in machine learning, such as color filtering and masking [2].…”

Machine Vision Systems in Precision Agriculture for Crop Farming

“…An analysis of the performance of RGB, NIR, and depth images for immature citrus fruit detection using machine vision techniques is conducted in [ 43 ]. First, circular object detection is performed to find potential fruit areas.…”

Experiments and Analysis of Close-Shot Identification of On-Branch Citrus Fruit with RealSense

“…Moreover, Choi et al [ 20 ] used Kinect V2 to synchronize the RGB, near-infrared (NIR) and depth images of on-tree green citrus. They applied a 2D Hough transform algorithm to RGB and NIR images and Choi’s circle estimation algorithm for the depth image to search the object fruit, then detected the fruit from the background with the AlexNet classification model.…”