Leaf Counting with Multi-Scale Convolutional Neural Network Features and Fisher Vector Coding

Abstract

The number of leaves in maize plant is one of the key traits describing its growth conditions. It is directly related to plant development and leaf counts also give insight into changing plant development stages. Compared with the traditional solutions which need excessive human interventions, the methods of computer vision and machine learning are more efficient. However, leaf counting with computer vision remains a challenging problem. More and more researchers are trying to improve accuracy. To this end, an automated, deep learning based approach for counting leaves in maize plants is developed in this paper. A Convolution Neural Network(CNN) is used to extract leaf features. The CNN model in this paper is inspired by Google Inception Net V3, which using multi-scale convolution kernels in one convolution layer. To compress feature maps generated from some middle layers in CNN, the Fisher Vector (FV) is used to reduce redundant information. Finally, these encoded feature maps are used to regress the leaf numbers by using Random Forests. To boost the related research, a relatively single maize image dataset (Different growth stage with 2845 samples, which 80% for train and 20% for test) is constructed by our team. The proposed algorithm in single maize data set achieves Mean Square Error (MSE) of 0.32.

Leaf-Counting in Monocot Plants Using Deep Regression Models

“…In this work, we advocate regression-based approaches because they require marginal efforts to annotate leaf numbers and can significantly reduce the costly and time-consuming labeling process. However, the existing regression-based approaches cannot fully address monocot plants due to leaf occlusion problems, unclear plant structures resulting from the illumination, and a lack of advanced models [ 6 , 9 ]. To address these issues, we note that a combination of advanced architecture, a distinct plant structure representation, and a wide variety of leaf shapes is promising in improving the accuracy of regression-based approaches for counting the leaves of monocot plants.…”

“…Most existing works required a large-scale dataset and usually failed to count the number of leaves with overlaps or occlusions. Jiang et al [ 6 ] used a regression-based approach with Google Inception Net V3 [ 7 ] associated with Fisher vector coding [ 8 ] to count the number of leaves in the maize plant. The absolute counting difference for the testing dataset was 0.35, with a mean squared error (MSE) of 0.31.…”

Leaf Counting: Fusing Network Components for Improved Accuracy

“…It enables growth rate estimation and is related to the health status of the plant and its yield potential (Telfer et al, 1997;Walter and Schurr, 1999). Manually measuring the traits of the visual plant is a slow, tedious, and expensive process (Jiang et al, 2019), and it usually requires the presence of specialized investigators (Giuffrida et al, 2016). Hence, these traits are measured on a small random sample of plants, which might lead to a measurement bias (Aich and Stavness, 2017).…”

“…With the growing need for systematic plant phenotyping (Großkinsky et al, 2015 ) and the development of recent Convolutional Neural Network (CNN)-based techniques (Ren et al, 2015 ; He et al, 2016 ), visual leaf counting has attracted considerable attention (Giuffrida et al, 2016 ; Dobrescu et al, 2017 ; Lu et al, 2017 ; Teimouri et al, 2018 ; Jiang et al, 2019 ; Kuznichov et al, 2019 ). A basic but appealing idea is to perform counting by using some standard detection or segmentation network architecture (Ren et al, 2015 ; Redmon et al, 2016 ; He et al, 2017 ; Lin et al, 2017b ) to detect the leaves.…”

Automated Counting Grains on the Rice Panicle Based on Deep Learning Method

“…Deep learning, which is a novel method for object detection with greater accuracy, has been widely used for agricultural applications [ 14 ]. These include the detection and counting of corn kernels [ 15 ], the leaf counting in maize plants [ 16 ], the detection and analysis of wheat spikes [ 17 ], seed-per-pod estimation for plant breeding [ 18 ], and automatic estimation of heading date of rice [ 19 ]. Deep-learning based image analysis has broad prospects and can be used to accurately and effectively detect and count grains per panicle.…”