Rootstocks for Apple - WSU Tree Fruit

Apple rootstocks can have a variety of desirable characteristics such as resistance to crown rot oomycetes (Phytophthora spp.), resistance to fire blight bacteria (Erwinia amylovora), resistance to Woolly apple aphids, dwarfing and tree branching modifications, increased precocity (early fruitfulness), increased productivity, and tolerance to apple replant disease (ARD). There are five major types of rootstocks used in the US. These include: Budagovsky (= Bud or B), Cornell/Geneva (=CG or G), Malling (=M) & Malling Merton (=MM), Michigan Apple Rootstock Clones (=MARK), and East Malling/Ashton Long (=EMLA) which are certified virus-free selections of Malling or Malling Merton. Rootstock names consist of the the type name or its abbreviation followed by the selection number. For instance, a Budagovsky selection 118 would be seen as either Budagovsky 118, Bud 118 or B.118. They all refer to the same stock material. As mentioned above, EMLA  are virus-free rootstocks derived from a corresponding M or MM selection. They have comparable characteristics to the uncertified material, but tend to be 5-10% more vigorous.

Guowang Product Page

Description of commonly available apple rootstocks

B.9 (Bud.9, Budagovsky 9)

Dwarfing rootstock resulting from a cross between M.8 x &#;Red Standard&#; (Krasnij Standard) from Russia. B.9 has been widely tested and is used commercially throughout the U.S. It is slightly more dwarfing than M.9 and is slightly more productive. Other traits of note: Very early precocity; very winter hardy; little suckering; requires support; adapted to well drained soil; very resistant to crown rot; more fireblight resistant than M.9.

B.118 (Bud.118, Budagovsky 118)

Semi-Dwarf rootstock resulting from a cross between &#;Moscow pear&#; x M.9 or M.8 from the former Soviet Union. It is reported to be very cold hardy, and produces a tree size of about 85% of seedling. It is more precocious than the seedling and can be grown without support. It is moderately resistant to fire blight and crown rot. Other traits of note: Somewhat early bearing; moderately productive; well anchored; needs well drained soils.

EMLA 7

 A tree on this rootstock will be 50 to 60 percent smaller than a standard tree. Trees on this clone are the most popular of all the rootstock grown. EMLA 7 does well on most soils. Some support may be needed in early years. EMLA 7 is very winter hardy. It is susceptible to suckering. EMLA 7 is extremely tolerant to fire blight.

EMLA 26

This rootstock is considered to be smaller than a half size tree. It is about 40 to 45 percent of a standard tree, needs some support in early years, but could be self-supporting in later years. EMLA 26 is very early and heavy bearing. This rootstock is very adaptable for close plantings and double rows.

EMLA 106

This rootstock produces a tree about half to two-thirds the size of a standard tree. It does not sucker and the rootstock is resistant to woolly aphid. EMLA 106 has been planted intensively in the East and West and is an excellent producer. It should be planted on well-drained soil as it is susceptible to crown rot.

EMLA 111

This rootstock produces a tree about two-thirds the size of a standard tree. Vigorous scion varieties and better soils may grow to three-quarter size or larger. EMLA 111 is a good producing rootstock, is well anchored and tolerant of drought conditions. It is widely adapted to most soil conditions.

M.26 (Malling 26)

Semi-dwarfing rootstock resulting from a cross between M.16 and M.9 made at the East Malling Research Station, Maidstone, Kent, England. M.26 is traditionally considered a dwarf rootstock, but is one of the more vigorous dwarfing rootstocks. M.26 is grown widely throughout the world and is included as a &#;standard&#; in many rootstock trials. M.26 is precocious and very productive, produces many burr knots, and is susceptible to crown rot and fire blight. In a joint effort to produce virus free rootstocks, the East Malling and Long Ashton Stations in England used heat treatments to eliminate known viruses and released M.26 EMLA. In general, virus-free rootstocks are slightly more vigorous than the original that contains viruses. Other traits of note: &#;  very early bearing; good productivity; may need support early on; winter hardy; prefers well drained soil; little suckering; susceptible to woolly apple aphid.

G.11 (Geneva 11®)

Semi-dwarfing rootstock resulting from the cross between M.26 x Robusta 5 crabapple and introduced by the New York State Agricultural Experiment Station, Geneva NY. G.11 produces trees of similar size to M.26 and is equally precocious. G.11 is moderately resistant to fire blight; moderately susceptible to woolly apple aphid and crown rot. Other traits of note: requires support in early years; produces few burr knots and root suckers; and it&#;s well adapted to most soils.

G.41 (Geneva 41®)

Dwarfing rootstock resulting from a cross between M.27 x Robusta 5 crabapple and introduced by the New York State Agricultural Experiement Station, Geneva NY.  G.41 is highly resistant to fire blight and phytophthora and it appears to be tolerant of replant disease (ARD). Other traits of note: Early bearing; winter hardy; very little suckering, requires tree support.

G.890 (Geneva 890®)

Semi-dwarfing rootstock; intermediate precocity; winter hardy; does not require support.

G.935 (Geneva 935®)

Dwarfing rootstock resulting from a cross between Ottawa 3 x Robust 5 crabapple and introduced by the New York State Agricultural Experiment Station, Geneva NY. Traits of note include: early bearing; winter hardy; moderate suckering; requires support; very productive; well adapted to most soils; highly resistant to crown rot; highly resistant to fireblight.

M9-337 (Malling 9 virus certified #337)

Dwarfing rootstock selection from M.9 virus-free certified clonal stock developed by the East Malling Research Station, Maidstone, Kent, England. M.9-337 is very precocious and tolerant to a wide range of soil and climatic conditions. Tree support is required.

Nic 29®, RN 29 cv. (Malling 9 selection NIC 29)

A dwarfing rootstock selection of M.9 made in Belgium and is slightly more vigorous than other M.9 selections such as M.9-337. M.9 is used to impart vigor to cultivars such as Empire or Honeycrisp.

Supporter 4&#;

A cross of M.9 x M.4, Supporter 4&#; is a dwarfing apple rootstock similar in vigor to EMLA 26. Anchorage is similar to EMLA 26, and trees on this root should be grown with some sort of support structure. The rootstock is relatively frost resistant. In tests, Supporter 4&#; showed better efficiency than both EMLA 26 and EMLA 106.

This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

The &#;Huangguan&#; pear disease spot detection and grading is the key to fruit processing automation. Due to the variety of individual shapes and disease spot types of &#;Huangguan&#; pear. The traditional computer vision technology and pattern recognition methods have some limitations in the detection of &#;Huangguan&#; pear diseases. In recent years, with the development of deep learning technology and convolutional neural network provides a new solution for the fast and accurate detection of &#;Huangguan&#; pear diseases. To achieve automatic grading of &#;Huangguan&#; pear appearance quality in a complex context, this study proposes an integrated framework combining instance segmentation, semantic segmentation and grading models. In the first stage, Mask R-CNN and Mask R-CNN with the introduction of the preprocessing module are used to segment &#;Huangguan&#; pears from complex backgrounds. In the second stage, DeepLabV3+, UNet and PSPNet are used to segment the &#;Huangguan&#; pear spots to get the spots, and the ratio of the spot pixel area to the &#;Huangguan&#; pear pixel area is calculated and classified into three grades. In the third stage, the grades of &#;Huangguan&#; pear are obtained using ResNet50, VGG16 and MobileNetV3. The experimental results show that the model proposed in this paper can segment the &#;Huangguan&#; pear and disease spots in complex background in steps, and complete the grading of &#;Huangguan&#; pear fruit disease severity. According to the experimental results. The Mask R-CNN that introduced the CLAHE preprocessing module in the first-stage instance segmentation model is the most accurate. The resulting pixel accuracy (PA) is 97.38% and the Dice coefficient is 68.08%. DeepLabV3+ is the most accurate in the second-stage semantic segmentation model. The pixel accuracy is 94.03% and the Dice coefficient is 67.25%. ResNet50 is the most accurate among the third-stage classification models. The average precision (AP) was 97.41% and the F1 (harmonic average assessment) was 95.43%.In short, it not only provides a new framework for the detection and identification of &#;Huangguan&#; pear fruit diseases in complex backgrounds, but also lays a theoretical foundation for the assessment and grading of &#;Huangguan&#; pear diseases.

It can effectively solve the problem of inaccurate grading of &#;Huangguan&#; pears caused by manual sorting, which is time-consuming and laborious and easy to distract. It provides a new idea for the automatic grading of &#;Huangguan&#; pear appearance quality.

A method for grading the severity of &#;Huangguan&#; pear disease was proposed. By calculating the ratio of the area of diseased spots to the area of &#;Huangguan&#; pear fruit, it provides technical support for the accurate classification of the appearance quality of &#;Huangguan&#; pear in actual production.

Pears are fruits produced and consumed around the world, growing on a tree and harvested in the Northern Hemisphere in late summer into October. The pear tree and shrub are a species of genus Pyrus, in the family Rosaceae, bearing the pomaceous fruit of the same name ( Ikinci et al., ). Several species of pears are valued for their edible fruit and juices, while others are cultivated as trees. China is the world&#;s largest producer and consumer of pears, and its pear cultivation area and output rank first in the world ( Oyom et al., ). &#;Huangguan&#; pear is a mid-early mature pear variety cultivated by China. It has the advantages of large fruit size, high quality, early fruit, and good yield. It can meet the demand for high-quality pears in the fruit market. After years of demonstration and promotion, &#;Huangguan&#; pear has become one of the main pear tree varieties in most regions, providing huge economic benefits for &#;Huangguan&#; pear producers and exporting countries. It is worth emphasizing that the economic value of &#;Huangguan&#; pear fruit depends to a large extent on the aesthetics of its appearance. The best-looking fruits are for export, the less diseased ones are reserved for domestic consumption, and the worst ones are used for further processing to make canned fruits or jams. However, the quality grading of &#;Huangguan&#; pear is a time-consuming and laborious process. So far, it has almost completely relied on human inspection and manual observation of disease symptoms to judge the grade of &#;Huangguan&#; pear. This method is costly and has highly subjective and low efficiency and timeliness. However, early automated grading systems have extensively utilized image processing algorithms and relied on manually defined image features to build classifiers ( Suykens, ; Zeng et al., ), limiting the robustness and generalization ( Xu and Mannor, ) of detection performance due to the variance of fruits types, appearances, and damage defects.

If you want to learn more, please visit our website Crown Pear Export.

2 Materials and methods

2.1 Data set production and processing

The data set used in this article has a total of images of &#;Huangguan&#; pear. Taking into account the diversity of lighting conditions in practical applications, The data was collected in three different periods from July to December : In the morning (8:30&#;10:00), noon (12:30&#;14:00) and afternoon (15:30&#;17:00) in the laboratory with camera. This leads to problems such as background noise, distance, location, and lighting conditions of &#;Huangguan&#; pear. It is the existence of these problems that can improve the generalization ability of the model in different scenarios and improve the robustness of the model. Part of the &#;Huangguan&#; pear image is shown in . According to the &#;Huangguan&#; pear samples displayed in the data set, the identification and segmentation of &#;Huangguan&#; pear fruit disease mainly have the following difficulties: 1) &#;Huangguan&#; pear background interferes with segmentation, and the different brightness of &#;Huangguan&#; pear imaging caused by factors such as light can easily be mistaken for disease; 2) &#;Huangguan&#; pear disease are irregular in shape, some are small, and the initial disease are difficult to detect with the naked eye, which increases the difficulty of disease segmentation; 3) &#;Huangguan&#; pear have different shooting backgrounds, and the quality of the background processing directly affects the classification of &#;Huangguan&#; pear.

2.2 Image data enhancement

The sample distribution of each type of disease in the data set is not uniform, and the limited training data is easy to overfit the deep learning model. In deep learning, the use of data augmentation methods to expand the data can improve the generalization ability of the model. The training data of this study uses the Image Data Generator online enhancement method under the Keras framework. That is, an enhancement method is randomly selected for each batch of data during the training process, without increasing the number of original data sets. In order to avoid changing the original data characteristics and better simulate the differences of samples under real shooting conditions, the training set of this research mainly adopts the following data enhancement methods: 1) Flip: Flip the image vertically to simulate the randomness of the shooting angle when the sample is collected, and will not change the shape of the diseased spot and the distribution of the diseased spot on the leaf. 2) Color jitter: Change the brightness of the image to randomly jitter between 0.8-1.2 times. Change the contrast of the image to randomly jitter between 0.6-1.6 times. Change the chromaticity of the image to jitter randomly between 0.7-1.4 times. Simulate lighting differences and ensure that the parameters conform to the actual shooting conditions to avoid image distortion. 3) Add noise: Add salt and pepper noise with a signal-to-noise ratio of 0.95 to the image to simulate the noise generated during the shooting process and weaken the high-frequency features to prevent the model from overfitting. The result of data enhancement is shown in .

2.3 Labeling of diseased spots of &#;Huangguan&#; pear fruit

To train the disease segmentation model, the disease need to be marked as shown in . The labeling of &#;Huangguan&#; pear disease is time-consuming and laborious, with a large number of small targets. The finer annotations help Mask R-CNN and DeepLabV3+ to perform finer segmentation of &#;Huangguan&#; pears and disease, laying the foundation for the classification of &#;Huangguan&#; pears. The labeling is divided into three scenes including background, pear and diseased spots, and labeling is carried out with LabelMe (Russell et al., ), an image semantic segmentation labeling tool.

2.4 Grading method for the severity of fruit diseases of &#;Huangguan&#; pear

The classification of disease severity is the basis for formulating prevention and control strategies. Three methods are usually used in practice. The first method is to calculate the ratio of the number of infected fruits per unit area to the total number of fruits. The second method is to calculate the ratio of the number of diseased fruits to the total number of fruits on the same plant. The third method is to calculate the ratio of the area of spots on the same fruit to the total area of the fruit. The third method is the basis for accurately estimating the severity of crop diseases in a region. Therefore, we used the third method, which uses the ratio of the spot area to the total area of the same fruit as the basis for classification of disease severity. This method is mainly based on the opinions and practical experience of fruit farmers who have been engaged in fruit grading for many years.

By calculating the ratio of the area of the diseased spot to the area of the fruit, the severity of the disease of &#;Huangguan&#; pear was classified. Since the &#;Huangguan&#; pear fruit to be divided is located in a complex background, the target &#;Huangguan&#; pear fruit and diseased spots are easily confused with other similar elements, resulting in over-segmentation or under-segmentation. Therefore, it is difficult to accurately segment &#;Huangguan&#; pear fruit and diseased spots at the same time using a single-stage network. In order to ensure the accuracy of disease segmentation, the &#;Huangguan&#; pear fruit in the complex background should be segmented first. Therefore, this study uses a two-stage segmentation network to classify the severity of &#;Huangguan&#; pear diseases, and classifies the &#;Huangguan&#; pear images according to the first, second and third levels. Specific steps are as follows. In the first stage, the segmentation target is the &#;Huangguan&#; pear fruit and the complex background. The mask image obtained from the test is used to extract the &#;Huangguan&#; pear fruit from the complex background, so as to obtain the &#;Huangguan&#; pear fruit in the simple background. In the second stage of segmentation, the diseased spots in the &#;Huangguan&#; pear fruit are taken as the target, and the proportion of the diseased spots in the &#;Huangguan&#; pear fruit is obtained. As the basis for the classification of disease severity of &#;Huangguan&#; pear. The formula is shown in formula (1).

P=SDiseaseSPear

(1)

Among them, S Pear represents the fruit area of &#;Huangguan&#; pear after segmentation; S Disease represents the area of the disease after segmentation; P represents the proportion of diseased spots on &#;Huangguan&#; pear fruit.

After calculating the area of &#;Huangguan&#; pear by the disease, refer to the &#;Huangguan&#; Pear Fruit Grade&#; DB 13/T &#; issued by China. According to local standards, the proportion of fruit diseases can be divided into three grades: good and bad. Among them, 0% of diseases are first-class fruits, 2% or less are second-class fruits, and diseases greater than 2% are third-class fruits.

2.5 Evaluation index

In order to reasonably evaluate the performance of the model, the first two segmentation stages of this study used 3 commonly used evaluation indicators: Pixel Accuracy (PA), dice and Intersection over Union (IoU). The pixel accuracy is the ratio of all correctly classified pixels to the total pixels, as shown in formula (2):

RPA=Σi=0kpiiΣi=0kΣj=0kpij

(2)

In the formula, k is the number of categories, pii is the number of pixels that are correctly predicted, and p ij represents the number of pixels whose category i is predicted to be category j

The Dice coefficient is a function that measures the similarity of two sets, and is one of the most commonly used evaluation indicators in semantic segmentation. As shown in formula (3):

Rdice=2|X&#;Y||X|+|Y|

(3)

Where X is the predicted pixel and Y is the ground truth.

The intersection ratio is the ratio of the intersection and union of a certain type of prediction result and the true value of the model. The intersection ratio is the most commonly used evaluation index in semantic segmentation, and the expression is shown in formula (4):

RIoU=A&#;BA&#;B

(4)

When the value of IOU is between 0 and 1, it represents the degree of overlap of the two boxes. The higher the value, the higher the degree of overlap.

The third grading stage uses 5 evaluation indicators commonly used in grading models, recall, precision, average precision (AP), F1 score and speed. Recall is the ratio of the number of correctly detected targets to all actual targets (Equation (5)). Precision is the number of correctly detected targets in all detected targets The ratio of (Equation (6)). F1 is the harmonic average of precision and recall (Equation (7)).

Recall(R)=TPTP+FN

(5)

Precision(P)=TPTP+FP

(6)

F1=2×Pre×RecPre+Rec

(7)

2.6 Model training

The hardware configuration used for training and testing in this research is as follows: Intel(R) Core(TM) i5-F CPU @ 2.90GHz, 16G RAM, NVIDIA GeForce GTX SUPER graphics card, 64-bit Windows 10 operating system, CUDA version 10.0 and TensorFlow version 1.13.2. In order to avoid the influence of hyperparameters on the experimental results, the hyperparameters of each network are uniformly configured. After trial and error, the following hyperparameters have been determined: The learning rate is 1e-4, the epochs is 50, and the batch size is 16. If training for more than 5 generations does not further improve the accuracy, start early stopping and stop training.

The company is the world’s best Snow Pear Exporter supplier. We are your one-stop shop for all needs. Our staff are highly-specialized and will help you find the product you need.