A dynamic growth inversion method for cherry fruit

Abstract

The application discloses a cherry fruit dynamic growth inversion method, which remotely monitors fruit growth and development, obtains the area of the fruit at each stage of the coloring period when the cherry fruit enters the coloring period, takes the coloring days as the independent variable and the area as the dependent variable, researches the growth trend and rhythm of the cherry fruit, fits the dynamic growth curve of the fruit through an inversion method, further gives the growth rule of the fruit, and finally constructs a cherry fruit dynamic growth inversion method; after the growth curve model is constructed, the reliability of the model is verified by using accurate factors and deviation factors. The application can realize rapid and accurate detection of the growth state of the fruit in a natural scene, dynamically master the growth rule of the greenhouse crops, and further improve the yield and quality of the cherry fruit.

Description

Technical Field

[0001] This invention relates to the field of greenhouse cherry cultivation technology, specifically to a method for dynamic growth inversion of cherry fruits. Background Technology

[0002] my country is a major agricultural country, and the intelligentization of agricultural production, the digitalization of agricultural management, and the precisionization of agricultural services are inevitable development trends. In recent years, my country's fruit industry has developed particularly rapidly, with planting area and output expanding rapidly. The fruit industry has formed a scale advantage and continues to grow. With the establishment of agricultural information systems and the increase in large amounts of agricultural data, manual data acquisition and analysis are becoming increasingly difficult. Therefore, data analysis technology has been effectively applied in agriculture, providing farmers with timely, predictive, seasonal, and guiding practical information. This has enabled a leap from traditional extensive planting and management models to modern intelligent and scientific methods, contributing to the rapid development of traditional agriculture towards smart agriculture.

[0003] Fruit growth and development is a complex process, exhibiting non-linear and abrupt changes over time under the influence of various environmental conditions and random factors. However, cherry fruits are neither drought-resistant nor waterlogging-resistant, making them highly sensitive to water availability; timely irrigation is crucial. Cherry fruit development can be divided into three stages: the first rapid growth stage, the pit-hardening stage, and the second rapid growth stage. The second rapid growth stage is the most vigorous, with the most sensitive water supply, determining fruit yield and quality. During this period, ovary cell division is rapid, and cells expand quickly; the size of the fruit at harvest depends primarily on the degree of development during this rapid growth stage. In recent years, the promotion of sweet cherries has accelerated, but systematic research on the dynamics of fruit growth and development is lacking. Environmental and climatic changes and imbalances in water and fertilizer management not only affect fruit growth and development but also fruit quality, leading to fruit drop, cracking, and deformed fruit.

[0004] Understanding the growth dynamics of cherries from fruit set to ripening is crucial for farmers to ensure proper water and fertilizer supply and aquatic ecosystem management. However, monitoring cherry growth requires measuring the size of the fruit at each stage, which is a time-consuming manual task. Summary of the Invention

[0005] The purpose of this invention is to provide a method for dynamic growth inversion of cherry fruits, enabling rapid and accurate detection of fruit growth status in natural settings, dynamically understanding the growth patterns of greenhouse crops, and thereby improving the yield and quality of cherry fruits.

[0006] To achieve the above objectives, this application proposes a method for dynamic growth inversion of cherry fruits, comprising:

[0007] Obtain images of cherry fruits during the coloring stage;

[0008] The cherry fruit images are preprocessed to make them more consistent with the original shape of the cherry fruit;

[0009] The projected area of ​​cherry fruit was obtained by improving the intelligent scissors algorithm;

[0010] The dynamic growth curve of the fruit was fitted using the inversion method.

[0011] Furthermore, to obtain images of cherry fruits during the coloring stage, the camera is mounted on a support frame at a fixed height, with the shooting distance controlled between 80 and 100 cm, and the shooting angle controlled between -15° and 90° vertically and 360° horizontally. The focal length of the camera lens is then adjusted to obtain images of cherry fruits during the coloring stage.

[0012] Furthermore, the cherry fruit image is preprocessed to make it more consistent with the original shape of the cherry fruit. Specifically, a 3×3 kernel is used to perform a closing operation on the cherry fruit image to fill in the concave edges or the hollow center, so that the segmented cherry fruit image is more consistent with the original shape of the cherry fruit.

[0013] Furthermore, the projected area of ​​the cherry fruit is obtained by improving the intelligent scissors algorithm, specifically as follows:

[0014] First, obtain a cherry fruit image in a natural environment obtained by closing operation. Use the Canny algorithm to detect the image edges and find candidate circle centers. Determine the radius based on the support of non-zero pixels on the edges of all candidate circle centers and obtain the optimal radius r. Extract the target contour and obtain the initial edge image by using the gradient Hough transform circle detection method. The obtained initial edge points are denoted as set C.

[0015] Using the initial edge point set obtained by gradient Hough transform circle detection as a prerequisite, and combining computer vision technology, semantic segmentation is performed pixel by pixel on the key feature regions of cherry fruit. The key feature regions of cherry fruit images in complex environments are adaptively searched, and the area of ​​the key feature regions is obtained, that is, the projected area of ​​the cherry fruit image during the coloring period. The area of ​​each fruit is measured in pixels.

[0016] Furthermore, the dynamic growth curve of the fruit is fitted using the inversion method, specifically as follows:

[0017] Let y be the projected area of ​​the cherry fruit image during the coloring period, and x be the number of days in the coloring period. Form a data pair x and y (x... i ,y i ), where i = 1, 2, ..., m, m is the total number of data pairs, and x is the number of each pair. i They are different;

[0018] Fit an analytical expression that is adapted to the dynamic growth law of fruit, namely y = f(x), to reflect the dependence between x and y. f(x) is called an inversion model, which is to "best" approximate or fit the known data in a certain sense.

[0019] The inversion model is based on the data (x) i ,y i To obtain an nth-degree polynomial, where n ≤ m:

[0020]

[0021] in, The fitting curves for the dynamic growth of the fruit, a0, a1…a n These are randomly selected variables.

[0022] Furthermore, the fruit dynamic growth fitting curve The method of obtaining it is:

[0023] Randomly selected variables a0, a1…a n Use a counter (count) to record the number of points falling on the fitted curve. This indicates that the number of points falling on the fruit dynamic growth fitting curve is small, resulting in a poor fitting effect. In this case, the selected variables should be discarded and the search should continue.

[0024] Let the step size of the variable be S0 = 0.2, S n =tan2°, substitute the data pairs formed by x and y into the following formula:

[0025]

[0026] When n=1, And so on; when the deviation between the observed value (x', y') and the fitted curve is large, n = n + 1, and the above formula is iteratively substituted until... When the data pair has essentially fallen onto the dynamic growth fitting curve of the fruit, it indicates that the data pair has basically fallen onto the curve.

[0027] Furthermore, by using the fruit dynamic growth fitting curve acquisition method, growth curves of Logistic, Logarithmic, Quadratic, Cubic, and Linear fitting models were obtained respectively, and the curve model with the highest degree of agreement with the actual observation value was determined through analysis.

[0028] Furthermore, select parameter R 2 The R-squared value is used to measure how well the fruit dynamic growth curve fits the observed values. 2 The larger the value of R, the better the fit. 2 The closer the value is to 1, the better the fitted curve fits the observed values;

[0029]

[0030] Where y is the observed value, For the mean, These are the fitted values.

[0031] Furthermore, after obtaining the curve model with the highest fit, the fitted value is obtained using a regression equation with the coloring period x as the dependent variable. Using the accuracy factor M f Sum of deviation factors N f To verify the reliability of the model, the closer the deviation factor is to 1, the smaller the fluctuation range between the fitted value and the observed value, thus proving the reliability of the curve model.

[0032]

[0033]

[0034] Compared with the prior art, the above technical solution adopted in this invention has the following advantages: This invention first remotely collects images of cherry fruits at each stage, promoting the automated monitoring of fruit growth. Under the premise of correctly obtaining the projected area of ​​the fruit during the coloring stage, it effectively derives the fruit growth curve and displays the dynamic growth pattern of the fruit, so as to take reasonable cultivation measures and manage water and fertilizer more accurately, thereby improving the yield and quality of cherry fruits. Attached Figure Description

[0035] Figure 1 This is a fruit segmentation effect based on computer vision technology.

[0036] Figure 2 A diagram showing the growth trend and growth rhythm of the fruit;

[0037] Figure 3 A curve fitting the dynamics of fruit growth. Detailed Implementation

[0038] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of this application and are not intended to limit the application; that is, the described embodiments are only a part of the embodiments of this application, and not all of them.

[0039] Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of the application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0040] Example 1

[0041] This invention remotely monitors the growth and development of cherry fruits. When cherry fruits enter the coloring stage, the area of ​​the fruit at each stage of coloring is obtained. The number of days of coloring is used as the independent variable, and the area as the dependent variable to study the growth trend and rhythm of cherry fruits. Furthermore, a dynamic growth curve of the fruit is fitted using an inversion method to further reveal the fruit growth pattern, ultimately constructing a dynamic growth inversion method for cherry fruits. After constructing the growth curve model, an accuracy factor A is used... f Sum of deviation factors B f Verify the reliability of the model. Specific implementation methods include:

[0042] Step 1: Obtain images of cherry fruits during the coloring stage;

[0043] Specifically, such as Figure 1 As shown, the fruit images were taken in a natural environment in the cherry-growing area of ​​Dalian City, Liaoning Province. The variety photographed is Meizao, and the target is fruit in the coloring stage. Considering that the size of the fruit displayed in the image varies at different shooting distances, which may have a certain impact on the projected area of ​​the target fruit, the camera was mounted on a support frame at a fixed height. The shooting distance was controlled between 80 and 100 cm, and the shooting angle was controlled within the range of -15° to 90° vertically and 360° horizontally. The focal length of the camera lens was adjusted to ensure that the acquired fruit images were clear. The captured images are automatically uploaded to the server daily.

[0044] Step 2: Preprocess the cherry fruit image to make it more consistent with the original shape of the cherry fruit;

[0045] Specifically, because cherries ripen at different times, the segmented fruits may have concave edges or hollow centers. Therefore, a closing operation is performed to smooth the image. A 3×3 kernel is used to perform the closing operation on the cherry fruit image to fill in the concave edges or hollow centers, making the segmented image more consistent with the original shape of the cherry fruit.

[0046] Step 3: Obtain the projected area of ​​the cherry fruit by improving the intelligent scissors algorithm;

[0047] Specifically, first, an image of cherry fruit in a natural environment obtained through closing operations is acquired. The Canny algorithm is used to detect the image edges and identify candidate circle centers. The radius is determined based on the degree of support of the non-zero pixels of the edges of all candidate circle centers, and the optimal radius r is obtained based on the circle center (X0, Y0). Next, the HoughCircles() function is used to perform gradient Hough change circle detection on the cherry fruit image, extract the target contour, and obtain the initial edge image. The obtained initial edge points are denoted as set C. Since the study focuses on a specific cluster of fruits on a tree, whose position is fixed, certain constraints are set on the horizontal and vertical coordinates to avoid interference from other fruits. The horizontal coordinate is controlled to be 600-700 pixels, and the vertical coordinate is controlled to be 200-300 pixels.

[0048] The initial edge point set C obtained from Hough circle detection is used as a prerequisite for the improved intelligent scissors algorithm to adaptively find key feature regions in cherry fruit images under complex environments. Simultaneously, a weight map is established, with weights ω in the improved intelligent scissors algorithm. Z ω D ω G The frozen path threshold is an adjustable parameter. This application finds that setting the weight to ω... Z =0.1, ω D =0.3, ω G =0.1, and the frozen path threshold is 28, at which point the algorithm performs optimally. Therefore, for a color image of a fruit, the local cost from one pixel p to another pixel q is calculated as follows:

[0049] I(p, q) = 0.1 × f z (q)+0.3×f D (p, q) + 0.1 × f G (q)

[0050] The above method can automatically acquire and analyze fruit images in agricultural scenarios with complex canopy structures, effectively obtaining the projected area of ​​the fruit during the coloring stage, i.e. the size of the fruit, and can provide a good understanding of various dynamics during fruit growth.

[0051] Step 4: Fit the dynamic growth curve of the fruit using the inversion method;

[0052] Specifically, taking the "Meizao" cherry variety from Dalian, Liaoning Province as an example, cherry fruit growth is divided into three stages: the first rapid growth period, the pit hardening period, and the second rapid growth period. By remotely monitoring the dynamic growth of the fruit, computer vision is used to segment the key feature regions of the fruit, i.e., the fruit's growth area, to obtain the growth area value of the fruit during the coloring period. Using the number of days the fruit changes color as the x-axis and the periodically obtained projected fruit area as the y-axis, a data pair of x and y is formed (x...y ... i ,yi (i = 1, 2, ..., m), the inversion method is used to fit quadratic, cubic, and logistic curves, and to plot cherry fruit growth curves and growth rhythm diagrams, such as... Figure 2 As shown;

[0053] from Figure 2 The growth rhythm diagram shows that the size of "Meizao" cherry fruit exhibits a slow-fast-slow growth trend throughout the coloring period. The growth curve shows that "Meizao" cherry fruit size has one main growth peak during the coloring period, which is mid-April. The fruit growth is the largest during this period, and it is the main growth stage for fruit enlargement during the entire coloring period. There are also minor growth peaks in late March and late April. Fruit growth is relatively slow at the beginning of the coloring period, accelerates in mid-April, and gradually enters a slow growth stage in late April. This indicates that the fruit enlargement of "Meizao" cherry is mainly concentrated in the first half of April, during which the fruit growth accounts for 81.88% to 93.74% of the fruit growth during the coloring period.

[0054] An analytical expression adapted to the dynamic growth pattern of fruit was fitted using the inversion method, as shown in Table 1. The growth curves of the Logistic, Logarithmic, Quadratic, Cubic, and Linear fitting models were obtained by the fruit dynamic growth fitting curve acquisition method. As can be seen from Table 1, the F-value and Sig value of the dynamic fitting model are both highly significant, indicating that the five fitting models can describe the dynamic growth and development process of fruit well.

[0055] Table 1. Model of Fruit Growth Dynamics for "Meizao" Cherries

[0056]

[0057]

[0058] The expressions for the five fitting models are as follows:

[0059] Logistic

[0060]

[0061] Logarithmic

[0062] y = 22604.12lnx + 37880.55

[0063] Linear

[0064] y = 2702.31x + 54659

[0065] Quadratic

[0066] y = -2.08x 2 +2758.33x+54398.23

[0067] Cubic

[0068] y = -4.04x 3 +161.53x 2 +957.8x+58826.57

[0069] The coefficient of determination R of the cubic polynomial 2 The Cubic model outperformed the other four fitting models, indicating that Cubic's fit to the size of "Meizao" cherry fruits was superior to Logistic, Logarithmic, Linear, and Quadratic models. Figure 3 It can be seen that the Cubic curve for the growth of "Meizao" cherry fruit size matches the actual observations better. And combined with M... f and N f To verify the reliability of the cubic polynomial's fitting model, the accuracy factor is a measure of the average difference between observed and predicted values, while the bias factor represents the fluctuation range of the predicted values. The average accuracy factor M of the cubic polynomial is... f =1.02, deviation factor N f =1.0002, indicating that the fitted value of the cubic polynomial model fluctuates less. Therefore, using a cubic polynomial can better and more stably describe the growth pattern of "Meizao" cherry fruit during the coloring period.

[0070] The foregoing description of specific exemplary embodiments of the invention is for illustrative and explanatory purposes. These descriptions are not intended to limit the invention to the precise forms disclosed, and it will be apparent that many changes and variations can be made in accordance with the foregoing teachings. The exemplary embodiments were chosen and described in order to explain the specific principles of the invention and its practical application, thereby enabling those skilled in the art to implement and utilize various different exemplary embodiments of the invention, as well as various different choices and variations. The scope of the invention is intended to be defined by the claims and their equivalents.

Claims

1. A method for dynamic growth inversion of cherry fruit, characterized in that, include: Obtain images of cherry fruits during the coloring stage; The cherry fruit images are preprocessed to make them more consistent with the original shape of the cherry fruit; The projected area of ​​cherry fruit was obtained by improving the intelligent scissors algorithm; The dynamic growth curve of the fruit was fitted using the inversion method; The projected area of ​​the cherry fruit was obtained by improving the intelligent scissors algorithm, specifically as follows: First, obtain a cherry fruit image in a natural environment obtained through closing operations. Then, use the Canny algorithm to detect image edges and identify candidate circle centers. Determine the radius based on the support of all candidate circle centers for their non-zero edge pixels, obtaining the optimal radius *r*. Finally, extract the target contour using the gradient Hough transform circle detection method to obtain the initial edge image. The obtained initial edge points are denoted as the set. ; Using the initial set of edge points obtained by gradient Hough transform circle detection as a prerequisite, and combining computer vision technology to perform semantic segmentation of key feature regions of cherry fruit pixel by pixel, the key feature regions of cherry fruit image are adaptively found, and the area of ​​the key feature regions is obtained, that is, the projected area of ​​cherry fruit image during the coloring period. The dynamic growth curve of the fruit is fitted using the inversion method, specifically as follows: Let y be the projected area of ​​the cherry fruit image during the coloring period, and x be the number of days in the coloring period. Form a data pair of x and y. , ), where i = 1, 2, ..., m, and m is the total number of data pairs. They are different; To fit an analytical expression that adapts to the dynamic growth pattern of fruit, To reflect the dependency relationship between x and y, This is called an inversion model; The inversion model is based on data ( , ), obtain an nth degree polynomial, where n m: in, To fit the dynamic growth curve of the fruit, The variables are randomly selected. Fruit dynamic growth fitting curve The method of obtaining it is: Random selection Use a counter (count) to record the number of points falling on the fitted curve. This indicates that the number of points falling on the fruit's dynamic growth fitting curve is small, resulting in a poor fitting effect. In this case, the selected points should be discarded. Continue the search; Let the variable step size be... , Substitute the data pairs formed by x and y into the following formula: When n=1, And so on; when the deviation between the observed value (x', y') and the fitted curve is large, Repeat the above formula until... When the data pair has essentially fallen onto the dynamic growth fitting curve of the fruit; By using the fruit dynamic growth fitting curve acquisition method, the growth curves of Logistic, Logarithmic, Quadratic, Cubic and Linear fitting models were obtained respectively, and the curve model with the highest degree of agreement with the actual observation value was determined through analysis. After obtaining the curve model with the highest matching degree, the coloring period is used. Using the regression equation as the dependent variable, its fitted value is obtained. and using accuracy factors Sum of deviation factors To verify the reliability of the model, the closer the deviation factor is to 1, the smaller the fluctuation range between the fitted value and the observed value, thus proving the reliability of the curve model. 。 2. The method for dynamic growth inversion of cherry fruit according to claim 1, characterized in that, To obtain images of cherry fruits during the coloring stage, the camera is mounted on a support frame at a fixed height, and the shooting distance is controlled at 80 mm. Within 100cm, the shooting angle should be controlled between -15° and 90° vertically, and horizontally... Rotate within this range, adjust the camera lens focal length, and acquire images of cherry fruits during the coloring stage.

3. The method for dynamic growth inversion of cherry fruit according to claim 1, characterized in that, The cherry fruit image is preprocessed to better reflect the original shape of the cherry fruit. Specifically, 3D processing is used. The kernel of step 3 performs a closing operation on the cherry fruit image, filling in the concave edges or the hollow center, so that the segmented cherry fruit image is more consistent with the original shape of the cherry fruit.

4. The method for dynamic growth inversion of cherry fruit according to claim 1, characterized in that, Select parameters The degree of fit between the fruit dynamic growth fitting curve and the observed values ​​is measured by: The larger the value, the better the fit. The closer the value is to 1, the better the fitted curve fits the observed values; Where y is the observed value, For the mean, These are the fitted values.

Images