Method for evaluating light efficiency of fruit tree canopy and application thereof

Abstract

The present application relates to a kind of fruit tree crown layer light effect evaluation method.The present application constructs tree crown layer light distribution prediction model, and its steps include: S1, obtain tree crown layer point cloud data and each layer leaf density data;S2, obtain the relative light intensity data of each grid of the fruit tree crown layer;S3, calculate crown layer effective leaf area;S4, according to the layer relative light intensity of the fruit tree crown layer and effective leaf area, establish the light distribution prediction model of the fruit tree crown layer.The light distribution of fruit tree crown layer can be quickly and accurately calculated using the above prediction model, which helps to improve the yield of fruit trees.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of fruit tree canopy light distribution of laser radar point cloud technology, in particular to a fruit tree canopy light efficiency evaluation method and application. BACKGROUND

[0002] Pear tree is one of the largest fruit tree species in cultivation area and yield in China, and pear tree planting is an important part of agricultural production. The pear tree canopy is dense and complex in structure due to the interlaced branches and leaves, resulting in differences in light distribution of the canopy. Therefore, suitable canopy structure has an important influence on the yield, growth and pest control of fruit trees. For the study of the canopy, three-dimensional point cloud technology is used to construct a canopy model to determine the relationship between canopy point cloud, growth characteristics and canopy light distribution, build an evaluation system based on fruit tree canopy light calculation, and realize automatic evaluation of pear tree light efficiency, providing quantitative basis and technical support for fruit tree shaping, high-quality production and scientific management.

[0003] Laser radar is a commonly used device for detecting the geometric parameters of fruit tree canopy and is applied to digital orchard management. Laser radar is a non-contact way to quickly obtain high-density and high-precision three-dimensional point cloud data, which does not damage the research object and provides better data support for subsequent research work. In the case of poor orchard environment, handheld laser radar is more suitable for use because it is a handheld device, and the point cloud data collection is synchronized with the movement of the operator, which can realize the reconstruction of the three-dimensional map of the scene and can be used in various complex situations with high precision.

[0004] The existing canopy light efficiency evaluation system generally has the problem of low accuracy of light efficiency prediction. The current traditional canopy light research is mainly based on three-dimensional digitizer, which first reconstructs the tree into a three-dimensional model, and then analyzes the light interception rate of the canopy using a sensor, but the reconstructed three-dimensional model of the tree has low precision and large error compared with the actual model, lacks description of the details of the canopy, and the three-dimensional digitizer is expensive and complex to operate. Other researchers generally only use ellipses, circles and ovals to construct pear tree canopy geometric models to determine the light radiation accumulation in different areas of the canopy, and do not study the internal detailed layering of the canopy. Many current methods are still subject to various factors and have not yet produced an efficient and fast light evaluation method. SUMMARY

[0005] In view of the defects in the prior art, the purpose of the present application is to provide a fruit tree canopy light efficiency evaluation method. The method calculates the effective leaf area of the fruit tree canopy point cloud data, and establishes a mathematical model with high accuracy between the relative light intensity data, which is used to evaluate the tree canopy light effect, and can realize the light calculation and evaluation of the fruit tree canopy, and provide accurate basis for fruit tree pruning management.

[0006] To achieve the above purpose, the technical solution adopted by the present application is:

[0007] A fruit tree canopy light efficiency evaluation method, characterized in that it comprises the following steps:

[0008] Step 1: Obtain fruit tree point cloud data through a laser radar device;

[0009] Step 2: Obtain the light intensity of multiple points of the fruit tree canopy through a light sensor, and obtain the relative light intensity of each layer of the canopy;

[0010] Step 3: Preprocess the fruit tree point cloud data obtained in step 1 to remove irrelevant point clouds, and obtain fruit tree canopy leaf point cloud data;

[0011] Step 4: Horizontally layer the fruit tree canopy leaf point cloud data preprocessed in step 3 to obtain point cloud data and corresponding layer dimension reduction images of each layer of the canopy;

[0012] Step 5: Extract the point cloud data dimension reduction images of each layer of the canopy obtained in step 4, and obtain the effective leaf area of each layer of the canopy based on the leaf area calculation method of the Freeman chain code of multi-target area optimization combined with the traditional pixel point ratio method;

[0013] Step 6: Confirm that the effective leaf area obtained in step 5 has a positive linear correlation with the actual layer leaf area value, and construct a relative light intensity estimation model based on the effective leaf area of each layer of the canopy according to the relative light intensity of each layer of the canopy obtained in step 2;

[0014] Step 7: Use the estimation model obtained in step 6 to evaluate the light condition of the fruit tree canopy to be tested.

[0015] On the basis of the above scheme,

[0016] The specific method for obtaining the light intensity of multiple points of the fruit tree canopy in step 2 is:

[0017] Take the main stem of the fruit tree as the center of the canopy, extend the same distance l in four directions from the ground as fixed points, and perform spatial layering at a certain step h in the vertical height based on the ground to obtain a series of cubic space;

[0018] n sensors are arranged at the horizontal junction of each cubic space, each sensor is horizontally l away from the trunk, to collect the light intensity E in the area of different height, different level and different shielding condition ji (j = 1, 2, 3…n), while measuring the absolute light intensity E of each layer under the condition of no shielding Ji . According to the above series of light intensity points, the relative light intensity E of the i-th layer is obtained i , as shown in the following formula:

[0019]

[0020] On the basis of the above scheme,

[0021] The horizontal layering and dimension reduction method of the fruit tree canopy leaf point cloud data in step 4 is specifically:

[0022] Step 4-1, read the fruit tree canopy leaf point cloud data, divide the single fruit tree point cloud to obtain the canopy layering point cloud at different heights with a certain thickness according to the layering height data of the light point collection scheme, set the single layer thickness, plane normal vector and inclination angle;

[0023] Step 4-2, use RANSAC algorithm to fit the plane to obtain the plane model coefficient, respectively, the horizontal projection of each layering point cloud of the canopy on XOY plane is obtained, that is, the two-dimensional projection of each layering point cloud of the canopy, which is each layering dimension reduction image.

[0024] On the basis of the above scheme,

[0025] The canopy projection tracking calculation method in step 5 is specifically:

[0026] Step 5-1, pre-process each layering dimension reduction image obtained in step 4, the above pre-processing process includes: (1) using median filtering to remove isolated noise without destroying the boundary contour; (2) using morphological erosion and expansion to improve the continuity of the point cloud image in two-dimensional plane; (3) using Canny operator to outline the approximate contour of the target, providing the tracking path for the subsequent optimized Freeman chain code algorithm;

[0027] Step 5-2, using the Freeman tracking algorithm based on multi-target area optimization, respectively, the unlabeled boundary pixel points and the adjacent boundary pixel points are vector labeled until the vector boundary is closed, the number of closed boundary pixel points is calculated and the nul points in the appropriate labeling are deleted, and the above operation is repeated until there is no unlabeled boundary, and finally the pixel points are accumulated and summed to output the result;

[0028] Step 5-3, the layer leaf area S E is calculated by using the traditional pixel point ratio method, as shown in the following formula, wherein P bFor optimizing the Freeman chain code to calculate the total pixel points occupied by the fruit tree hierarchical projection plane, P t For the pixel points of the fruit tree hierarchical projection rectangle, S t For the actual area of the fruit tree hierarchical projection rectangle;

[0029]

[0030] On the basis of the above scheme,

[0031] The construction method of the relative light intensity estimation model in step 6 is specifically as follows:

[0032] Taking the hierarchical effective leaf area as the independent variable x and the layer relative light intensity value of each hierarchical layer of the crown layer as the dependent variable y, linear regression is used for regression analysis to obtain the distribution relationship between the hierarchical effective leaf area and the layer relative light intensity, as follows:

[0033] y = b + ax;

[0034] In the formula, a and b are regression equation coefficients.

[0035] Another object of the present application is to provide an application of the fruit tree crown layer light efficiency evaluation method.

[0036] To achieve the above object, the technical scheme adopted by the present application is as follows:

[0037] Step 1, selecting a target fruit tree of the same variety at the right time, collecting its point cloud data in a standardized manner, and obtaining the effective leaf area of each hierarchical layer of the crown layer according to steps 3 to 5 of claim 1;

[0038] Step 2, substituting the effective leaf area of each hierarchical layer obtained in step 1 into the distribution relationship between the hierarchical effective leaf area and the relative light intensity to obtain the corresponding relative light intensity prediction value and evaluate the light condition of the fruit tree crown layer.

[0039] The fruit tree crown layer light efficiency evaluation method and application have the following beneficial effects:

[0040] A reasonable point cloud crown layer analysis reconstruction method is constructed, the light receiving condition of the tree crown is introduced into the point cloud tree crown reconstruction algorithm, the relationship between the crown layer leaf area model and the light intensity is analyzed, and an accurate light efficiency analysis model is formed to realize the light calculation and evaluation of the fruit tree crown layer, provide accurate basis for fruit tree pruning, management and protection, and help improve the yield of fruit trees. BRIEF DESCRIPTION OF DRAWINGS

[0041] The present application has the following drawings:

[0042] Figure 1 The present application has the following drawings:

[0043] Figure 2The experimental schematic diagram II of the present application;

[0044] Figure 3 The flow chart of the method for calculating the effective area of the crown leaf projection according to the present application;

[0045] Figure 4 The flow chart of the light efficiency evaluation method according to the present application;

[0046] Figure 5 The analysis effect diagram of the embodiment of the present application.

[0047] Figure 6 S Ei Linear regression result of the actual layer effective leaf area DETAILED DESCRIPTION

[0048] The present application will be further described in detail below with reference to the accompanying drawings.

[0049] The present application takes pear trees as an example to specifically describe the method.

[0050] 1. As shown in Figure 1 , 2 , during the stable period of the pear tree canopy, a clear day without wind is selected, and the following operations are performed:

[0051] (1) Place the handheld laser radar device (3D-BOX three-dimensional SLAM laser scanner) at a position 10 cm higher than the head, and obtain the pear tree point cloud data (P) by circling around the target fruit tree. Figure 2 ) through the target fruit tree.

[0052] (2) At 9:00-17:00 in the morning, set up cubic grid points at a distance l (l = 0.5 m) from the ground in the east, west, south, and north directions with the pear tree trunk as the center, and set up line layers with a step size l = 0.7 m in vertical height to form a number of cubic grid spaces with a volume of 0.5 m x 0.5 m x 0.7 m (i.e., the canopy is processed in layers). Use a rod or wire to build a square grid to specifically divide the measurement area according to the above-mentioned divided cubic grid space. Set n sensors at a distance l (l = 0.5 m) from the trunk at the intersection of each cubic grid (i.e., the horizontal intersection between each layer of the canopy) to collect the light intensity of the points at different heights, different levels, and different shading conditions in the canopy.

[0053] Figure 1 For an embodiment of the present application, for the convenience of understanding, only 4 sensors set at the intersection of the 1st and 2nd layers are shown, and 4 fixed points where sensors are set at the intersection of other layers are also marked. The 4 sensors set at each layer can be used to measure the corresponding 4 light intensities E 1i , E 2i , E 3i , E4i At the same time, the absolute light intensity E0 under the condition of no shelter is measured J The relative light intensity Ei of the ith layer is obtained i As shown in the following formula:

[0054]

[0055] 2. Use CloudCompare software to remove irrelevant points and points below the crown.

[0056] 3. Read the fruit tree crown leaf point cloud data obtained in the previous step, set the single-layer thickness, plane method phasor, and inclination angle, use RANSAC to fit the plane to obtain the plane model coefficient, project the target point cloud to the plane, and save the projected point cloud in pcd format. Finally, the point cloud is output in png format.

[0057] 4. Preprocess each layer of point cloud pictures, including (1) using median filtering to remove isolated noise without destroying the boundary contour; (2) using morphological erosion and dilation to improve the continuity of the point cloud image in the two-dimensional plane; (3) using the Canny operator to outline the approximate outline of the target, providing a tracking path for the subsequent optimized Freeman chain code algorithm. Then use the Freeman tracking algorithm to find the nearest boundary point and label the pixel point, calculate the closed boundary area size after judging the vector boundary closure, and then accumulate the area to obtain the effective leaf area S of each crown layer. Ei .

[0058] 5. Use the layer effective leaf area as the independent variable and the layer relative light intensity value (the average value of the point light intensity of layer 4) as the dependent variable, and use linear regression to perform regression analysis. The distribution relationship between the layer effective leaf area and the layer relative light intensity is obtained, that is:

[0059] y = 27093.665 + 0.577x

[0060] 6. During the time period of implementing step (2) above, use the quadrat method to measure the light intensity for verification of the estimation model:

[0061] Use a square quadrat with a side length of 40 cm to select the layer leaf density sampling range, place the above quadrat at the interface of each layer obtained in step 1 (the height is at 0.7 m, 1.4 m, and 2.1 m from the ground), ensure that the center of each quadrat coincides with the center of the trunk, and the projection of each quadrat on the ground coincides. Count the number of leaves in i (i = 3) quadrats to obtain the number of leaves n i , and the average leaf size data S = (2 / 3) * a * b aiAs the blade area, 'a' represents the longest transverse blade length, and 'b' represents the longest longitudinal blade length. Using S... zi =S ai *n i Calculate the leaf area S within each quadrat. zi Using the projected area of ​​the quadrat (S 投 =50cm*50cm) in the canopy area S 冠i The actual effective leaf area of ​​the layer is obtained by estimating the proportion, as shown in the following formula:

[0062] S 实i =S zi / (S 投 / S 冠i ).

[0063] 7. Select 5 trees, divide each tree into 3 layers according to step 6, and calculate the actual effective leaf area of ​​each layer. Compare this with the corresponding leaf area S of the point cloud layer of the fruit tree obtained by optimizing the Preeman algorithm and pixel ratio method described in steps 1-5. Ei Compare (e.g.) Figure 6 The table below shows the detailed regression data and relative error values.

[0064] Detailed regression analysis data and relative error values

[0065]

[0066] The results above show that for canopies with a larger effective leaf area, the calculated error is smaller, and vice versa. The final average error between the two calculations is 6.74%.

[0067] 8. By selecting target fruit trees and stratifying them using the method described above, the distribution relationship between the effective leaf area and relative light intensity of each layer (obtained through optimization of the Preeman algorithm and pixel ratio method) was used to obtain a series of predicted light intensity values ​​at different canopy heights (0.7m, 1.4m, and 2.1m from the ground), as shown in the table below. The light intensity was rated as I, II, and III based on relative light intensity of 30%, 60%, and 80%.

[0068]

[0069] The rating level is positively correlated with the quality of the fruit, so the quality of the fruit can be preliminarily predicted by predicting the rating level through light intensity.

[0070] The above embodiments are only used to illustrate the present invention and are not intended to limit the present invention. Those skilled in the art can make various changes and modifications without departing from the essence and scope of the present invention. Therefore, all equivalent technical solutions also fall within the protection scope of the present invention.

[0071] That which is not described in detail in the specification is considered to be of prior art by those skilled in the art.

Claims

1. A method for evaluating the light efficacy of fruit tree canopies, characterized in that, Includes the following steps: Step 1: Acquire point cloud data of fruit trees using lidar equipment; Step 2: Use a light sensor to obtain the light intensity at multiple points in the fruit tree canopy and calculate the relative light intensity of each layer of the canopy. Step 3: Preprocess the fruit tree point cloud data obtained in Step 1, remove irrelevant point clouds, and obtain the fruit tree canopy leaf point cloud data. Step 4: Perform horizontal layering and dimensionality reduction on the point cloud data of the fruit tree canopy leaves after the preprocessing in Step 3 to obtain the point cloud data of each layer of the canopy and the corresponding dimensionality reduction images of each layer. Step 5: Extract the point cloud data of each layer of the canopy obtained in Step 4 and the corresponding dimensionality reduction images of each layer. Based on the Freeman chain code leaf area calculation method of multi-objective area optimization, and combined with the traditional pixel point ratio method, obtain the effective leaf area of ​​each layer of the canopy. Step 6: Confirm that there is a positive linear correlation between the effective leaf area of ​​each layer of the canopy obtained in Step 5 and the actual leaf area value of each layer. Based on the relative light intensity of each layer of the canopy obtained in Step 2, construct a relative light intensity estimation model based on the effective leaf area of ​​each layer of the canopy. Step 7: Using the estimation model obtained in Step 6, predict the light conditions of the fruit tree canopy based on the effective leaf area of ​​each layer of the actual fruit tree canopy.

2. The method for evaluating the light efficiency of fruit tree canopy as described in claim 1, characterized in that: The specific method for obtaining the multi-point light intensity of the fruit tree canopy in step 2 is as follows: Using the main trunk of the fruit tree as the center of the canopy, and extending the same distance l in four directions from the ground as fixed points, and using the ground as a reference, spatial layers are formed in the vertical height with a certain step size h, resulting in a series of cubic spaces. n sensors are deployed at the horizontal boundaries of the aforementioned cubic spaces, with each sensor at a horizontal distance l from the main trunk, to collect the light intensity E in areas with different heights, layers, and occlusion conditions. ji (j=1,2,3…n), and simultaneously measure the absolute light intensity E of each layer under unobstructed conditions. Ji The relative illumination intensity E of the i-th layer is obtained based on the above points. i As shown in the following formula:

3. The method for evaluating the light efficacy of fruit tree canopy as described in claim 1, characterized in that: The method for horizontal layering and dimensionality reduction of fruit tree canopy leaf point cloud data described in step 4 is as follows: Step 4-1: Read the point cloud data of the fruit tree canopy leaves, and according to the light point acquisition scheme, layer height data, set the single layer thickness, plane normal vector, and tilt angle to divide the point cloud of a single fruit tree to obtain canopy layered point clouds with a certain thickness at different heights. Step 4-2: Use the RANSAC algorithm to fit the plane and obtain the plane model coefficients. Then, perform horizontal projection of the XOY plane on each layer of the canopy point cloud to obtain the two-dimensional projection of each layer of the canopy point cloud, which is the dimension-reduced image of each layer.

4. The method for evaluating the light efficiency of fruit tree canopy as described in claim 1, characterized in that: The specific method for calculating canopy projection tracing in step 5 is as follows: Step 5-1: Preprocess the layered dimensionality reduction images obtained in step 4. The preprocessing process includes: (1) using median filtering to remove isolated noise without destroying the boundary contour; (2) using morphological erosion and dilation to improve the continuity of the point cloud image in the two-dimensional plane; (3) using the Canny operator to outline the general contour of the target and provide a tracking path for the subsequent use of the optimized Freeman chain code algorithm. Step 5-2: Using the Freeman tracking algorithm based on multi-target area optimization, vector labels are applied to unlabeled boundary pixels and adjacent boundary pixels respectively until the vector boundary is closed. The number of closed boundary pixels is calculated and an appropriate number of null points in the labels are deleted. The above operation is repeated until there are no unlabeled boundaries. Finally, the pixels are summed and the result is output. Step 5-3: Calculate the leaf area S using the traditional pixel ratio method. E As shown in the following formula, where P b To optimize the calculation of the total number of pixels occupied by the fruit tree layer projection surface using Freeman chain code; P t S represents the number of pixels in the layered projection rectangle of the fruit tree. t The actual area of ​​the layered projected rectangle for the fruit trees; 5. The method for evaluating the light efficacy of fruit tree canopy as described in claim 1, characterized in that: The specific method for constructing the relative illuminance intensity estimation model in step 6 is as follows: Using the effective leaf area of ​​each layer as the independent variable x and the relative light intensity of each layer of the canopy as the dependent variable y, linear regression analysis was performed to obtain the distribution relationship between the effective leaf area of ​​each layer and the relative light intensity of each layer, as shown in the following formula: y = b + ax; In the above formula, a and b are the coefficients of the regression equation.

6. A method for evaluating the light conditions of fruit tree canopies, characterized in that: Step 1: Select target fruit trees of the same variety at the appropriate time, collect their point cloud data in a standardized manner, and obtain the effective leaf area of ​​each layer of the canopy according to the method described in claim 1. Step 2: Substitute the effective leaf area of ​​each layer obtained in Step 1 into the relationship between the effective leaf area of ​​each layer and the relative light intensity distribution as described in claim 5 to obtain the corresponding predicted value of relative light intensity and evaluate the light conditions of the fruit tree canopy.

Images