A method for measuring leaf area index based on multi-exposure fusion

Abstract

The present invention discloses a method for measuring leaf area index based on multi-exposure fusion. First, a fisheye camera is used to capture multiple canopy images at the same sampling point. The canopy images are then converted into canopy grayscale images. The canopy grayscale images are then subjected to multi-exposure fusion processing to obtain a multi-exposure fused image. The multi-exposure fused image is binarized, and then the canopy porosity at multiple angles is extracted from the binarized image to invert the leaf area index.

Description

Technical Field

[0001] The present invention belongs to the technical field of vegetation parameter measurement, and more specifically, relates to a method for measuring leaf area index based on multi-exposure fusion. Background Art

[0002] The Leaf Area Index (LAI) is an important vegetation structure variable that plays a crucial role in studying plant feedback to climate. As a fundamental property of global vegetation, LAI is listed as a key climate variable by the global climate change research community. It is a crucial parameter for characterizing canopy photosynthesis and determining vegetation growth status, and is also a fundamental parameter for numerous mathematical models in agricultural science, ecological science, and remote sensing science. The Leaf Area Index can be considered the ratio of leaf area to ground area. It reflects the number and distribution of leaves and is a fundamental factor for studying light penetration through the canopy, canopy productivity, total evaporation and transpiration losses from the soil beneath the canopy, canopy interception, and soil temperature. Therefore, by monitoring the Leaf Area Index of crops, we can understand the growth status of crops, such as whether they are susceptible to pests and diseases and whether fertilization is needed, and thus manage them accordingly. The Leaf Area Index of crops can also be used to estimate crop yields.

[0003] Currently, leaf area index (LAI) is measured using two methods: direct and indirect. Direct measurement typically involves two steps: collecting samples and measuring leaf area. While highly accurate, it is highly destructive and labor-intensive, limiting measurements to small canopies and essentially impossible for large tree canopies such as forests. Indirect measurement, using a light propagation model, allows for a relatively simple measurement of light penetration and provides an accurate estimate of LAI based on the Lambert-Beer law.

[0004] Indirect measurement methods mainly include oblique sampling, radiometric methods, and porosity inversion methods, which have the advantages of being fast and simple. The problem with oblique sampling is that it requires a large number of specialized probes and is quite cumbersome to operate, making it unsuitable for obtaining parameters of plants with larger canopies. Radiometric methods often require the use of lidar (which can be divided into ground-based, vehicle-mounted, airborne, and satellite-based types according to the carrier platform) to obtain spatial information of the target object, which makes data processing complex and costly. The representative method of porosity inversion, the hemispherical imaging method, uses a fisheye lens to obtain canopy image information at nearly 180°, obtains canopy porosity through image processing, and then inverts the leaf area index. Since Even et al. used it for radiation monitoring of vegetation canopies, the hemispherical imaging method has been widely studied and applied due to its ultra-large viewing angle range and the low cost of equipment for obtaining fisheye images.

[0005] In recent years, with the improvement and progress of digital cameras and digital image processing technology, the function of hemispherical imaging method has been increasingly improved. It has now become an important research method in various fields. A typical hemispherical photography method is the invention patent application "A Method for Measuring Leaf Area Index" published on January 19, 2018 with publication number CN107610066A. However, this method has the problem of overexposure of canopy images, which leads to underestimated LAI values. Summary of the Invention

[0006] The purpose of the present invention is to overcome the shortcomings of the existing technology and provide a method for measuring leaf area index based on multi-exposure fusion, which improves the overexposure problem of canopy images and reduces the loss of vegetation canopy information in canopy images, thereby reducing the problem of underestimation of LAI measurement values.

[0007] To achieve the above object of the invention, the present invention provides a method for measuring leaf area index based on multi-exposure fusion, characterized by comprising the following steps:

[0008] (1) Collect canopy images;

[0009] (1.1) Use the camera's automatic exposure mode to photograph the plant canopy, collect canopy images, and record the exposure time when photographing the canopy images;

[0010] (1.2) Using the recorded exposure time as a reference, set K different exposure time levels. These K exposure time levels must include the exposure time of the automatic exposure mode.

[0011] (1.3) Fix the exposure time of the camera at K exposure time levels in sequence, and then take pictures at the same sampling point in sequence, so as to collect K canopy images with different exposure times at the same sampling point;

[0012] (2) Perform multi-exposure fusion processing on canopy images;

[0013] (2.1) K canopy images of the same sampling point are converted into grayscale images, where the kth canopy grayscale image is recorded as I k , k=1,2,…,K;

[0014] (2.2), the canopy grayscale image I k Perform global average filtering to obtain the global filtered image L k , and then L k Substitute each pixel point in formula (1) to calculate the canopy grayscale image I k Global exposure weight at each pixel location;

[0015]

[0016] in, Indicates I k The global exposure weight at the pixel point (x, y) in the image is the base layer, g is the global layer, and σ is the global exposure weight. g are the parameters of the global Gaussian distribution, express The pixel value at the pixel point (x, y);

[0017] (2.3), the canopy grayscale image I k After the guided filter, I k Basic Layer B k ;

[0018] B k =G r,ε (I k ,I k ) (2)

[0019] Among them, G r,ε () represents a guided filter;

[0020] (2.4), calculate the canopy grayscale image I k Local exposure weight at each pixel location

[0021]

[0022] Among them, the superscript l indicates local, B k (x,y) represents the base layer B k The pixel value at the pixel point (x, y), σ l is the parameter of the local Gaussian distribution;

[0023] (2.5) The global exposure weight of each pixel and local exposure weights Blended together as base weights

[0024]

[0025] (2.6), calculate the canopy grayscale image I k Detail layer D k ;

[0026] D k =I k -B k (5)

[0027] (2.7), the canopy grayscale image I k The filtered image is processed by a 7×7 average filter Then Substitute each pixel point in formula (6) to calculate the canopy grayscale image I k Detail weight

[0028]

[0029] Among them, the superscript D represents the detail layer, express The pixel value at the pixel point (x, y), σ D is the parameter that controls the Gaussian distribution;

[0030] (2.8), in the basic layer B k and detail layer D k , the weight value at each pixel position is divided by the sum of the weights at each pixel position to obtain the normalized weight at each pixel position, and the normalized weights at each pixel position constitute a normalized weight matrix;

[0031] (2.9) Calculate the canopy image after multi-exposure fusion processing;

[0032]

[0033] in, Base layer B k The corresponding normalized weight matrix, Represents the detail layer D k The corresponding normalized weight matrix, α, represents the control of the fusion contrast of the multi-exposure fusion image;

[0034] (3) Calculate the leaf area index of the canopy image after multi-exposure fusion using hemispherical photography method;

[0035] (3.1) For the canopy image P after multi-exposure fusion, out Perform binarization processing. After the binarization processing, the white pixels in the obtained binary image are marked as background pixels, and the black pixels are marked as vegetation canopy pixels;

[0036] (3.2), set the calibration projection equation r = fθ, f is the radius of the effective circular imaging area of ​​the fisheye lens, θ is the incident angle, and r represents the ring radius;

[0037] According to the calibration projection equation, n rings are taken in the binary image according to the incident angle range (0~θ0). Each ring corresponds to the incident angle interval with a width of dθ=θ0 / n. From the inside to the outside, the radius of each ring is r1=f·θ0 / n, r2=2f·θ0 / n, r3=3f·θ0 / n......r n =f·θ0, the incident angles of each ring are θ1=1 / (2n)θ0, θ2=3 / (2n), θ3=5 / (2n)θ0,…,θn =(2n-1) / (2n)θ0;

[0038] (3.3), calculate the canopy porosity T(θ i );

[0039]

[0040] Among them, T(θ i ) represents the incident angle θ i Canopy porosity under the f (θ i ) represents the incident angle θ i The number of vegetation canopy pixels in the binary image, N b (θ i ) represents the incident angle θ i The number of background pixels in the binary image;

[0041] (3.4) Calculate the leaf area index LAI;

[0042]

[0043] in, is the normalized weight factor of each ring, satisfying

[0044] The object of the invention of the present invention is achieved like this:

[0045] The present invention is based on a method for measuring leaf area index by multi-exposure fusion. First, a fisheye camera is used to capture multiple canopy images at the same sampling point. The canopy images are then converted into canopy grayscale images. The canopy grayscale images are then subjected to multi-exposure fusion processing to obtain a multi-exposure fused image. The multi-exposure fused image is binarized, and then the canopy porosity at multiple angles is extracted from the binarized image to invert the leaf area index.

[0046] At the same time, the method for measuring leaf area index based on multi-exposure fusion of the present invention also has the following beneficial effects:

[0047] (1) The general LAI measurement method is greatly affected by the imaging light. The present invention has good applicability and reduces the impact of light changes on imaging quality.

[0048] (2) In the automatic exposure mode, the LAI inverted by the hemispherical photography method will have the problem of low overall values. The present invention adopts a multi-exposure fusion method and the LAI inverted by the hemispherical photography method, which reduces the LAI underestimation problem compared with the automatic exposure mode and improves the LAI inversion accuracy.

[0049] (3) There is a problem of missing information in the canopy image obtained in the automatic exposure mode. The present invention reduces the information missing of the canopy image through multi-exposure fusion. The obtained canopy image has more complete information, which greatly improves the accuracy of subsequent vegetation processing and analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 This is a flow chart of a method for measuring leaf area index based on multi-exposure fusion according to the present invention;

[0051] Figure 2 This is a schematic diagram of a canopy image;

[0052] Figure 3 This is a schematic diagram of the canopy image after grayscale processing;

[0053] Figure 4 It is a schematic diagram of the ring division and the image in automatic exposure mode;

[0054] Figure 5 This is a schematic diagram of the ring division. DETAILED DESCRIPTION

[0055] The following describes the specific embodiments of the present invention in conjunction with the accompanying drawings so that those skilled in the art can better understand the present invention. It should be noted that in the following description, when detailed descriptions of known functions and designs may dilute the main content of the present invention, such descriptions will be omitted here.

[0056] Example

[0057] Figure 1 This is a flow chart of the method for measuring leaf area index based on multi-exposure fusion of the present invention.

[0058] In this embodiment, a shooting area of ​​102°54′E~104°53′E and 30°05′~31°26′N was selected. The area has a flat terrain and rich vegetation types, including evergreen forests dominated by trees such as ginkgo, camphor, banyan, osmanthus, and nanmu. The vegetation canopy image was taken by a Nikon D90 digital camera, which has a wide exposure time range, a maximum resolution of 4288×2848, a LENSBABY Circular Fisheye5.8 F3.5 full-circle fisheye lens, and a field of view of 185°. In order to ensure that the shooting position is fixed, the camera is mounted on a horizontal tripod. In order to quickly take a set of canopy pictures with different exposure times (exposure time ranges from 1 / 2s to 1 / 4000s), the camera is connected to a laptop, and the digiCamControl software is used to control the camera shooting using a program. Then, the camera is set to automatic exposure mode to obtain automatically exposed canopy pictures for comparison of experimental results.

[0059] like Figure 1 As shown, a method for measuring leaf area index based on multi-exposure fusion of the present invention is described in detail, which specifically includes three steps: S1, collecting canopy images; S2, performing multi-exposure fusion processing on the canopy images; S3, calculating the leaf area index of the canopy images after multi-exposure fusion using hemispherical photography method;

[0060] Below we explain the three steps in detail, as follows:

[0061] S1, collect canopy images;

[0062] S1.1. Use the camera's automatic exposure mode to photograph the plant canopy, collect canopy images, and record the exposure time when photographing the canopy images;

[0063] S1.2. Using the recorded exposure time as a reference, set six different exposure time levels. These six exposure time levels must include the exposure time for the automatic exposure mode.

[0064] S1.3. Fix the exposure time of the camera at 6 exposure time levels in sequence, and then take pictures at the same sampling point in sequence, so as to collect 6 canopy images with different exposure times at the same sampling point.

[0065] In this embodiment, the six canopy images are as follows: Figure 2 As shown, the exposure values ​​of the six canopy images from a to f are 1 / 80s, 1 / 100s, 1 / 125s, 1 / 160s, 1 / 200s, and 1 / 250s, respectively.

[0066] S2, perform multi-exposure fusion processing on the canopy image;

[0067] S2.1. Convert the six canopy images of the same sampling point with different exposure times into grayscale images, where the kth canopy grayscale image is denoted as I k , k=1,2,…,6; the converted grayscale image is as follows Figure 3 As shown;

[0068] S2.2, the canopy grayscale image I k Perform global average filtering to obtain a global filtered image Then Substitute each pixel point in formula (1) to calculate the canopy grayscale image I k Global exposure weight at each pixel location;

[0069]

[0070] in, Indicates I kThe global exposure weight at the pixel point (x, y) in the image is the base layer, g is the global layer, and σ is the global exposure weight. g is the parameter of the global Gaussian distribution, with a value of 0.2. express The pixel value at the pixel point (x, y);

[0071] S2.3, the canopy grayscale image I k After the guided filter, I k Basic Layer B k ;

[0072] B k =G r,ε (I k ,I k ) (2)

[0073] Among them, G r,ε () represents a guided filter;

[0074] S2.4. Calculate canopy grayscale image I k Local exposure weight at each pixel location

[0075]

[0076] Among them, the superscript l indicates local, B k (x,y) represents the base layer B k The pixel value at the pixel point (x, y), σ l is the parameter of the local Gaussian distribution, and its value is 0.5;

[0077] S2.5. The global exposure weight of each pixel and local exposure weights Blended together as base weights

[0078]

[0079] S2.6. Calculate canopy grayscale image I k Detail layer D k ;

[0080] D k =I k -B k (5)

[0081] S2.7, the canopy grayscale image I k The filtered image is processed by a 7×7 average filter Then Substitute each pixel point in formula (6) to calculate the canopy grayscale image Ik Detail weight

[0082]

[0083] Among them, the superscript D represents the detail layer, express The pixel value at the pixel point (x, y), σ D To control the parameters of Gaussian distribution, the value is 0.12;

[0084] S2.8, in the basic layer B k and detail layer D k , the weight value at each pixel position is divided by the sum of the weights at each pixel position to obtain the normalized weight at each pixel position, and the normalized weights at each pixel position constitute a normalized weight matrix;

[0085] S2.9. Calculate the canopy image after multi-exposure fusion processing;

[0086]

[0087] in, Base layer B k The corresponding normalized weight matrix, Represents the detail layer D k The corresponding normalized weight matrix, α, represents the fusion contrast of the multi-exposure fusion image, α≥1, and the α value in this example is 1.3;

[0088] In this embodiment, the multi-exposure fused image is as follows: Figure 4 The image in b and the automatic exposure mode are as shown in Figure 4 As shown in a in the figure, by comparison, it can be seen that the image after multi-exposure fusion has clearer contours than the automatic exposure image, retains more canopy information, and greatly reduces the image overexposure problem.

[0089] S3, calculate the leaf area index of the canopy image after multi-exposure fusion using hemispherical photography method;

[0090] S3.1. Multi-exposure fused canopy image P out Perform binarization processing. After the binarization processing, the white pixels in the obtained binary image are marked as background pixels, and the black pixels are marked as vegetation canopy pixels;

[0091] S3.2. Set the calibration projection equation r = fθ, where f is the radius of the effective circular imaging area of ​​the fisheye lens, θ is the incident angle, and r represents the ring radius;

[0092] According to the calibration projection equation, 5 rings are taken in the binary image according to the incident angle range of (0~θ0). Each ring corresponds to the incident angle interval with a width of dθ=75° / 5=15°. The radius of each ring from the inside to the outside is r1=f*15, r2=f*30, r3=f*45, r4=f*60, r5=f*75, and the incident angle of each ring is represented by θ1=7°, θ2=23°, θ3=38°, θ4=53°, θ5=68° respectively. The schematic diagram of the ring division is shown in the figure. Figure 5 As shown;

[0093] S3.3. Calculate the canopy porosity T(θ i );

[0094]

[0095] Among them, T(θ i ) represents the incident angle θ i Canopy porosity under the f (θ i ) represents the incident angle θ i The number of vegetation canopy pixels in the binary image, N b (θ i ) represents the incident angle θ i The number of background pixels in the binary image;

[0096] S3.4. Calculate the leaf area index LAI;

[0097]

[0098] in, is the normalized weight factor of each ring, satisfying The weight value w before normalization i =2sinθ i cosθ i Δθ i , the relationship between the normalized weight factor and the weight factor before normalization

[0099] As shown in Table 1, the weight factors of each ring are as follows:

[0100]

[0101] Table 1

[0102] Although the above describes the illustrative specific embodiments of the present invention to facilitate understanding of the present invention by those skilled in the art, it should be clear that the present invention is not limited to the scope of the specific embodiments. For those skilled in the art, as long as various changes are within the spirit and scope of the present invention as defined and determined by the appended claims, these changes are obvious, and all inventions and creations using the concepts of the present invention are protected.

Claims

1. A method for measuring leaf area index based on multi-exposure fusion, characterized in that: The following steps are involved: (1) Collect canopy images; (1.1) Use the camera's automatic exposure mode to photograph the plant canopy, collect canopy images, and record the exposure time when photographing the canopy images; (1.2) Using the recorded exposure time as a reference, set K different exposure time levels, where the K exposure time levels include the exposure time of the automatic exposure mode; (1.3) Fix the exposure time of the camera at K exposure time levels in sequence, and then take pictures at the same sampling point in sequence, so as to collect K canopy images with different exposure times at the same sampling point; (2) Perform multi-exposure fusion processing on canopy images; (2.1) K canopy images of the same sampling point are converted into grayscale images, where the kth canopy grayscale image is recorded as I k , k=1,2,…,K; (2.2), the canopy grayscale image I k Perform global average filtering to obtain the global filtered image L k , and then L k Substitute each pixel point in formula (1) to calculate the canopy grayscale image I k Global exposure weight at each pixel location; in, Indicates I k The global exposure weight at the pixel point (x, y) in the image is the base layer, g is the global layer, and σ is the global exposure weight. g are the parameters of the global Gaussian distribution, express The pixel value at the pixel point (x, y); (2.3), the canopy grayscale image I k After the guided filter, I k Basic Layer B k ; B k =G r,ε (I k ,I k ) (2) Among them, G r,ε () represents a guided filter; (2.4), calculate the canopy grayscale image I k Local exposure weight at each pixel location Among them, the superscript l indicates local, B k (x,y) represents the base layer B k The pixel value at the pixel point (x, y), σ l is the parameter of the local Gaussian distribution; (2.5) The global exposure weight of each pixel and local exposure weights Blended together as base weights (2.6), calculate the canopy grayscale image I k Detail layer D k ; D k =I k -B k (5) (2.7), the canopy grayscale image I k The filtered image is processed by a 7×7 average filter Then Substitute each pixel point in formula (6) to calculate the canopy grayscale image I k Detail weight Among them, the superscript D represents the detail layer, express The pixel value at the pixel point (x, y), σ D is the parameter that controls the Gaussian distribution; (2.8), in the basic layer B k and detail layer D k , the weight value at each pixel position is divided by the sum of the weights at each pixel position to obtain the normalized weight at each pixel position, and the normalized weights at each pixel position constitute a normalized weight matrix; (2.9) Calculate the canopy image after multi-exposure fusion processing; in, Base layer B k The corresponding normalized weight matrix, Represents the detail layer D k The corresponding normalized weight matrix, α, represents the control of the fusion contrast of the multi-exposure fusion image; (3) Calculate the leaf area index of the canopy image after multi-exposure fusion using hemispherical photography method; (3.1) For the canopy image P after multi-exposure fusion, out Perform binarization processing. After the binarization processing, the white pixels in the obtained binary image are marked as background pixels, and the black pixels are marked as vegetation canopy pixels; (3.2), set the calibration projection equation r = fθ, f is the radius of the effective circular imaging area of ​​the fisheye lens, θ is the incident angle, and r represents the ring radius; According to the calibration projection equation, n rings are taken in the binary image according to the incident angle range (0~θ0). Each ring corresponds to the incident angle interval with a width of dθ=θ0 / n. From the inside to the outside, the radius of each ring is r1=f·θ0 / n, r2=2f·θ0 / n, r3=3f·θ0 / n......r n =f·θ0, the incident angles of each ring are θ1=1 / (2n)θ0, θ2=3 / (2n), θ3=5 / (2n)θ0,…,θ n =(2n-1) / (2n)θ0; (3.3), calculate the canopy porosity T(θ i ); Among them, T(θ i ) represents the incident angle θ i Canopy porosity under the f (θ i ) represents the incident angle θ i The number of vegetation canopy pixels in the binary image, N b (θ i ) represents the incident angle θ i The number of background pixels in the binary image; (3.4) Calculate the leaf area index LAI; in, is the normalized weight factor of each ring, satisfying

Images