Method for monitoring growth vigor of rape in flowering period based on near-surface visible light orthoimage

Abstract

The invention discloses a method for monitoring rape flowering stage growth vigor based on a near-surface visible light ortho-image, and relates to the technical field of rape flowering stage growth vigor monitoring. Within the near-earth height range, an unmanned aerial vehicle is used for shooting orthographic projection images of different periods of the flowering period of a rape group, related two-dimensional image parameters representing the flowering amount of the rape in the images are extracted, function fitting is carried out on the two-dimensional image parameters and the flowering period and the flowering amount of field investigation, and therefore the flowering process and dynamic state of the rape are described through the image parameters. Real-time and rapid monitoring of the rape flowering stage state is achieved, when the rape flowering stage is finished and whether fertilizer lacks or not can be judged, and a theoretical basis is provided for field management and regulation and control in the rape reproductive growth stage.

Description

Technical Field

[0001] The invention relates to the technical field of rapeseed flowering growth monitoring, and in particular to a method for monitoring rapeseed flowering growth based on near-ground visible light orthophoto images. Background Art

[0002] Rapeseed is the largest oil crop in my country, with a wide distribution range. Its quality and yield have an important impact on agricultural and economic development. The flowering period of rapeseed can accurately reflect the growth status of rapeseed and indicate the final grain yield, thus providing a basis for field management and fertilization.

[0003] Traditional rapeseed flowering period monitoring usually relies on ground surveys. These methods are time-consuming, labor-intensive, costly, and have low data accuracy, making it difficult to meet the needs of large-scale rapeseed flowering period monitoring. Existing rapeseed monitoring based on drone remote sensing has the following problems: (1) Expensive multispectral drones are used to obtain multi-band image information in large-scale airspace. The image acquisition is large-scale, time-consuming, not targeted, and has large interference noise. (2) The pixel extraction methods are mainly single color space (RGB, HSV, and Lab) and k-means extraction methods, which cannot separate rapeseed pixels well. There are under-segmentation (incomplete segmentation) and over-segmentation (too many segmentations), which can easily cause large statistical errors and take a long time to calculate. It is not suitable for pixel extraction of complex and changeable rapeseed images of different periods.

[0004] Therefore, the present invention provides a method for monitoring the growth of rapeseed during its flowering period based on near-ground visible light orthophoto images. Summary of the invention

[0005] In view of the above shortcomings, the present invention provides a method for monitoring the growth of rapeseed during the flowering period based on near-ground visible light orthophoto images, which solves the problem of low accuracy of monitoring results caused by poor rapeseed image segmentation effect and long calculation time. The specific technical solution is as follows:

[0006] The method for monitoring the growth of rapeseed during flowering period based on near-ground visible light orthophoto images comprises the following steps:

[0007] S1. Arrangement of field experiment:

[0008] (1) Preparation before sowing: The surface of the test plot should be relatively flat, without any drop, and without any obstacles around or above. Ensure that the drainage and irrigation water is unobstructed. The surface stubble height of the plot with rice as the previous crop should not be greater than 300 mm, and the soil moisture for sowing should be moderate, with a relative humidity of 40% to 60%. If it is dry land, the previous crop cannot be cruciferous crops, and the planting area should be more than 50 square meters. The test seeds are selected according to the local ecological conditions and production characteristics, and are excellent early-maturing semi-winter varieties suitable for local environment planting, with short growth period, fast germination, concentrated flowering period, and good resistance. The base fertilizer is prepared according to the local rapeseed habit of fertilization, and is applied to the soil before turning over and preparing the land. Generally, it is (15-15-15) compound fertilizer 30kg / mu-40kg / mu, and boron fertilizer (B content 12%) 0.5kg / mu. Fertilizers should be granular fertilizers for easy spreading and to prevent expired fertilizers from caking.

[0009] (2) Direct seeding: The sowing date is selected according to the sowing date of local planting habits, which is generally from October to late November. Sowing should be done early. If the sowing date is late, the sowing amount can be appropriately increased to make up for the late sowing. The sowing density can be achieved by manual broadcasting or row sowing. The direct sowing amount is 400g / mu-500g / mu, which can be mixed with granular boron fertilizer for sowing; the row sowing opens a 3-5 cm deep planting trough, the row spacing is 20-30 cm, the seed amount is 200g / mu-400g / mu, and the soil needs to be covered after sowing.

[0010] (3) Post-sowing management: Chemical closed weeding is to spray the entire field with 70-90 ml of 50% acetochlor emulsifiable concentrate or 40L-50L of Jindul water per mu within 3 days after sowing to prevent weeds. In terms of drainage and waterlogging prevention, rapeseed is a dryland crop and waterlogging in the field should be avoided. Drainage should be done in time during the rainy season. Thinning and finalization is to remove small and weak seedlings from the cotyledon stage to the three-leaf and one-heart stage to ensure uniform density during the flowering period. If there are broken ridges, re-sow or re-seedlings should be done in time. Water and fertilizer management is to irrigate and replenish water in time if drought occurs during the seedling stage and the initial flowering stage when water demand is high; when the relative soil moisture content is 70%, no irrigation is required, and timely irrigation should be carried out by furrow irrigation, and flooding should be avoided; according to the growth of rapeseed seedlings and the requirements of production agronomic regulations, nitrogen, phosphorus, potassium and boron fertilizers should be applied reasonably during the seedling stage. Generally, 5-10 kg / mu of urea should be applied during the seedling stage to ensure the growth of rapeseed. To prevent and control pests, apply insect remover powder in the two-leaf and one-heart stage to control flea beetles, spray high-efficiency cypermethrin in the late seedling stage to control cabbage loopers and diamondback moths, and use acetamiprid to control aphids in arid areas.

[0011] S2. Obtaining field survey data: Select 5 to 10 representative rapeseed plants in the test field and mark them. From the beginning of the flowering period, survey the total flowering amount of each marked rapeseed every 5 to 10 days until the end of the flowering period; at the same time, survey the density of the rapeseed population to estimate the canopy flower amount of the population, and record the dates of the initial flowering period, the peak flowering period and the final flowering period;

[0012] "Representative plants" refer to plants that can reflect the overall growth of rapeseed in the sampling area, and avoid taking plants that are too high or too low, underdeveloped or prematurely flowering. The same applies below;

[0013] S3. Obtaining image data: Starting from the flowering period, while conducting field surveys on the flowering amount at each period of the flowering period, select representative rapeseed groups, and divide the rapeseed groups into a priori group and a verification group. Use drones to take orthographic projection photos of the flowering canopy of each group of rapeseed groups near the ground to obtain images;

[0014] S4, extracting feature parameters: first convert the image from RGB color space to HSV color space, use H, V, a, b color channels to perform preliminary processing of image parameters, then use Matlab software to use k-means image segmentation method to extract yellow area pixels in the image, count the number of pixels, and name it S;

[0015] S5. Construct the fitting equation of characteristic parameters and field survey data: Use the Gussian function to fit the change of the number of pixels over time, and obtain the equation of image parameters representing the flowering period; Use the Gussian function to fit the number of pixels and the number of flowers in each time period, and obtain the characteristic equation of image parameters representing the growth of the flowering period;

[0016] S6. Verify the fitting equation: Use the S value data extracted from the verification group images and the field survey data to verify the flowering period equation and the rapeseed growth potential equation;

[0017] S7. Predicting the growth of rapeseed during flowering period: obtaining the orthophoto image of the rapeseed during flowering period in the designated monitoring area, and predicting the growth of rapeseed during flowering period through the verified fitting equation.

[0018] Furthermore, in step S1, the sowing is carried out by manual broadcasting or row broadcasting; the seed amount for manual broadcasting is 500g / mu, which can be mixed with granular boron fertilizer and broadcast together; the row broadcasting is to open a 5 cm deep planting trough, with a row spacing of 30 cm, a seed amount of 400g / mu, and cover the soil after sowing.

[0019] Furthermore, in step S2, the total number of flowers surveyed is to survey the total number of fully opened flowers.

[0020] Furthermore, in step S3, the a priori group is repeated three times.

[0021] Furthermore, in step S3, the drone is a small civilian unmanned aerial vehicle that can be equipped with a visible light camera.

[0022] Furthermore, in step S3, the drone takes photos at a flight altitude of 10-20m, and images of different flowering stages are taken at the same altitude. During shooting, the lens depression angle is 90°, the lens direction is perpendicular to the ground, and the canopy of the rapeseed group is located at the center of the shooting.

[0023] Furthermore, in step S3, the drone adopts an automatic exposure mode for shooting, sets the shooting image resolution to 5472×3648, and the image storage format to JPEG.

[0024] Furthermore, in step S4, the object of feature parameter extraction is the rapeseed canopy image at the center of the picture, with a size of 400×400 pixels, and one feature parameter is extracted from each photo.

[0025] Compared with the prior art, the present invention has the following beneficial effects:

[0026] 1. The present invention utilizes ordinary small UAVs to collect orthographic projection images of rapeseed groups at different periods of flowering period in a small-scale airspace near the ground. Through pre-processing + multi-channel combination + kmeans method, the two-dimensional image parameters characterizing the flowering amount of rapeseed in the image are extracted, and the function is fitted with the flowering period and flowering amount of field survey, so as to describe the flowering process and dynamics of rapeseed through image parameters, and realize rapid and accurate monitoring of the growth of rapeseed during flowering period.

[0027] 2. The present invention uses ordinary small drones to collect images in small-scale airspace near the ground, and only collects visible light band images for analysis. It does not need to use expensive multi-spectral drones to perform remote sensing in large-scale airspace to obtain multi-band image information. Therefore, the image acquisition scale of the present invention is small, time-consuming, more targeted, the details of rapeseed growth are richer, the interference noise is small, the amount of information required is small, the acquisition process is simpler and the cost is low.

[0028] 3. The present invention uses pre-processing + multi-channel combination + kmeans method to firstly perform a rough screening of sample image pixels through an empirical formula to reduce interference and calculation amount, and then convert the visible light image (generally RGB) into multiple color channel images (such as HSV, Lab), and combine them. The pixels are classified through the k-means clustering method, so as to accurately extract rapeseed flower pixels. Compared with the ordinary image feature parameter extraction, the extraction accuracy of this method is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for describing the embodiments are briefly introduced below.

[0030] Figure 1 The specific operation flow chart of characteristic parameter extraction of the present invention;

[0031] Figure 2 Aerial images of different periods, selected images, color space conversion and extracted rapeseed flower pixel comparison images of Example 1;

[0032] Figure 3 The fitting graph of the change of the number of pixels over time in Example 1 and the fitting graph of the number of pixels and the number of flowers blooming in each period;

[0033] Figure 4 This is a linear fitting diagram of flowering amount and rapeseed flower pixels in Example 1;

[0034] Figure 5 It is a comparison chart of the results of the traditional pixel extraction method and the "pre-processing + multi-channel combination + k-means extraction method" of the present invention;

[0035] Figure 6 The present invention is a general flow chart of monitoring the growth of rapeseed during flowering period based on near-ground visible light orthographic projection images. DETAILED DESCRIPTION

[0036] The specific embodiments of the present invention are described in detail below, but it should be understood that the protection scope of the present invention is not limited by the specific embodiments.

[0037] Example 1

[0038] The method for monitoring the growth of rapeseed during flowering period based on near-ground visible light orthophoto images comprises the following steps:

[0039] S1. Dynamic field test arrangement during rapeseed flowering period

[0040] (1) Preparation before sowing: Select a flat rice-oil rotation field with smooth drainage and irrigation for the test plot; dig furrows and build beds, with a bed width of 2.5m and a bed length of 5m, and plant a total of 4 beds; dig box furrows, surrounding furrows, and waist furrows with a furrow width of 30cm; set up 2m protection rows around the perimeter, and the planting area is about 150㎡. The test variety is the local main variety, the cabbage-type winter rapeseed hybrid variety Jingyou 69. Base fertilizer preparation: Apply (15-15-15) compound fertilizer 40kg / mu, boron fertilizer (B content 12%) 0.5kg / mu, and apply it to the soil at one time before land preparation.

[0041] (2) Sowing: The sowing date is October 28, 2022. The artificial row sowing method is adopted. The planting trough is 3-5 cm deep on the bed surface. The sowing amount is 200g / mu and the row spacing is about 30cm. Soil needs to be covered after sowing. Remove the small and weak seedlings from the cotyledon stage to the three-leaf and one-heart stage to ensure the same density during the flowering period. Thin out or supplement the seedlings in time according to the row spacing of 30cm×10cm.

[0042] (3) Field management: Irrigate once after sowing, irrigate once during the period of high water demand such as the seedling stage and the early flowering stage, and drain water in time during the rainy season. Chemical closed weeding. After sowing and covering the soil, spray 45L of Jindul with water on the soil surface for closed weeding. According to the seedling conditions, apply 5kg / mu of urea at the seedling stage to ensure the growth of rapeseed. Apply beetle powder once at the two-leaf and one-heart stage to prevent flea beetles, and spray high-efficiency chlorpyrifos once at the late seedling stage to prevent cabbage worms and diamondback moths.

[0043] S2. Field survey of rapeseed flowering quantity at different stages of flowering period

[0044] Five representative rapeseed plants were selected for marking in each compartment. From the beginning of the flowering period, the total flowering amount (the total number of fully opened flowers) of each rapeseed plant was surveyed every 5 days until the end of the flowering period. A total of 5 surveys were conducted on February 16, February 21, February 28, March 6, and March 14, 2023. The dates when the rapeseed population entered the initial flowering period, the peak flowering period, and the final flowering period were February 21, March 2, and March 10, 2023, respectively. Three 1m×1m sample plots were surveyed, and the average planting density was converted to 18,000 plants / mu.

[0045]

[0046] S3. Obtain image data

[0047] (1) Acquisition time: The survey was conducted simultaneously with the flowering period survey. Four representative rapeseed populations were selected (one for each of the four fields, three for replication, and one for verification). Each time, a drone was used to take multiple orthographic projection images of the rapeseed canopy at the four locations. Four images were selected for feature function extraction.

[0048] (2) Environmental requirements: Choose clear and windless weather during the rapeseed flowering period.

[0049] (3) UAV parameter setting: A small civilian unmanned aerial vehicle (DJI Air 2S) equipped with a visible light camera was used to take orthographic projection photos near the ground. The lens parameters were viewing angle: 88°, 35mm equivalent focal length: 22mm, aperture: f / 2.8, focal range: 0.6m~∞m; photo size 20 million pixels. The shooting flight altitude range was 10m~20m to obtain clear images. Images of different flowering stages were taken at the same altitude. The lens depression angle was 90°, the lens direction was perpendicular to the ground, and the research object (the canopy of the rapeseed group) was located in the center of the shooting. The automatic exposure mode was used for taking photos, and the image resolution was set to 5472×3648, and the image storage format was JPEG.

[0050] S4. Extract feature parameters

[0051] The digital image of the rapeseed canopy at the center of the photo was selected as the object for image data feature parameter extraction, and one pixel with a size of 400×400 was extracted from each photo.

[0052] In order to remove the interference factor of brightness, the image is converted from RGB color space to HSV color space, and the H (hue), a, and b color channels are used for preliminary processing of image parameters. The yellow area (representing rapeseed flowers) pixels in the image are extracted using the MATLAB software k-means image segmentation method, and the number of pixels is counted and named S. The specific operation process of image feature parameter extraction is shown in Figure 1 .

[0053] Comparison of aerial images, selected images, color space conversion and extracted rapeseed pixels at different times Figure 2 ;

[0054] The pixel values ​​of rapeseed flowers at different periods in the 400*400 pixel image samples are counted, and the results are shown in Table 1:

[0055] Table 1

[0056]

[0057] S5. Construct fitting equations between characteristic parameters and field survey data

[0058] The Gaussian function is used to fit the number of days after sowing, the rapeseed pixel value, and the flowering amount to obtain the fitting image (such as Figure 3 Specific equation parameters are shown in Table 2:

[0059] Table 2

[0060]

[0061] There is a linear positive correlation between rapeseed flower pixels and flowering amount (see Figure 4 ), and the performance is significant. The linear fitting equation parameters are shown in Table 3:

[0062] Table 3

[0063] Linear equations y=a+b*x Drawing Flowering amount Weight Unweighted intercept 7.57515±5.59719 Slope 6.58069±0.27813 Residual sum of squares 233.04448 Pearson's 0.99733 R-squared (COD) 0.99467 Adjusted R-squared 0.99289

[0064] S6. Verify the fitting equation

[0065] Substitute the verification data (verification group S repeated 4 times) into the Gaussian fitting equation, calculate the days after sowing and the amount of flowering to verify the accuracy of the equation. If the equation is accurate, you can randomly take orthophoto images of the flowering period and use the equation to predict the flowering period and amount at that time; if the verification accuracy is poor, you need to adjust the equation.

[0066] The comparison between the traditional pixel extraction method (single color space (RGB, HSV and Lab) + k-means extraction method) and the present invention's "pre-processing + multi-channel combination + k-means extraction method" is shown in Figure 5 ,

[0067] like Figure 5 As shown in the figure, the traditional pixel extraction method (single color space (RGB, HSV and Lab) + k-means extraction method) cannot separate the rapeseed flower pixels well, and there are under-segmentation (incomplete segmentation) and over-segmentation (too many segmentations) phenomena, and the operation time is long. The segmentation method of the present invention can separate and extract the rapeseed flower pixels in each time period well, and the operation time is short (ranked second).

[0068] In summary, the present invention uses ordinary small drones to collect orthographic projection images of rapeseed groups at different periods of flowering period in small-scale airspace near the ground. Through pre-processing + multi-channel combination + kmeans method, the two-dimensional image parameters related to the flowering amount of rapeseed in the image are extracted, and the function is fitted with the flowering period and flowering amount of field survey, so that the flowering process and dynamics of rapeseed are described by image parameters, and the growth of rapeseed during flowering period is realized. Rapid and accurate monitoring.

[0069] The foregoing description of specific exemplary embodiments of the present invention is for the purpose of illustration and demonstration. These descriptions are not intended to limit the present invention to the precise form disclosed, and it is clear that many changes and variations can be made based on the above teachings. The purpose of selecting and describing the exemplary embodiments is to explain the specific principles of the present invention and its practical application, so that those skilled in the art can realize and utilize various different exemplary embodiments of the present invention and various different selections and changes. The scope of the present invention is intended to be limited by the claims and their equivalents.

Claims

1. A method for monitoring the growth of rapeseed during flowering period based on near-ground visible light orthophoto images, characterized in that: The following steps are involved: S1. Arrangement of field experiment: select a field with a flat surface, no drop, no obstacles around and above, and unobstructed drainage and irrigation as the test field, apply basal fertilizer to the soil of the test field before tilling and preparing the land, select excellent early-maturing semi-winter varieties suitable for local environment planting from October to late November, with short growth period, fast germination, concentrated flowering period, and good resistance, for sowing, and carry out field management after sowing; S2. Obtaining field survey data: Select 5 to 10 representative rapeseed plants in the test field and mark them. From the beginning of the flowering period, survey the total flowering amount of each marked rapeseed every 5 to 10 days until the end of the flowering period; at the same time, survey the density of the rapeseed population to estimate the canopy flower amount of the population, and record the dates of the initial flowering period, the peak flowering period and the final flowering period; S3. Obtaining image data: Starting from the flowering period, while conducting field surveys on the flowering amount at each period of the flowering period, select representative rapeseed groups, and divide the rapeseed groups into a priori group and a verification group. Use drones to take orthographic projection photos of the flowering canopy of each group of rapeseed groups near the ground to obtain images; S4, extracting feature parameters: first convert the image from RGB color space to HSV color space, use H, V, a, b color channels to perform preliminary processing of image parameters, then use Matlab software to use k-means image segmentation method to extract yellow area pixels in the image, count the number of pixels, and name it S; S5. Construct the fitting equation of characteristic parameters and field survey data: Use the Gussian function to fit the change of the number of pixels over time, and obtain the equation of image parameters representing the flowering period; Use the Gussian function to fit the number of pixels and the number of flowers in each time period, and obtain the characteristic equation of image parameters representing the growth of the flowering period; S6. Verify the fitting equation: Use the S value data extracted from the verification group images and the field survey data to verify the flowering period equation and the rapeseed growth potential equation; S7. Predicting the growth of rapeseed during flowering period: obtaining the orthophoto image of the rapeseed during flowering period in the designated monitoring area, and predicting the growth of rapeseed during flowering period through the verified fitting equation.

2. The method for monitoring the growth of rapeseed during flowering period based on near-ground visible light orthophoto images according to claim 1 is characterized in that: In step S1, the sowing is carried out by manual broadcasting or row broadcasting; the seed amount for manual broadcasting is 400-500g / mu, which can be mixed with granular boron fertilizer for broadcasting; the row broadcasting is to open a planting trough 3-5 cm deep, with a row spacing of 20-30 cm, a seed amount of 200-400g / mu, and cover the soil after sowing.

3. The method for monitoring the growth of rapeseed during flowering period based on near-ground visible light orthophoto images according to claim 1 is characterized in that: In step S2, the total number of flowers bloomed is investigated to determine the total number of fully opened flowers.

4. The method for monitoring rapeseed flowering growth based on near-ground visible light orthophoto images according to claim 1 is characterized in that: In step S3, the a priori group is repeated three times.

5. The method for monitoring the growth of rapeseed during flowering period based on near-ground visible light orthophoto images according to claim 1 is characterized in that: In step S3, the drone is a small civilian unmanned aerial vehicle that can be equipped with a visible light camera.

6. The method for monitoring rapeseed flowering growth based on near-ground visible light orthophoto images according to claim 1 is characterized in that: In step S3, the drone takes photos at a flight altitude of 10-20m, and images of different flowering stages are taken at the same altitude. During shooting, the lens depression angle is 90°, the lens direction is perpendicular to the ground, and the canopy of the rapeseed group is located at the center of the shooting.

7. The method for monitoring the growth of rapeseed during flowering period based on near-ground visible light orthophoto images according to claim 1 is characterized in that: In step S3, the drone uses an automatic exposure mode for shooting, sets the image resolution to 5472×3648, and the image storage format to JPEG.

8. The method for monitoring rapeseed flowering growth based on near-ground visible light orthophoto images according to claim 1 is characterized in that: In step S4, the object of feature parameter extraction is the rapeseed canopy image at the center of the picture, with a size of 400×400 pixels, and one feature parameter is extracted from each photo.

Images