A method for identifying the growth trend of highland barley in a low-temperature disaster environment

Abstract

The present application relates to the technical field of image processing, and particularly relates to a kind of identification method of low-temperature disaster environment under the growth trend of highland barley, comprising: obtaining the original image of target highland barley growth area, and the original image is preprocessed, and the target gray image after enhanced texture is obtained;According to target gray image, determine the interference peak critical threshold;According to target gray image, generate gradient direction histogram, and based on the peak value information in gradient direction histogram and interference peak critical threshold, determine a plurality of effective slope values;According to a plurality of effective slope values, determine slope variance;In the case where the difference of slope variance relative to preset growth reference value exceeds first preset threshold, determine the growth trend of highland barley in target highland barley growth area as lodging growth.The present application improves the accuracy of the growth trend identification of highland barley in low-temperature disaster environment.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, specifically to a method for recognizing the growth status of highland barley under low-temperature disaster environments. Background Technology

[0002] As a staple food crop in highland areas, barley is frequently subjected to natural disasters such as low-temperature freezing damage during its growth process. In severe cases, this can lead to lodging of the plants, resulting in reduced yields or even total crop failure. Therefore, timely and accurate monitoring of crop growth after a disaster is crucial for implementing effective agronomic measures and minimizing losses.

[0003] Currently, in the field of agricultural monitoring, acquiring field images using remote sensing platforms such as drones has become a common method, and related technologies are constantly evolving. However, in practical applications, especially in complex field environments affected by low temperatures, existing methods are often not ideal in determining crop growth status, and the accuracy and reliability of the results still need further improvement. Summary of the Invention

[0004] To address the technical problem of inaccurate identification of barley growth status under low-temperature disaster environments, the present invention aims to provide a method for identifying barley growth status under low-temperature disaster environments. The specific technical solution adopted is as follows: Firstly, a method for identifying the growth status of barley under low-temperature disaster environments is provided. This method includes: acquiring an original image of the target barley growth area and preprocessing the original image to obtain an enhanced textured grayscale image; determining a critical threshold for interference peaks based on the target grayscale image, where the critical threshold is used to characterize the number of interference peaks in the gradient direction histogram; generating a gradient direction histogram based on the target grayscale image, and determining multiple effective slope values ​​based on the critical threshold for interference peaks and the peak information in the gradient direction histogram, where the multiple effective slope values ​​are used to characterize key transition features in the image texture direction; determining a slope variance based on the multiple effective slope values, where the slope variance is used to characterize the degree of fluctuation and dispersion of the gradient direction slope; and determining that the growth status of the barley in the target barley growth area is lodging growth when the difference between the slope variance and a preset growth benchmark value exceeds a first preset threshold.

[0005] In one possible design, the original image is preprocessed, including: converting the original image to grayscale to obtain an initial grayscale image; denoising the initial grayscale image to obtain a denoised grayscale image; performing texture enhancement processing on the denoised grayscale image using a preset directional filter to obtain a directionally enhanced image; and performing grayscale normalization processing on the directionally enhanced image to obtain a target grayscale image.

[0006] In one possible design, determining the critical threshold of the interference peak based on the target grayscale image includes: obtaining an initial critical threshold preset based on historical data; determining the environmental interference intensity of the target barley growing area based on the target grayscale image; and correcting the initial critical threshold based on the environmental interference intensity to obtain the critical threshold of the interference peak.

[0007] In one possible design, determining the environmental interference intensity of the target barley growing area based on the target grayscale image includes: dividing the target grayscale image into multiple non-overlapping regular image sub-regions; for each image sub-region, determining the variance of the grayscale values ​​of all pixels in the image sub-region as the local grayscale variance of the image sub-region; determining a segmentation threshold using an adaptive threshold segmentation algorithm based on the local grayscale variances of all image sub-regions; determining the number of image sub-regions whose local grayscale variance is greater than the variance segmentation threshold as the number of abnormal sub-regions; and determining the environmental interference intensity based on the number of abnormal sub-regions and the total number of image sub-regions.

[0008] In one possible design, generating a gradient orientation histogram based on the target grayscale image includes: calculating the gradient direction of each pixel in the target grayscale image using an edge detection operator; uniformly dividing the range of all gradient directions into multiple continuous direction intervals; determining the number of pixel gradient directions contained in each direction interval, and generating a gradient orientation histogram.

[0009] In one possible design, multiple effective slope values ​​are determined based on the critical threshold of the interference peak and the peak information in the gradient direction histogram. This includes: identifying the two directional intervals with the highest count values ​​in the gradient direction histogram as the first and second main peak intervals; determining the target directional interval among multiple intervals between the first and second main peak intervals; determining the first effective slope value based on the peak point of the first main peak interval and the peak point of the target directional interval; determining the second effective slope value based on the peak point of the target directional interval and the peak point of the second main peak interval; and for every two adjacent directional intervals in the gradient direction histogram, excluding the directional interval between the first and second main peak intervals, determining a third effective slope value based on the peak points of the two adjacent directional intervals. The multiple effective slope values ​​include the first effective slope value, the second effective slope value, and all third effective slope values.

[0010] In one possible design, the target direction interval is determined among multiple intervals between the first and second main peak intervals, including: for each candidate interval between the first and second main peak intervals, determining a first slope from the peak point of the first main peak interval to the peak point of the candidate interval, and a second slope from the peak point of the candidate interval to the peak point of the second main peak interval; determining a first mean based on the absolute values ​​of the first and second slopes; determining a second mean based on the absolute values ​​of the slopes corresponding to all adjacent direction intervals between the first and second main peak intervals; determining an initial slope difference corresponding to the candidate interval based on the first and second mean; determining a slope smoothness based on the environmental interference intensity and the critical threshold of the interference peak, whereby the slope smoothness characterizes the degree of neutralization effect of the interference peak on the calculation of the inter-peak slope; compensating the initial slope difference based on the slope smoothness to obtain a compensated slope difference; and determining the candidate interval with the largest compensated slope difference as the target direction interval.

[0011] In one possible design, the texture enhancement processing of the denoised grayscale image is performed by a preset directional filter, including: constructing at least two directional Gaussian kernels corresponding to the field ridge direction of the target barley growth area; convolving each directional Gaussian kernel with the denoised grayscale image to obtain the corresponding convolution result; and weighting and fusing the convolution results to obtain the directional enhancement image.

[0012] In one possible design, obtaining an initial critical threshold based on historical data includes: acquiring multiple sets of historical barley images labeled with growth status; constructing a test image sequence with different numbers of simulated interference peaks for each set of historical barley images; determining the similarity between the slope variance of the gradient direction histogram of normal growth samples and lodged growth samples under each number of simulated interference peaks; determining the number of simulated interference peaks that maximizes the similarity based on the relationship between the similarity and the number of simulated interference peaks, as the initial critical threshold for a single fitting; repeating the fitting process multiple times and determining the mean of all initial critical thresholds for single fitting as the final initial critical threshold.

[0013] In one possible design, determining the preset growth baseline value includes: acquiring multiple sets of barley sample images in a normal growth state under low temperature conditions; determining the sample slope variance corresponding to each set of barley sample images; and determining the mean of the multiple sample slope variances as the preset growth baseline value.

[0014] The present invention has the following beneficial effects: Understandably, the barley growth status identification method under low-temperature disaster environments provided in this embodiment of the invention addresses the core issues of insignificant differences in image features between lodged and normally growing barley under low-temperature disaster scenarios, and susceptibility to interference factors such as weeds and plant damage. By dynamically adapting the critical threshold of interference peaks and accurately screening effective slope values, the method effectively eliminates irrelevant interference from the identification results, significantly improving the accuracy of growth status identification. Simultaneously, by constructing a judgment standard using the quantitative indicator of slope variance, the method avoids subjective judgment bias, making the identification results more objective and reliable. The entire method requires no complex hardware support and can be implemented using conventional UAV image acquisition equipment and digital image processing technology. It is adaptable to barley planting scenarios with different low-temperature intensities, different planting terrains, and different growth stages on plateaus, enabling rapid and accurate identification of lodged growth status. This provides scientific and reliable technical support for timely disaster intervention measures in agricultural production and reducing losses caused by low-temperature disasters, demonstrating strong practical application value. Attached Figure Description

[0015] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a flowchart illustrating a method for identifying the growth status of highland barley under low-temperature disaster conditions, provided in one embodiment of the present invention. Figure 2 This is a schematic diagram of a barley growth status identification system under low-temperature disaster environment provided in an embodiment of the present invention. Detailed Implementation

[0017] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of a method for identifying the growth status of highland barley under low-temperature disaster environments proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0018] In embodiments of the present invention, the terms "exemplary" or "for example" are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" or "for example" in embodiments of the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the terms "exemplary" or "for example" is intended to present the relevant concepts in a specific manner.

[0019] In the description of this invention, unless otherwise stated, " / " means "or". For example, A / B can mean A or B. The term "and / or" in this document is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, and B alone. Furthermore, "at least one" and "more than one" refer to two or more. The terms "first," "second," etc., do not limit the quantity or order of execution, and "first," "second," etc., do not necessarily imply differences.

[0020] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0021] The following description, in conjunction with the accompanying drawings, details the specific scheme of the method for identifying the growth status of highland barley under low-temperature disaster environments provided by the present invention.

[0022] Please see Figure 1 The diagram shows a flowchart of a method for identifying the growth status of highland barley under low-temperature disaster conditions according to an embodiment of the present invention, including the following steps S101-S106.

[0023] S101. Obtain the original image of the target barley growing area and preprocess the original image to obtain the target grayscale image with enhanced texture.

[0024] As one possible approach, the first step is image acquisition. A drone equipped with a high-resolution RGB camera is selected, and aerial photography parameters such as flight altitude, flight speed, and image overlap are set according to actual operational needs. The target barley growing area is then photographed during periods of suitable lighting conditions to obtain raw RGB color images.

[0025] Next, a preprocessing procedure is performed on the acquired original RGB image to obtain a target grayscale image with enhanced texture features.

[0026] The preprocessing process includes converting the original image to grayscale to obtain an initial grayscale image.

[0027] Optionally, a weighted average method is used to linearly combine the brightness values ​​of the red, green, and blue channels of each pixel in the original RGB image according to preset weights, calculate its grayscale value, and thus convert the color image into an initial grayscale image.

[0028] The initial grayscale image is denoised to obtain the denoised grayscale image.

[0029] Optionally, to suppress image noise, the initial grayscale image is subjected to two-stage filtering. First, a medium-range filter is applied to remove salt-and-pepper noise from the image; then, a Gaussian filter is applied to smooth the image to filter out local minor fluctuations, resulting in a denoised grayscale image.

[0030] A texture enhancement process is performed on the denoised grayscale image using a preset directional filter to obtain a directionally enhanced image.

[0031] In some embodiments, to highlight the texture features characterizing the directionality of barley growth, directional filtering enhancement is applied to the denoised grayscale image. Specifically, a directional Gaussian kernel adapted to the direction of the ridges in the target barley growth area is constructed. Taking a typical embodiment as an example, for orthogonal ridges, directional angle parameters can be constructed separately. and Two Gaussian kernels in each direction are used. These two Gaussian kernels are convolved with the denoised grayscale image to obtain the convolution results in the two directions. Then, the two convolution results are weighted and fused, that is, the values ​​of the corresponding pixels are linearly combined according to the preset weight coefficients (for example, both can be set to 0.5, or can be adjusted according to the importance of the actual field ridge direction), thereby generating a directionally enhanced image that integrates texture information from multiple dominant directions.

[0032] The direction-enhanced image is subjected to grayscale normalization to obtain the target grayscale image.

[0033] Optionally, a linear transformation is performed on the pixel grayscale values ​​of the orientation-enhanced image to normalize its grayscale dynamic range to the standard range of 0 to 255, resulting in the final target grayscale image with enhanced texture features for subsequent analysis.

[0034] Understandably, this embodiment of the invention efficiently converts the original RGB color image into a single grayscale image. While reducing data complexity, it effectively suppresses salt-and-pepper noise and random fluctuations introduced during image acquisition through targeted filtering operations. Furthermore, it utilizes a directional filter adapted to the field ridge direction to enhance key texture information reflecting the arrangement and growth direction of barley plants. Finally, normalization processing ensures that the output image has a consistent grayscale range, laying a solid foundation for the stable calculation and reliable analysis of subsequent gradient direction features. This overall improves the efficiency of the subsequent growth status recognition algorithm in utilizing the original image information and its anti-interference capability.

[0035] S102. Determine the critical threshold of the interference peak based on the target grayscale image.

[0036] Among them, the critical threshold of the interference peak is used to characterize the number of interference peaks in the gradient direction histogram.

[0037] As one possible approach, an initial critical threshold is obtained based on historical data.

[0038] It should be noted that the initial critical threshold preset based on historical data is the critical value of the number of interference peaks that distinguishes between normal growth and lodging growth when there is no environmental interference, obtained by fitting a large amount of historical barley image data.

[0039] In some embodiments, multiple sets of historical barley images with growth status annotations are first acquired. This includes collecting multiple sets of images of barley growing areas in high-altitude low-temperature environments as historical samples. The samples need to cover different low-temperature intensity levels (light frost, moderate frost, cold wave), different barley growth stages (jointing stage, heading stage, grain-filling stage), and different planting terrains (plains, gentle slopes, mountains) to ensure the diversity and representativeness of the samples. Each set of historical images is annotated with its growth status through manual field surveys, clearly labeled as "normal growth sample" or "lodged growth sample," ultimately forming a historical dataset containing more than 500 annotated samples.

[0040] Furthermore, for each set of historical barley images, test image sequences with different numbers of simulated interference peaks were constructed. This included, for each sample, first obtaining a historical target grayscale image according to the preprocessing procedure described in the previous embodiment, then generating random Gaussian noise blocks by computer, and progressively superimposing noise of different intensities onto the historical target grayscale image to generate a series of test image sequences with the number of simulated interference peaks n increasing from 1 to 20. The simulated interference peaks were used to simulate peaks not related to the growth direction of barley caused by factors such as weeds, soil particles, and image transmission noise in real-world scenarios. The noise intensity was controlled by the standard deviation of the Gaussian noise (standard deviation range 0.1 to 1.0), and the number of interference peaks n was the statistical number of independently generated simulated interference peaks in each set of test images.

[0041] Further investigation was conducted to determine the similarity between the slope variance of the gradient direction histogram of normally growing samples and fallen growing samples for each number of simulated interference peaks. This included calculating the slope variance of the gradient direction histogram for each image in the simulation sequence, for each specific number of interference peaks. Calculate the values ​​of all “normal growth” samples here. The average slope variance and the mean slope variance of all “lodging growth” samples. Next, the similarity between these two variances is calculated. The calculation formula is as follows: In the formula, The denominator represents the absolute value of the difference in slope variance between the two classes of samples. Used to achieve normalization of molecules. It is a very small positive number, and an empirical value of 0.001 can be taken to prevent the denominator from being zero. The value range is [0,1]. The larger the value, the greater the number of interference peaks. The closer the slope variance characteristics of normal and lodged slopes are, the more difficult it is to distinguish them. The smaller the value, the greater the difference between the two and the easier it is to distinguish them.

[0042] Further, based on the relationship between similarity and the number of simulated interference peaks, the number of simulated interference peaks that maximizes similarity is determined as the initial critical threshold for a single fitting iteration. This fitting process is repeated multiple times, and the average of all initial critical thresholds for single fitting iterations is used as the final initial critical threshold. This includes using the number of simulated interference peaks in each set of test images. As the independent variable, the similarity is calculated based on the corresponding slope and variance. Using the dependent variable, a data mapping relationship is constructed. This mapping relationship is then fitted using a third-order polynomial and the least squares method to obtain the fitted polynomial. This polynomial reflects the influence of changes in the number of interference peaks on the difficulty of distinguishing between normal and lodged growth samples. By differentiating the fitted polynomial and setting the derivative to zero, the number of interference peaks corresponding to a derivative of zero can be obtained. This value is the initial critical threshold for fitting a single set of historical samples (at which point the similarity is...). (Reaching the maximum value makes differentiation the most difficult). To avoid the randomness of a single fit, the above fitting process is repeated for all groups of samples in the historical dataset, resulting in multiple initial critical thresholds for single fits. The average of all single fit results is then taken as the final preset initial critical threshold. .

[0043] Furthermore, the intensity of environmental disturbance in the target barley growing area is determined based on the target grayscale image.

[0044] In some embodiments, the target grayscale image to be analyzed is divided into multiple non-overlapping regular image sub-regions (e.g., divided into a 7×7 grid). For each image sub-region, the variance of all pixel grayscale values ​​within the sub-region is calculated, and this variance is determined as the local grayscale variance of the sub-region. The distribution of the local grayscale variances of all image sub-regions in the entire image is statistically analyzed, and a segmentation threshold is adaptively determined using Otsu's method. Sub-regions with local grayscale variances greater than this segmentation threshold are marked as abnormal sub-regions, and the number of marked abnormal sub-regions is obtained. The ratio of the number of abnormal sub-regions to the total number of image sub-regions is then determined as the environmental interference intensity. Environmental disturbance intensity It is used to quantify the degree of irregular grayscale disorder caused by factors such as low-temperature disasters in the current image.

[0045] Finally, the initial critical threshold is corrected based on the intensity of environmental interference to obtain the critical threshold of the interference peak.

[0046] In some embodiments, the "environmental disturbance intensity" is based on a pre-established value learned from a large number of labeled samples. "and the number of additional weak peaks in the resulting gradient direction histogram" The mapping relationship between them (which is usually a positively correlated monotonic function) is based on the currently calculated... The number of corresponding environmental disturbance peaks can be obtained by looking up a table or calculating through a fitting function. This leads to the initial critical threshold. Subtract the number of environmental interference peaks corresponding to the current target grayscale image This yields the corrected critical threshold for interference peaks. The critical threshold for interference peaks characterizes the number of valid peak values ​​and interference peak values ​​that must be distinguished in the actual scenario of the target barley growing area. Its formula is expressed as: In addition, to ensure The validity of this can be determined by setting its lower bound to 1, i.e. .

[0047] Understandably, this embodiment of the invention achieves dynamic threshold correction and scene adaptation by combining an initial threshold pre-trained based on historical data with the intensity of environmental interference evaluated in real time for the current target image. This method not only utilizes historical prior knowledge to establish a stable discrimination benchmark, but more importantly, it quantifies the degree of irregular grayscale disorder (i.e., environmental interference intensity) caused by low-temperature disasters in the current image and adjusts the benchmark threshold accordingly, effectively overcoming the shortcomings of fixed thresholds in adapting to different disaster levels and field environments. This strategy, combining "global prior" and "local adaptation," significantly improves the accuracy and robustness of the critical parameter of the interference peak threshold in complex real-world application scenarios, thus providing crucial support for ensuring the reliability of subsequent effective slope value extraction and final growth status determination.

[0048] S103. Generate a gradient direction histogram based on the target grayscale image.

[0049] As one possible implementation, an edge detection operator is used to calculate the gradient direction of each pixel in the target grayscale image.

[0050] In some embodiments, a classical edge detection operator (such as the Sobel operator) is used to perform convolution operations on the image to obtain approximate gradient values ​​for each pixel in the horizontal (x-axis) and vertical (y-axis) directions, respectively. and Then, the gradient direction angle of the pixel is calculated based on these two components. Its formula is expressed as ,in, This is a two-parameter arctangent function, whose return value is typically in the range of (-π, π] or [0, 2π) radians, representing the direction in which the image grayscale changes most drastically at that point. This refers to the gradient direction of a pixel.

[0051] Furthermore, the entire gradient direction range is uniformly divided into multiple continuous directional intervals. Specifically, the entire angular range (e.g., 0° to 360°) covered by the gradient direction values ​​of all pixels is uniformly divided into... A series of consecutive directional intervals (Bins). The angular span of each directional interval. .

[0052] Then, iterate through all valid pixels in the image and calculate their gradient directions. This is then assigned to the corresponding direction interval. For a given gradient direction... The corresponding direction interval index It can be calculated using the following formula: In the formula, This indicates the floor function. Indicates taking Divide by The remainder, For modulo notation, since the gradient direction is periodic (i.e., 0° and 360° are equivalent), to ensure that all gradient directions can be correctly mapped to [0, ... Within the range of the interval index [-1], for Perform the modulo operation (i.e., calculate the division by) The remainder after (the remainder), thus seamlessly mapping the entire circumferential direction to In a discrete statistical interval.

[0053] Further statistics on each directional interval (in The number of gradient directions of the pixels contained within a given interval is the count value for that interval. Finally, the interval is indexed by direction. Using the central angle value (or its corresponding value) as the x-axis and the count value of the corresponding interval as the y-axis, the resulting statistical chart is the gradient direction histogram of the target grayscale image. This histogram intuitively reflects the overall directional distribution of all edges or texture structures in the target grayscale image. The intervals with higher count values ​​correspond to the dominant texture directions in the image, which are the basic data for identifying "peak intervals" in subsequent steps.

[0054] S104. Based on the critical threshold of the interference peak and the peak information in the gradient direction histogram, determine multiple effective slope values.

[0055] Among them, multiple effective slope values ​​are used to characterize key transition features of image texture direction.

[0056] One possible implementation involves first iterating through the count values ​​of all directional intervals in the gradient orientation histogram. The two directional intervals with the highest count values ​​are then identified as the first and second dominant peak intervals, which typically correspond to the most prominent texture-dominant directions in the image. All directional intervals located between these two dominant peak intervals are considered candidate intervals. Target directional intervals are then determined from among these candidate intervals; these target directional intervals are the points selected from the candidate intervals that best represent the "true" directional transition between the two dominant peaks.

[0057] In some embodiments, for the directional intervals between the first main peak interval and the second main peak interval, there may be invalid intervals with too low count values. Before performing subsequent calculations, each directional interval between the first main peak interval and the second main peak interval can be traversed by using a pre-set second preset threshold (e.g., an empirical value of 10), and directional intervals with count values ​​less than the second preset threshold can be eliminated, leaving the remaining directional intervals as candidate intervals.

[0058] In some embodiments, for each candidate interval (denoted as the ), (1 candidate interval), determine the first slope from the peak point of the first main peak interval to the peak point of the candidate interval, denoted as . And the second slope from the peak point of the candidate interval to the peak point of the second main peak interval. The first mean is determined based on the absolute values ​​of the first and second slopes. Its formula is expressed as: .

[0059] Furthermore, based on the absolute values ​​of the slopes of all adjacent directional intervals between the first and second main peak intervals, a second mean is determined, denoted as . Based on the first mean and the second mean, the initial slope difference corresponding to the candidate interval is determined. Its formula is expressed as This difference reflects the difference between the local transition steepness and the global average transition level when this candidate interval is used as a transition point. The larger the difference, the more likely the candidate interval is to be a significant transition "trough".

[0060] The slope smoothness is further determined based on the environmental interference intensity and the critical threshold of the interference peak. The slope smoothness characterizes the degree to which the interference peak neutralizes the calculation of the inter-peak slope. The formula for calculating the slope smoothness is as follows: In the formula, For the first The slope flatness corresponding to each candidate interval A preset constant greater than 0 (e.g.) ), used to control the maximum possible magnification, For environmental disturbance intensity, The critical threshold for interference peaks; For the first The local neighborhood variance corresponding to the nth candidate interval, such as the first... Each candidate interval is on the left and right. (like Each directional interval constitutes its local neighborhood. The variance of all gradient directions within this neighborhood is then calculated as the local neighborhood variance. The local neighborhood variance directly reflects the degree of dispersion of the local directional distribution. The larger the variance, the more chaotic the region may be and the greater the impact of disturbance. The saturation parameter is a small positive constant (e.g., ...). ), used to prevent local neighborhood variance When it reaches its maximum, the value of the fraction increases infinitely, ensuring... There exists an upper bound, and avoidance when The denominator is zero to ensure computational stability.

[0061] Among them, the fractional terms in the above formula Integrating global interference levels ( ) and local disorder ( When the local neighborhood variance Compared to When the value is small, the fractional value is approximately equal to... Proportional to local changes; when When it is very large, the value of the fraction approaches The saturation point is primarily determined by global parameters. Ultimately, The range is [1, ]. The value is always greater than or equal to 1, and increases with the increase of global interference ( Increased) or increased local disorder ( (Increases) and increases monotonically.

[0062] Furthermore, the initial slope difference is compensated based on the slope smoothness to obtain the compensated slope difference. For each candidate interval... The slope difference after compensation is calculated as follows: The formula passes through As a multiplier, it compensates for the initial slope difference, especially in cases of severe interference. Large) or locally disordered candidate intervals ( In the large directional range, the true difference signal may be severely "neutralized," thus requiring a larger amplification factor ( The amplification effect is weaker in regions with minor interference and local stability, and more significant interference is used to restore its significance. Approximately 1). Finally, compare all candidate intervals. ,Will The candidate interval with the largest value is determined as the target direction interval.

[0063] Furthermore, the effective slope values ​​are calculated, including the first effective slope value, the second effective slope value, and the third effective slope value.

[0064] The first effective slope value is the absolute value of the slope from the peak point of the first main peak interval to the peak point of the target direction interval; the second effective slope value is the absolute value of the slope from the peak point of the target direction interval to the peak point of the second main peak interval; and the third effective slope value is the absolute value of the slope of every two adjacent directional intervals excluding the first main peak interval, the second main peak interval, and the intervals in between in the gradient direction histogram. The third effective slope value reflects the transition of other local directions in the non-main bimodal regions of the image.

[0065] S105. Determine the slope variance based on multiple effective slope values.

[0066] Among them, the slope variance is used to characterize the degree of fluctuation and dispersion of the gradient direction slope.

[0067] As one possible implementation, the variance is calculated based on the first effective slope value, the second effective slope value, and the third effective slope value obtained from the above steps. The calculated variance is the slope variance corresponding to the target grayscale image. .

[0068] In some embodiments, let the total number of valid slope values ​​obtained be... , let them be a sequence .in, These correspond to the first effective slope value and the second effective slope value. This corresponds to all third effective slope values.

[0069] slope variance The calculation uses the standard statistical variance formula, as follows: First, the arithmetic mean of multiple effective slope values. , , The number of valid slope values ​​must be greater than zero. For the sequence corresponding to multiple effective slope values, the first one is... One effective slope value. The slope variance is then calculated using the standard statistical variance formula. , Calculated The value is the slope variance.

[0070] S106. When the difference between the slope variance and the preset growth benchmark value exceeds the first preset threshold, the growth status of barley in the target barley growth area is determined to be lodging growth.

[0071] It should be noted that the preset growth baseline value is a standard reference value for the slope variance of highland barley under normal growth conditions at low temperatures. Its determination needs to be based on a large number of representative normal growth samples to ensure the reliability and universality of the baseline value. The specific process is as follows: Multiple sets of images of barley samples under normal growth conditions in high-altitude low-temperature environments were collected. The samples needed to meet diversity requirements: covering different levels of low-temperature intensity (mild frost, moderate frost), different barley growth stages (jointing stage, heading stage, grain-filling stage), different planting terrains (plains, gentle slopes, mountains), and different ridge distribution patterns (orthogonal, parallel). The total number of samples was no less than 30 sets to avoid baseline deviations caused by samples from a single scene. All samples were confirmed through manual field surveys to be free of lodging, with regular plant growth direction and no obvious leaf damage or wilting tissue, ensuring that the samples represented purely normal growth scenes.

[0072] For each group of normal growth sample images, following the complete workflow of the aforementioned embodiment (image preprocessing → determining the critical threshold of interference peaks → generating gradient direction histograms → extracting effective slope values ​​→ calculating slope variance), the sample slope variance for each group of samples is calculated respectively. ( , (Number of sample groups). This process strictly follows a unified calculation standard to ensure that the calculation logic of the slope variance of all samples is consistent with the slope variance corresponding to the target grayscale image. Consistency is maintained to avoid the invalidation of benchmark values ​​due to calculation discrepancies.

[0073] To eliminate the interference of extreme samples on the benchmark value, a pre-defined outlier removal algorithm is used to filter the slope variance of all samples. This is achieved by calculating the mean of the slope variance of all samples. and standard deviation The range for outlier detection is set as [ Samples with slope variances outside this range are identified as outliers and removed. If the number of sample groups after removal is less than 20, additional samples are collected and the above steps are repeated to ensure that the remaining samples have sufficient statistical representativeness.

[0074] Finally, the mean of the slope variance of the remaining samples after removing outliers is calculated to obtain the preset growth baseline value. .

[0075] As one possible implementation, the difference between the slope variance of the target grayscale image and a preset growth baseline value is determined. Its formula is expressed as In the formula, The slope variance corresponding to the target grayscale image. To preset the growth baseline value, It is a very small positive number, and an empirical value of 0.001 can be taken to prevent the denominator from being zero. It is used to reflect the degree of deviation of the slope variance corresponding to the target grayscale image from the preset growth baseline value.

[0076] Then, if the difference between the slope variance and the preset growth benchmark value exceeds the first preset threshold (for example, an empirical value of 1 can be taken), the growth status of barley in the target barley growth area is determined to be lodging growth.

[0077] If the difference exceeds a first preset threshold, it indicates the slope variance of the current target grayscale image. Significantly greater than the normal growth baseline value (More than double). This means that the gradient slope of the barley image in this area fluctuates abnormally sharply, and the transition of the texture direction is extremely messy and disordered. This is a typical feature of the plant stem losing regular support and growing lodging. Under these circumstances, the growth trend of the barley in the target barley growth area is determined to be lodging growth.

[0078] Understandably, the barley growth status identification method under low-temperature disaster environments provided in this embodiment of the invention addresses the core issues of insignificant differences in image features between lodged and normally growing barley under low-temperature disaster scenarios, and susceptibility to interference factors such as weeds and plant damage. By dynamically adapting the critical threshold of interference peaks and accurately screening effective slope values, the method effectively eliminates irrelevant interference from the identification results, significantly improving the accuracy of growth status identification. Simultaneously, by constructing a judgment standard using the quantitative indicator of slope variance, the method avoids subjective judgment bias, making the identification results more objective and reliable. The entire method requires no complex hardware support and can be implemented using conventional UAV image acquisition equipment and digital image processing technology. It is adaptable to barley planting scenarios with different low-temperature intensities, different planting terrains, and different growth stages on plateaus, enabling rapid and accurate identification of lodged growth status. This provides scientific and reliable technical support for timely disaster intervention measures in agricultural production and reducing losses caused by low-temperature disasters, demonstrating strong practical application value.

[0079] Please see Figure 2 The diagram illustrates a structural schematic of a barley growth status identification system under low-temperature disaster conditions, as provided in an embodiment of the present invention. Figure 2 As shown, the barley growth status recognition system 20 under low temperature disaster environment includes an image acquisition unit 21, a threshold determination unit 22, a feature extraction unit 23, and a growth status determination unit 24.

[0080] The image acquisition unit 21 is used to acquire the original image of the target barley growing area and preprocess the original image to obtain the target grayscale image with enhanced texture. The specific implementation can be referred to the description in step S101 in the above embodiment, and will not be repeated here.

[0081] The threshold determination unit 22 is used to determine the critical threshold of the interference peak based on the target grayscale image. Its specific implementation can be referred to the description in step S102 of the above embodiment, and will not be repeated here.

[0082] The feature extraction unit 23 is used to generate a gradient direction histogram based on the target grayscale image, and to determine multiple effective slope values ​​based on the critical threshold of the interference peak and the peak information in the gradient direction histogram. Its specific implementation can be referred to the description in steps S103 and S104 in the above embodiment, and will not be repeated here.

[0083] The growth status determination unit 24 is used to determine the slope variance based on multiple effective slope values, and when the difference between the slope variance and the preset growth benchmark value exceeds the first preset threshold, it determines that the growth status of barley in the target barley growth area is lodging growth. The specific implementation can be referred to the description in steps S105 and S106 in the above embodiment, and will not be repeated here.

[0084] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0085] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

Claims

1. A method for identifying the growth status of highland barley under low-temperature disaster conditions, characterized in that, The method includes: The original image of the target barley growing area is acquired, and the original image is preprocessed to obtain a target grayscale image with enhanced texture. The critical threshold for interference peaks is determined based on the target grayscale image, and the critical threshold for interference peaks is used to characterize the number of interference peaks in the gradient direction histogram. A gradient orientation histogram is generated based on the target grayscale image, and multiple effective slope values ​​are determined based on the critical threshold of the interference peak and the peak information in the gradient orientation histogram. The multiple effective slope values ​​are used to characterize the key transition features of the image texture direction. The slope variance is determined based on the plurality of effective slope values, and the slope variance is used to characterize the degree of fluctuation and dispersion of the gradient direction slope; If the difference between the slope variance and the preset growth benchmark value exceeds a first preset threshold, the barley growth pattern of the target barley growth area is determined to be lodging growth. Determining the critical threshold of the interference peak based on the target grayscale image includes: Obtain the initial critical threshold based on historical data; The intensity of environmental disturbance in the target barley growing area is determined based on the target grayscale image; The initial critical threshold is corrected based on the intensity of the environmental interference to obtain the critical threshold of the interference peak; Obtain the initial critical threshold based on historical data, including: Acquire multiple sets of historical barley images with growth status annotations; For each set of historical barley images, a test image sequence with a different number of simulated interference peaks was constructed; Determine the similarity between the slope variance of the gradient direction histogram of normal growth samples and lodged growth samples for each number of simulated interference peaks; Based on the relationship between the similarity and the number of simulated interference peaks, the number of simulated interference peaks that makes the similarity reach the maximum value is determined as the initial critical threshold for a single fitting. The fitting process is repeated multiple times, and the mean of all the initial critical thresholds for each single fitting is determined as the final initial critical threshold. Based on the critical threshold of the interference peak and the peak information in the gradient direction histogram, multiple effective slope values ​​are determined, including: The two directional intervals with the highest count values ​​in the gradient direction histogram are determined as the first main peak interval and the second main peak interval. Determine the target direction interval among multiple intervals between the first main peak interval and the second main peak interval; The first effective slope value is determined based on the peak point of the first main peak interval and the peak point of the target direction interval; The second effective slope value is determined based on the peak point of the target direction interval and the peak point of the second main peak interval; For every two adjacent directional intervals in the gradient direction histogram, excluding the directional interval between the first main peak interval and the second main peak interval, a third effective slope value is determined based on the peak points of the two adjacent directional intervals. The plurality of effective slope values ​​include the first effective slope value, the second effective slope value, and all of the third effective slope values; Determining the target direction interval among multiple intervals between the first main peak interval and the second main peak interval includes: For each candidate interval between the first main peak interval and the second main peak interval, a first slope from the peak point of the first main peak interval to the peak point of the candidate interval is determined, and a second slope from the peak point of the candidate interval to the peak point of the second main peak interval is determined. The first mean is determined based on the absolute values ​​of the first slope and the second slope. The second mean is determined based on the absolute values ​​of the slopes of all adjacent directional intervals between the first and second main peak intervals. Based on the first mean and the second mean, determine the initial slope difference value corresponding to the candidate interval; The slope smoothness is determined based on the environmental interference intensity and the critical threshold of the interference peak. The slope smoothness is used to characterize the degree of neutralization effect of the interference peak on the calculation of the inter-peak slope. The initial slope difference is compensated based on the slope smoothness to obtain the compensated slope difference. The candidate interval with the largest slope difference after compensation is determined as the target direction interval.

2. The method for identifying the growth status of highland barley under low-temperature disaster environment according to claim 1, characterized in that, Preprocessing the original image includes: The original image is converted to grayscale to obtain an initial grayscale image; The initial grayscale image is subjected to noise reduction processing to obtain a noise-reduced grayscale image; A texture enhancement process is performed on the denoised grayscale image using a preset directional filter to obtain a directionally enhanced image. The direction-enhanced image is subjected to grayscale normalization processing to obtain the target grayscale image.

3. The method for identifying the growth status of highland barley under low-temperature disaster environment according to claim 1, characterized in that, Determining the environmental disturbance intensity of the target barley growing area based on the target grayscale image includes: The target grayscale image is divided into multiple non-overlapping regular image sub-regions; For each image sub-region, the variance of the gray values ​​of all pixels in the image sub-region is determined as the local gray variance of the image sub-region; Based on the local gray-level variance of all image sub-regions, the segmentation threshold is determined by an adaptive threshold segmentation algorithm. The number of image sub-regions whose local gray-level variance is greater than the variance segmentation threshold is determined as the number of abnormal sub-regions. The intensity of environmental interference is determined based on the number of abnormal sub-regions and the total number of image sub-regions.

4. The method for identifying the growth status of highland barley under low-temperature disaster environment according to claim 1, characterized in that, Generating a gradient orientation histogram based on the target grayscale image includes: The gradient direction of each pixel in the target grayscale image is calculated using an edge detection operator; The entire range of gradient directions is uniformly divided into multiple continuous directional intervals; The number of gradient directions of pixels contained in each of the directional intervals is determined, and the gradient direction histogram is generated.

5. The method for identifying the growth status of highland barley under low-temperature disaster environment according to claim 2, characterized in that, Texture enhancement processing is performed on the denoised grayscale image using a preset directional filter, including: Construct at least two Gaussian kernels corresponding to the ridge direction of the target barley growing area; Each Gaussian kernel in the specified direction is convolved with the noise-reduced grayscale image to obtain the corresponding convolution result. The convolution results are weighted and fused to obtain the orientation-enhanced image.

6. The method for identifying the growth status of highland barley under low-temperature disaster environment according to claim 1, characterized in that, Determining the preset growth benchmark value includes: Multiple images of highland barley samples in a normal growth state under low temperature conditions were collected; Determine the sample slope variance corresponding to each group of highland barley sample images; The mean of the slope variances of multiple samples is determined as the preset growth baseline value.

Images