Planting information monitoring method and system based on machine vision

Abstract

The invention discloses a planting information monitoring method and system based on machine vision, and belongs to the technical field of agricultural informationization, and the method comprises the steps: S1, obtaining image data of crop leaves, and collecting real-time illumination intensity data through an ambient light sensor array; s2, calculating a reflectivity compensation factor of the surface of the blade according to the real-time illumination intensity data; s3, performing fusion processing on the plurality of reflectivity compensation factors and the image data to generate compensated leaf color feature data; and S4, analyzing the compensated leaf color feature data through an adaptive feature compensation network to obtain a crop nutrition state evaluation result. According to the method, the problem of inaccurate leaf color feature extraction under the dynamic illumination condition is effectively solved by constructing the non-linear compensation model of the illumination intensity and the reflectivity and combining the ambient light sensor array and the self-adaptive feature compensation network, and the precision and the reliability of a monitoring system are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the field of agricultural information technology, and in particular to a planting information monitoring method and system based on machine vision. Background Art

[0002] With the rapid development of modern agriculture, machine vision-based crop monitoring technology, due to its high efficiency and accuracy, plays a vital role in assessing crop growth status and collecting environmental parameters. However, existing crop monitoring systems still have significant deficiencies in the accuracy of leaf color feature extraction under dynamic lighting conditions. This can easily lead to misjudgments of nutritional status due to light fluctuations, compromising the effectiveness of crop management decisions.

[0003] After searching, a mobile platform and monitoring method for all-terrain agricultural crop information monitoring with the publication number CN115384643B was disclosed. This patent realizes the multi-station and multi-directional agricultural information collection function through a crawler walking mechanism, a navigation system and an information monitoring device, and can maintain a high level of travel stability on complex road surfaces. However, this technical solution mainly focuses on the stability and applicability design of the hardware platform, and does not involve image acquisition and processing issues under dynamic lighting conditions. In particular, it lacks a compensation mechanism for the relationship between the reflective characteristics of the leaf surface and the light intensity. Therefore, in actual applications, when the external lighting conditions change drastically, it may be difficult for the system to accurately extract the color characteristics of the leaves, thereby affecting the accuracy of judging the nutritional status of the crops.

[0004] After searching, a hydrological information monitoring device, monitoring system, and hydropower station with publication number CN105698773B were disclosed. This patent obtains first hydrological information of the watershed section through a detection device and generates second hydrological information through a processing device to support the operation optimization of the hydropower station. Although this technical solution emphasizes the collection and processing capabilities of sensor data, its core focus is on the monitoring of hydrological information and fails to fully consider the impact of light changes on visual monitoring in agricultural planting scenes. In addition, the solution lacks the design of an illumination compensation model for image data and cannot effectively address the problem of color feature extraction errors under dynamic lighting conditions.

[0005] The above issues demonstrate that existing agricultural information monitoring systems still have shortcomings in extracting leaf color features and determining nutritional status under dynamic lighting conditions. In particular, they lack effective solutions for compensating for the nonlinear relationship between light intensity and reflectivity. Therefore, the present invention provides a machine vision-based planting information monitoring method and system. This approach aims to significantly improve monitoring accuracy under dynamic lighting conditions by constructing a nonlinear compensation model for light intensity and reflectivity, combined with an ambient light sensor array and an adaptive feature compensation network. This approach, in turn, addresses the need for efficient and precise planting information monitoring in modern agriculture. Summary of the Invention

[0006] In order to solve the technical problems existing in the prior art, the present invention provides a planting information monitoring method and system based on machine vision. By constructing a nonlinear compensation model of light intensity and reflectivity, combining an ambient light sensor array and an adaptive feature compensation network, the problem of inaccurate leaf color feature extraction under dynamic lighting conditions is effectively solved, and the accuracy and reliability of the monitoring system are significantly improved.

[0007] In a first aspect, the present invention provides a method for monitoring planting information based on machine vision, the method comprising the following steps: S1, acquiring image data of crop leaves and collecting real-time light intensity data through an ambient light sensor array; S2. Calculating a reflectivity compensation factor of the blade surface according to the real-time light intensity data; S3, fusing the plurality of reflectivity compensation factors with the image data to generate compensated leaf color feature data; S4. Analyzing the compensated leaf color feature data through an adaptive feature compensation network to obtain a crop nutritional status assessment result; S5. Generate a planting management recommendation report based on the nutritional status assessment results.

[0008] As a preferred technical solution of the present invention, the method of acquiring image data of crop leaves and collecting real-time light intensity data through an ambient light sensor array includes the following steps: Using a multispectral camera to capture image data of crop leaves at a fixed frequency; the image data is composed of a plurality of pixel points; By building an ambient light sensor array, we can collect light intensity data of multiple pixel points in real time. The light intensity data is synchronized with the pixel points of the image data by time stamp to form real-time light intensity data; each of the real-time light intensities represents the light intensity data of a single pixel point.

[0009] As a preferred technical solution of the present invention, the calculation of the reflectivity compensation factor of the blade surface according to the real-time light intensity data includes the following steps: Constructing a nonlinear relationship model between the real-time light intensity data and the reflectivity; the core formula of the nonlinear relationship model is a nonlinear mapping function; Set the standard light intensity and reflectivity under the standard light intensity; Inputting the real-time light intensity data into the nonlinear relationship model; The nonlinear mapping function fits the reflectivity to generate a reflectivity compensation factor; The nonlinear mapping function calculation formula is: ; in, is real-time light intensity data, The real-time light intensity data is The reflectivity compensation factor, 、 and These are fitting parameters used to adjust the function’s sensitivity to illumination changes; is the reflectivity under standard light intensity.

[0010] As a preferred technical solution of the present invention, the method of fusing the plurality of reflectivity compensation factors with the image data to generate compensated leaf color feature data comprises the following steps: Performing illumination correction on each pixel of the image data according to the reflectivity compensation factor; Perform color space conversion on the corrected image data, from RGB color space to Lab color space; The a component and the b component in the Lab color space are extracted as compensated leaf color feature data.

[0011] The illumination correction calculation formula is: ; in, is the original pixel value, The real-time light intensity data is The corrected pixel value.

[0012] As a preferred technical solution of the present invention, the analysis of the compensated leaf color feature data by the adaptive feature compensation network includes the following steps: Constructing a training data set, wherein the training data set includes leaf color feature data and nutritional status labels; Dividing the training data set into a training set, a validation set, and a test set; Constructing an adaptive feature compensation network, wherein the adaptive feature compensation network includes a weight matrix and a bias vector; Preliminary learning of model parameters of the adaptive feature compensation network is performed using a training set, and parameter adjustment and model evaluation are performed using a validation set and a test set; The weight matrix and bias vector of the adaptive feature compensation network are updated through a back-propagation algorithm.

[0013] As a preferred technical solution of the present invention, the method of updating the weight matrix and bias vector of the adaptive feature compensation network by back propagation algorithm includes the following steps: S401, initializing a weight matrix and a bias vector; the initialization of the weight matrix needs to adapt to the distribution characteristics of the leaf color features; S402, feature extraction and prediction generation: the leaf color feature data is input into the adaptive feature compensation network; the abstract features of the leaf color feature data are extracted layer by layer through the weight function, and a nutritional status probability score is generated; the nutritional status probability distribution represents the confidence level of the nutritional status category; for example, the probability of nitrogen deficiency is 85%, and the probability of water deficiency is 15%.

[0014] As a preferred technical solution of the present invention, the method of updating the weight matrix and bias vector of the adaptive feature compensation network by the back propagation algorithm further includes the steps of: S403, quantifying the error between the nutritional status probability distribution and the nutritional status label: using a cross entropy loss function to calculate the loss value between the nutritional status probability distribution and the nutritional status label; S404, back-propagation gradient calculation and weight matrix and bias vector update: Based on the chain rule, reversely calculate the partial derivative of the cross-entropy loss function with respect to the weight matrix and bias vector, and then update the weight matrix and bias vector through the gradient descent method, thereby improving the accuracy of mapping the nutritional status of the crop from the leaf color feature data.

[0015] As a preferred technical solution of the present invention, the updating of the weight matrix and bias vector of the adaptive feature compensation network by the back propagation algorithm further includes: The cross entropy loss function calculation formula is: ; The calculation formula for updating the weight matrix is: ; ; The calculation formula for the updated bias vector is: ; ; in, is the loss value of nutritional status probability distribution and nutritional status label, For the The probability distribution value of the nutritional status of the class, For the The value of the nutritional status of the class is manually annotated or experimentally determined, is the total number of nutritional status categories; is the error value of the current layer, For the The activation value of the layer, To update the step size, it is used to prevent oscillation or underfitting; For the The weight matrix of the layer, For the The bias vector of the layer.

[0016] As a preferred technical solution of the present invention, the analysis of the compensated leaf color feature data by the adaptive feature compensation network includes the following steps: Inputting the compensated leaf color feature data into the adaptive feature compensation network to generate a nutritional status probability distribution; The nutritional status probability distribution is mapped to a predefined nutritional status level to obtain a crop nutritional status assessment result.

[0017] In a second aspect, the present invention further provides a planting information monitoring system based on machine vision, which implements the planting information monitoring method based on machine vision, and the system includes an image data acquisition module, a compensation factor calculation module, a feature data generation module, a nutritional status assessment module, and a planting management suggestion module; The image data acquisition module is used to acquire image data of crop leaves and collect real-time light intensity data through an ambient light sensor array; The compensation factor calculation module is used to calculate the blade surface reflectivity compensation factor according to the real-time light intensity data; The feature data generating module is used to fuse the nonlinear compensation factor with the image data to generate compensated leaf color feature data; The nutritional status assessment module is used to analyze the compensated leaf color feature data through an adaptive feature compensation network to obtain a crop nutritional status assessment result; The planting management suggestion module is used to generate a planting management suggestion report based on the nutritional status assessment result.

[0018] Compared with the prior art, the present invention has the following beneficial effects: The present invention constructs a nonlinear compensation model of light intensity and reflectivity, combines an ambient light sensor array and an adaptive feature compensation network, and effectively solves the problem of inaccurate leaf color feature extraction under dynamic lighting conditions, thereby significantly improving the accuracy and reliability of the monitoring system. The ambient light sensor array collects light intensity data in real time and synchronizes it with the image data by time stamp, thereby ensuring the real-time and accuracy of light compensation and avoiding misjudgment of nutritional status due to light changes. At the same time, the adaptive feature compensation network analyzes the compensated leaf color feature data, which can fully explore the complex relationship between leaf color features and crop nutritional status, thereby improving the scientificity and comprehensiveness of nutritional status assessment. In addition, the present invention also ensures the adaptability of the model to different crop types and environmental conditions by training and optimizing the adaptive feature compensation network, further enhancing the practicality and promotion value of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0020] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0021] Figure 1 This is the first flow chart of a planting information monitoring method based on machine vision of the present invention.

[0022] Figure 2 This is the second flow chart of a planting information monitoring method based on machine vision of the present invention.

[0023] Figure 3 It is a structural schematic diagram of a planting information monitoring system based on machine vision of the present invention. DETAILED DESCRIPTION

[0024] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0025] It should be noted that all directional indications (such as up, down, left, right, front, back, etc.) in the embodiments of the present invention are only used to explain the relative position relationship, movement status, etc. between the various components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indication will also change accordingly.

[0026] In addition, the descriptions of "first", "second", etc. in the present invention are for descriptive purposes only and should not be understood as indicating or implying their relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" or "second" may explicitly or implicitly include at least one of such features. In addition, the technical solutions between the various embodiments can be combined with each other, but this must be based on the fact that they can be implemented by ordinary technicians in this field. When the combination of technical solutions is contradictory or cannot be implemented, it should be deemed that such combination of technical solutions does not exist and is not within the scope of protection required by the present invention.

[0027] Example 1 See Figure 1 As shown, the present invention provides a planting information monitoring method based on machine vision, which specifically includes the following steps: S1, obtain image data of crop leaves and collect real-time light intensity data through the ambient light sensor array; S2. Calculating a reflectivity compensation factor of the blade surface according to the real-time light intensity data; S3, fusing the plurality of reflectivity compensation factors with the image data to generate compensated leaf color feature data; S4. Analyzing the compensated leaf color feature data through an adaptive feature compensation network to obtain a crop nutritional status assessment result; S5. Generate a planting management recommendation report based on the nutritional status assessment results.

[0028] Specifically, the method of acquiring image data of crop leaves and collecting real-time light intensity data through an ambient light sensor array includes the following steps: Using a multispectral camera to capture image data of crop leaves at a fixed frequency; the image data is composed of a plurality of pixel points; By building an ambient light sensor array, we can collect light intensity data of multiple pixel points in real time. The light intensity data is synchronized with the pixel points of the image data by time stamp to form real-time light intensity data; each of the real-time light intensities represents the light intensity data of a single pixel point.

[0029] Specifically, the calculation of the reflectivity compensation factor of the blade surface according to the real-time light intensity data includes the following steps: Constructing a nonlinear relationship model between the real-time light intensity data and the reflectivity; the core formula of the nonlinear relationship model is a nonlinear mapping function; Set the standard light intensity and reflectivity under the standard light intensity; Inputting the real-time light intensity data into the nonlinear relationship model; The nonlinear mapping function fits the reflectivity according to the real-time light intensity data to generate a reflectivity compensation factor; The nonlinear mapping function calculation formula is: ; in, is real-time light intensity data, The real-time light intensity data is The reflectivity compensation factor, 、 and These are fitting parameters used to adjust the function’s sensitivity to illumination changes; is the reflectivity under standard light intensity.

[0030] It can be understood that this embodiment decomposes the leaf image data into several pixel points for analysis to improve the detection accuracy of local diseases and nutritional imbalances. The present invention first collects light intensity data of several pixel points through an ambient light sensor array; since the leaf reflectivity is affected by light intensity and the relationship is not a simple linear relationship, a nonlinear relationship model can be constructed to fit the relationship between real-time light intensity data and the reflectivity compensation factor to generate a reflectivity compensation factor. The reflectivity compensation factor reflects the difference between the actual light reflection ability of the leaf surface at this position and the light reflection ability under standard light intensity, and is used to eliminate the influence of ambient light differences on color characteristics; when When , it means that the light intensity is higher than the standard light intensity and the reverse reflectivity needs to be weakened; when When it is on, it means that the light intensity is weaker than the standard light intensity and forward reflectivity compensation is needed.

[0031] Specifically, the fusion processing of the plurality of reflectivity compensation factors with the image data to generate compensated leaf color feature data includes the following steps: Performing illumination correction on each pixel of the image data according to the reflectivity compensation factor; Perform color space conversion on the corrected image data, from RGB color space to Lab color space; The a component and the b component in the Lab color space are extracted as compensated leaf color feature data.

[0032] The illumination correction calculation formula is: ; in, is the original pixel value, The real-time light intensity data is The corrected pixel value.

[0033] It can be understood that in this embodiment, illumination correction is first performed on each pixel of the image data according to the reflectivity compensation factor, and the corrected pixel value replaces the original pixel value; since the three channels of the RGB color space contain both illumination intensity and chromaticity information, color distortion may also occur when the illumination intensity correction is performed on the pixel, and therefore it is necessary to convert the color space of the image data from the RGB color space to the Lab color space, where the Lab color space includes an L component, an a component, and a b component, where the L component is a brightness channel, and the a component and the b component are color information channels, which are used to ensure that color differences are determined only by color information and are not affected by light intensity. degree interference; the a component represents the chromaticity change between green and red. When the leaves are healthy, they reflect more green light and the a component is low. When the leaves are nitrogen-deficient or aged, the chlorophyll degrades, the leaves appear red, and the a component increases; the b component represents the chromaticity change between blue and yellow. When the leaves are healthy, the b component is low because chlorophyll covers carotene. When the leaves are dry or aged, the proportion of carotenoids increases and the b component increases. In addition, fungal infection or insect pests may cause the a component of a local area to be abnormally high; after the Lab color space conversion is completed, the a component and b component in the Lab color space are extracted as the compensated leaf color feature data.

[0034] Specifically, analyzing the compensated leaf color feature data through the adaptive feature compensation network includes the following steps: Constructing a training data set, wherein the training data set includes leaf color feature data and nutritional status labels; Dividing the training data set into a training set, a validation set, and a test set; Constructing an adaptive feature compensation network, wherein the adaptive feature compensation network includes a weight matrix and a bias vector; Using the training set to perform preliminary learning of model parameters of the adaptive feature compensation network, and using the validation set and the test set to perform parameter adjustment and model evaluation; The weight matrix and bias vector of the adaptive feature compensation network are updated through a back-propagation algorithm.

[0035] Based on the foregoing, the weight matrix and bias vector of the adaptive feature compensation network are updated by the back propagation algorithm, including the steps of: S401, initializing a weight matrix and a bias vector; the initialization of the weight matrix needs to adapt to the distribution characteristics of the leaf color features; S402, feature extraction and prediction generation: The leaf color feature data is input into the adaptive feature compensation network; abstract features of the leaf color feature data are extracted layer by layer using a weighting function, and a nutritional status probability distribution is generated; the nutritional status probability distribution represents the confidence level of the nutritional status category; for example, the probability of nitrogen deficiency is 85%, and the probability of water deficiency is 15%; S403, quantifying the error between the nutritional status probability distribution and the nutritional status label: using a cross entropy loss function to calculate the loss value between the nutritional status probability distribution and the nutritional status label; The cross entropy loss function calculation formula is: ; in, is the loss value of nutritional status probability distribution and nutritional status label, For the The probability distribution value of the nutritional status of the class, For the The value of the nutritional status of the class is manually annotated or experimentally determined, is the total number of nutritional status categories; S404, back-propagation gradient calculation and weight matrix and bias vector update: Based on the chain rule, the partial derivative of the cross entropy loss function with respect to the weight matrix and bias vector is reversely calculated. The weight matrix and bias vector are then updated using the gradient descent method, thereby improving the accuracy of mapping the crop nutritional status from the leaf color feature data. The calculation formula for updating the weight matrix is: ; ; The calculation formula for the updated bias vector is: ; ; in, is the error value of the current layer, For the The activation value of the layer, To update the step size, it is used to prevent oscillation or underfitting; For the The weight matrix of the layer, For the The bias vector of the layer; It can be understood that in this embodiment, the leaf color feature data are leaf image features of various leaves under standard light intensity, and the nutritional status label is the nutritional status manually marked or experimentally measured, such as nitrogen deficiency, water deficiency, and fungal interference; 70% of the training data set is divided into a training set for preliminary learning of model parameters; 15% of the training data set is divided into a validation set for adjusting hyperparameters and preventing overfitting; 15% of the training data set is divided into a test set for the final evaluation of the generalization ability of the model; then an adaptive feature compensation network is constructed, and the adaptive feature compensation network is used to construct a mapping relationship between the leaf color feature data and the nutritional status label, the weight matrix reflects the mapping relationship from the input leaf color feature data to the output nutritional status, and the bias vector is used to adjust the baseline value output by the adaptive feature compensation network to enhance the adaptive feature compensation network's ability to fit complex patterns; then the training set is imported into the adaptive feature compensation network for training.

[0036] During the training process, the present invention first initializes the weight matrix and bias vector, and then inputs the leaf color feature data into the adaptive feature compensation network. The adaptive feature compensation network generates a nutritional status probability distribution based on the initial weight matrix and bias vector, and then calculates the loss value of the nutritional status probability distribution and the nutritional status label; based on the loss value, the weight matrix and bias vector are continuously iteratively updated until the nutritional status probability distribution approaches the true nutritional status label. This step significantly improves the monitoring accuracy of leaf color feature data under dynamic lighting conditions by continuously iteratively optimizing the weight matrix and bias vector.

[0037] Specifically, the analyzing the compensated leaf color feature data by the adaptive feature compensation network further includes the following steps: Inputting the compensated leaf color feature data into the adaptive feature compensation network to generate a nutritional status probability distribution; The nutritional status probability distribution is mapped to a predefined nutritional status level to obtain a crop nutritional status assessment result.

[0038] Example 2 The present invention also provides a planting information monitoring system based on machine vision, which implements the above-mentioned planting information monitoring method based on machine vision. The system includes an image data acquisition module, a compensation factor calculation module, a feature data generation module, a nutritional status assessment module and a planting management suggestion module; The image data acquisition module is used to acquire image data of crop leaves and collect real-time light intensity data through an ambient light sensor array; The compensation factor calculation module is used to calculate the blade surface reflectivity compensation factor according to the real-time light intensity data; The feature data generating module is used to fuse the nonlinear compensation factor with the image data to generate compensated leaf color feature data; The nutritional status assessment module is used to analyze the compensated leaf color feature data through an adaptive feature compensation network to obtain a crop nutritional status assessment result; The planting management suggestion module is used to generate a planting management suggestion report based on the nutritional status assessment result.

[0039] The foregoing description is intended only to provide specific embodiments of the present invention, which will enable those skilled in the art to understand and implement the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not intended to be limited to the embodiments shown herein, but is intended to be accorded the widest scope consistent with the principles and novel features claimed herein.

Claims

1. A planting information monitoring method based on machine vision, characterized in that: The following steps are involved: S1, acquiring image data of crop leaves and collecting real-time light intensity data through an ambient light sensor array; S2. Calculating a reflectivity compensation factor of the blade surface according to the real-time light intensity data; S3, fusing the plurality of reflectivity compensation factors with the image data to generate compensated leaf color feature data; S4. Analyzing the compensated leaf color feature data through an adaptive feature compensation network to obtain a crop nutritional status assessment result; S5. Generate a planting management recommendation report based on the nutritional status assessment results.

2. The method for monitoring planting information based on machine vision according to claim 1, characterized in that: The method of acquiring image data of crop leaves and collecting real-time light intensity data through an ambient light sensor array includes the following steps: Using a multispectral camera to capture image data of crop leaves at a fixed frequency; the image data is composed of a plurality of pixel points; By building an ambient light sensor array, we can collect light intensity data of multiple pixel points in real time. The light intensity data is synchronized with the pixel points of the image data by time stamp to form real-time light intensity data; each of the real-time light intensities represents the light intensity data of a single pixel point.

3. The method for monitoring planting information based on machine vision according to claim 1, characterized in that: Calculating the reflectivity compensation factor of the blade surface according to the real-time light intensity data comprises the steps of: Constructing a nonlinear relationship model between the real-time light intensity data and the reflectivity; The core formula of the nonlinear relationship model is a nonlinear mapping function; Set the standard light intensity and reflectivity under the standard light intensity; Inputting the real-time light intensity data into the nonlinear relationship model; The nonlinear mapping function fits the reflectivity to generate a reflectivity compensation factor; The nonlinear mapping function calculation formula is: ； in, is real-time light intensity data, The real-time light intensity data is The reflectivity compensation factor, 、 and These are fitting parameters used to adjust the function’s sensitivity to illumination changes; is the reflectivity under standard light intensity.

4. The method for monitoring planting information based on machine vision according to claim 2, characterized in that: The step of fusing the plurality of reflectivity compensation factors with the image data to generate compensated leaf color feature data comprises the following steps: Performing illumination correction on each pixel of the image data according to the reflectivity compensation factor; Perform color space conversion on the corrected image data, from RGB color space to Lab color space; The a component and the b component in the Lab color space are extracted as compensated leaf color feature data. The illumination correction calculation formula is: ； in, is the original pixel value, The real-time light intensity data is The corrected pixel value.

5. The method for monitoring planting information based on machine vision according to claim 4, characterized in that: The analyzing of the compensated leaf color feature data by the adaptive feature compensation network comprises the following steps: Constructing a training data set, wherein the training data set includes leaf color feature data and nutritional status labels; Dividing the training data set into a training set, a validation set, and a test set; Constructing an adaptive feature compensation network, wherein the adaptive feature compensation network includes a weight matrix and a bias vector; Preliminary learning of model parameters of the adaptive feature compensation network is performed using a training set, and parameter adjustment and model evaluation are performed using a validation set and a test set; The weight matrix and bias vector of the adaptive feature compensation network are updated through a back-propagation algorithm.

6. The method for monitoring planting information based on machine vision according to claim 5, characterized in that: The method of updating the weight matrix and bias vector of the adaptive feature compensation network by back propagation algorithm comprises the steps of: S401, initializing a weight matrix and a bias vector; the initialization of the weight matrix needs to adapt to the distribution characteristics of the leaf color features; S402, feature extraction and prediction generation: inputting the leaf color feature data into the adaptive feature compensation network; The abstract features of the leaf color feature data are extracted layer by layer through the weight function, and a nutritional status probability score is generated; the nutritional status probability distribution represents the confidence level of the nutritional status category; for example, the probability of nitrogen deficiency is 85%, and the probability of water deficiency is 15%.

7. The method for monitoring planting information based on machine vision according to claim 6, characterized in that: The method of updating the weight matrix and bias vector of the adaptive feature compensation network by back propagation algorithm further includes the steps of: S403, quantifying the error between the nutritional status probability distribution and the nutritional status label: using a cross entropy loss function to calculate the loss value between the nutritional status probability distribution and the nutritional status label; S404, back-propagation gradient calculation and weight matrix and bias vector update: Based on the chain rule, reversely calculate the partial derivative of the cross-entropy loss function with respect to the weight matrix and bias vector, and then update the weight matrix and bias vector through the gradient descent method, thereby improving the accuracy of mapping the nutritional status of the crop from the leaf color feature data.

8. The method for monitoring planting information based on machine vision according to claim 7, characterized in that: The updating of the weight matrix and bias vector of the adaptive feature compensation network by the back propagation algorithm further includes: The cross entropy loss function calculation formula is: ； The calculation formula for updating the weight matrix is: ； ； The calculation formula for the updated bias vector is: ； ； in, is the loss value of nutritional status probability distribution and nutritional status label, For the The probability distribution value of the nutritional status of the class, For the The value of the nutritional status of the class is manually annotated or experimentally determined, is the total number of nutritional status categories; is the error value of the current layer, For the The activation value of the layer, To update the step size, it is used to prevent oscillation or underfitting; For the The weight matrix of the layer, For the The bias vector of the layer.

9. The method for monitoring planting information based on machine vision according to claim 7, characterized in that: The analyzing of the compensated leaf color feature data by the adaptive feature compensation network comprises the following steps: Inputting the compensated leaf color feature data into the adaptive feature compensation network to generate a nutritional status probability distribution; The nutritional status probability distribution is mapped to a predefined nutritional status level to obtain a crop nutritional status assessment result.

10. A planting information monitoring system based on machine vision, applying the planting information monitoring method based on machine vision according to any one of claims 1 to 9, characterized in that: The system includes an image data acquisition module, a compensation factor calculation module, a feature data generation module, a nutrition status assessment module and a planting management suggestion module; The image data acquisition module is used to acquire image data of crop leaves and collect real-time light intensity data through an ambient light sensor array; The compensation factor calculation module is used to calculate the blade surface reflectivity compensation factor according to the real-time light intensity data; The feature data generating module is used to fuse the nonlinear compensation factor with the image data to generate compensated leaf color feature data; The nutritional status assessment module is used to analyze the compensated leaf color feature data through an adaptive feature compensation network to obtain a crop nutritional status assessment result; The planting management suggestion module is used to generate a planting management suggestion report based on the nutritional status assessment result.

Images