Method, device and terminal equipment for monitoring the distribution of poisonous weeds in grasslands

Abstract

This application provides a method, apparatus, and terminal device for monitoring the distribution of noxious weeds in grasslands, applicable to the field of image processing technology. The method comprises: obtaining first spectral image information and second spectral image information; determining initial location information of noxious weeds in the first spectral image based on a preset noxious weed sample information database and the first spectral image information; calculating target location information of noxious weeds in the first spectral image based on the initial location information of noxious weeds in the first spectral image and a preset noxious weed band classification model; determining location information of noxious weeds in the second spectral image based on the target location information of noxious weeds in the first spectral image and the second spectral image information; and generating noxious weed location distribution information based on the second spectral image noxious weed location information, a preset noxious weed spectral image feature extraction model, and a preset noxious weed location detection model. This application combines multiple spectral images to improve the efficiency and accuracy of noxious weed distribution monitoring.

Description

Technical Field

[0001] The present application belongs to the field of image processing technology, and in particular relates to a method, device and terminal equipment for monitoring the distribution of poisonous weeds in grasslands. Background Art

[0002] Grassland ecosystems are among the largest terrestrial ecosystems on Earth, playing a vital role in global ecological balance, soil and water conservation, and the development of animal husbandry. However, in recent years, grassland ecosystems have been severely degraded due to overgrazing, climate change, and human activities. This has led to a decrease in grassland area, increased desertification, and the spread of noxious weeds. These weeds are not only toxic to livestock, causing poisoning, weight loss, and even death, but also directly impact meat and milk production, threatening the economic benefits of herders. Furthermore, the spread of noxious weeds further disrupts grassland biodiversity and ecological balance, accelerating grassland degradation. Therefore, monitoring and control of noxious weeds have become critical tasks for grassland conservation and the sustainable development of animal husbandry.

[0003] Monitoring noxious weeds has always been a major challenge in grassland ecological protection and management. Traditional methods rely primarily on manual ground surveys, where workers dynamically move across grasslands to collect data on noxious weeds such as Leontopodiumnanum, Ajania tenuifolia, Aster hispidus, Saussurea ajaponica, Aconitum pendulum, and Ligularia virgaurea. This data is then used for subsequent follow-up surveys. Currently, remote sensing technology can be used to monitor the distribution of noxious weeds on grasslands.

[0004] However, surveying the distribution of noxious weeds across vast grasslands is time-consuming and labor-intensive, extremely inefficient, and fails to fully capture the actual distribution of noxious weeds, hindering accurate detection. Existing remote sensing monitoring technologies have limitations in distinguishing the spectra of noxious weeds from common vegetation, particularly in mixed vegetation. Spectral overlap makes accurate identification of noxious weeds difficult. Furthermore, they generally lack the ability to monitor the growth cycles and dynamic distribution of noxious weeds, making them difficult to monitor in grassland ecosystems with complex and fragmented noxious weed populations. This results in poor monitoring effectiveness and little room for improvement. Summary of the Invention

[0005] In view of this, the embodiments of the present application provide a method, apparatus, and terminal device for monitoring the distribution of poisonous weeds in grasslands, aiming to solve the problems existing in the prior art of long time consumption, small coverage, low efficiency in monitoring the distribution of poisonous weeds, and inability to quickly and comprehensively monitor the distribution of poisonous weeds. At the same time, it solves the problems of the prior art of difficulty in tracking the dynamic distribution of poisonous weeds and redundant spectral bands, thereby improving the accuracy of monitoring poisonous weeds in grasslands.

[0006] A first aspect of the embodiments of the present application provides a method for monitoring the distribution of poisonous weeds in grasslands, comprising:

[0007] Acquire a plurality of first spectral image information and second spectral image information;

[0008] Determining initial position information of poisonous weeds in the first spectral image according to a preset poisonous weed sample information database and the first spectral image information;

[0009] Calculating target position information of poisonous weeds in the first spectral image according to the initial position information of poisonous weeds in the first spectral image and a plurality of preset poisonous weed band classification models;

[0010] determining the poisonous weed position information of the second spectral image according to the target position information of the poisonous weeds in the first spectral image and the second spectral image information;

[0011] The poisonous weed position distribution information is generated according to the poisonous weed position information of the second spectral image, a preset poisonous weed spectral image feature extraction model, and a preset poisonous weed position detection model.

[0012] A second aspect of an embodiment of the present application provides a device for monitoring the distribution of poisonous weeds in grasslands, comprising:

[0013] A spectral image information acquisition module, configured to acquire a plurality of first spectral image information and a second spectral image information;

[0014] A first spectral image poisonous weed initial position information determination module, configured to determine the first spectral image poisonous weed initial position information based on a preset poisonous weed sample information library and the first spectral image information;

[0015] A first spectral image poisonous weed target position information calculation module, configured to calculate the first spectral image poisonous weed target position information based on the first spectral image poisonous weed initial position information and a plurality of preset poisonous weed band classification models;

[0016] A second spectral image poisonous weed position information determining module, configured to determine the second spectral image poisonous weed position information based on the first spectral image poisonous weed target position information and the second spectral image information;

[0017] The poisonous weed location distribution information generating module is configured to generate poisonous weed location distribution information based on the poisonous weed location information of the second spectral image, a preset poisonous weed spectral image feature extraction model, and a preset poisonous weed location detection model.

[0018] A third aspect of an embodiment of the present application provides a terminal device, which includes a memory and a processor. The memory stores a computer program that can be run on the processor. When the processor executes the computer program, it implements the steps of the grassland poisonous weed distribution monitoring method as described in the first aspect above.

[0019] A fourth aspect of the embodiments of the present application provides a computer-readable storage medium, comprising: storing a computer program, which, when executed by a processor, implements the steps of the grassland poisonous weed distribution monitoring method as described in the first aspect above.

[0020] Compared with the prior art, the embodiments of the present application have the following advantages: the positions of poisonous weeds in a high-resolution local spectral image are located by using a preset poisonous weed sample information library, and the spectral information in the local spectral image is then extracted to determine the position distribution of poisonous weeds in a lower-resolution global spectral image, thereby combining the obtained multiple spectral images and automatically identifying the distribution status of poisonous weeds in the global spectral image, thereby improving the efficiency, comprehensiveness and accuracy of monitoring the distribution status of poisonous weeds. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the embodiments or descriptions of the prior art. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0022] Figure 1 This is a schematic diagram of the implementation process of the method for monitoring the distribution of poisonous weeds in grasslands provided in Example 1 of the present application;

[0023] Figure 2 This is a schematic diagram of the implementation process of the method for monitoring the distribution of poisonous weeds in grasslands provided in Example 2 of the present application;

[0024] Figure 3 This is a schematic diagram of the implementation process of the method for monitoring the distribution of poisonous weeds in grasslands provided in Example 3 of the present application;

[0025] Figure 4 This is a schematic diagram of the implementation process of the method for monitoring the distribution of poisonous weeds in grasslands provided in Example 4 of the present application;

[0026] Figure 5 This is a schematic diagram of the implementation process of the method for monitoring the distribution of poisonous weeds in grasslands provided in Example 5 of the present application;

[0027] Figure 6 This is a schematic diagram of the implementation process of the method for monitoring the distribution of poisonous weeds in grasslands provided in Example 6 of the present application;

[0028] Figure 7 Schematic diagram of the structure of a device for monitoring the distribution of poisonous weeds in grasslands provided in an embodiment of the present application;

[0029] Figure 8 It is a schematic diagram of the terminal device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0030] In the following description, specific details such as specific system structures and techniques are provided for purposes of illustration rather than limitation to facilitate a thorough understanding of the embodiments of the present application. However, it will be apparent to those skilled in the art that the present application may be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to avoid obscuring the description of the present application with unnecessary detail.

[0031] In order to illustrate the technical solution described in this application, specific embodiments are provided below.

[0032] Figure 1 The following is a flowchart of the method for monitoring the distribution of poisonous weeds in grasslands provided in Example 1 of the present application, which is described in detail as follows:

[0033] Step S101: Acquire a plurality of first spectral image information and second spectral image information.

[0034] In this embodiment, the first and second spectral image information represent two different types of spectral image information, respectively. The first spectral image information can represent high-resolution spectral image information reflecting the local conditions of the grassland, i.e., it has more spectral details but a smaller field of view. The second spectral image information can represent lower-resolution spectral image information reflecting the overall conditions of the grassland, i.e., it has less spectral details but a larger field of view. Spectral image information includes key parameters such as reflectance and radiance in each band. The first spectral image information can be a drone hyperspectral image, obtained by dynamically capturing the grassland using a drone. As you can see, while drone hyperspectral imagery has high spatial resolution and rich spectral information, drones have limited flight range and their capture range is typically limited to a small area. Therefore, the first spectral image information obtained is a high-resolution localized image of the grassland. The second spectral image information can be a satellite remote sensing image, obtained using satellite remote sensing technology. As you can see, satellite remote sensing has wide coverage, but its spatial and spectral resolution are relatively low, making it difficult to capture the detailed characteristics of small-scale noxious weeds, especially those that are scattered and intermingled.

[0035] Step S102: determining initial position information of poisonous weeds in the first spectral image according to a preset poisonous weed sample information database and the first spectral image information.

[0036] In this embodiment, the pre-set noxious weed sample information database can be obtained by collecting high-quality noxious weed samples through field surveys, recording and summarizing their phenological information. Noxious weeds exhibit significant spectral differences in different bands (such as red edge, near-infrared, and shortwave infrared). These differences are primarily reflected in their reflectivity or absorption properties, creating a unique signature in the spectral curve compared to ordinary vegetation. By combining this noxious weed sample information database, a band selection algorithm can be used to extract key band features and distinguish noxious weeds from ordinary vegetation background. Specifically, taking an alpine meadow at an altitude of 3500-4500 meters as an example, 1.0 m × 1.0 m sample quads were laid out within the selected study area. Each quadrat was sequentially numbered, and photographs of the samples within the quadrat were taken, with the GPS coordinates of the center location recorded. The samples collected include common noxious weed types in the area, such as dwarf tinder, fine-leaved chrysanthemum, dogweed, saussurea, scutellaria, and yellow broom, along with their phenological information (e.g., flowering period, peak growth period, and degeneration period). The sampling process spans different seasons to ensure that the database contains the spectral characteristics of noxious weeds from different growth periods. The first spectral image information can be spliced ​​together, that is, multiple images reflecting local grassland conditions are spliced ​​together to obtain a high-resolution global image that reflects the grassland's overall condition. The noxious weeds in the spliced ​​global image are then located and annotated based on the spectral characteristics of each noxious weed in a preset noxious weed sample database. The resulting annotated information, i.e., the initial location information of the noxious weeds in the first spectral image, is used for subsequent screening to obtain more accurate noxious weed band information. Specifically, the first spectral image information can be preprocessed to remove noise, perform atmospheric correction, and perform geometric correction to ensure spectral image quality and data accuracy. The first spectral image information is stitched. Specifically, the pre-processed spectral image information is first subjected to lens calibration, black and white frame calibration, atmosphere and reflectance calibration in Spec View Version software. Then, in order to eliminate the influence of environmental noise and the instrument itself on the spectral data, the spectral data is smoothed using a smoothing method, and the bands affected by water vapor at the beginning and end are eliminated to obtain valid spectral band data. Finally, the corrected data is imported into the image processing software Hi Spectral Stitcher for spectral band data stitching to obtain a complete hyperspectral image of the monitoring area, i.e., the first spectral stitched image information.

[0037] In this embodiment, it can be understood that since poisonous weeds have significant spectral feature differences in different bands (such as red edge, near infrared and shortwave infrared), the poisonous weed sample information library can be combined to screen band features to distinguish poisonous weeds from ordinary vegetation, reduce spectral redundancy, and further improve the classification ability of the poisonous weed band classification model in the future.

[0038] Step S103 : calculating target position information of poisonous weeds in the first spectral image according to the initial position information of poisonous weeds in the first spectral image and a plurality of preset poisonous weed band classification models.

[0039] In this embodiment, the multiple preset noxious weed band classification models can include support vector machines (SVMs), random forests (RFs), BP neural networks, and XGBoost, or other neural network models. The preset noxious weed band classification models have been trained using the spectral features of noxious weeds in the noxious weed sample database. During the training process, appropriate feature scaling methods can be used to improve the model's convergence speed and accuracy. The initial noxious weed position information in the first spectral image can be subjected to band classification processing using the noxious weed band classification model, separating the initial noxious weed position information in the first spectral image into noxious weed position information and normal vegetation position information. The resulting noxious weed position information is the target noxious weed position information in the first spectral image. It will be appreciated that in this embodiment, the position information is represented by the relative position of the spectral information within the image.

[0040] Step S104: determining the poisonous weed position information in the second spectral image according to the poisonous weed target position information in the first spectral image and the second spectral image information.

[0041] In this embodiment, the high-resolution global image obtained by splicing the first spectral image information is aligned with the second spectral image, so that the position of the poisonous weeds in the second spectral image is located and marked by the target position information of the poisonous weeds in the first spectral image. The marked poisonous weed position in the second spectral image, i.e., the poisonous weed position information of the second spectral image, is used to represent the relative position of the poisonous weed spectrum in the second spectral image.

[0042] Step S105 : generating poisonous weed location distribution information based on the poisonous weed location information in the second spectral image, a preset poisonous weed spectral image feature extraction model, and a preset poisonous weed location detection model.

[0043] In this embodiment, the preset poisonous weed spectral image feature extraction model can be a Transformer model. The preset poisonous weed location detection model can be an LSTM model. After first converting the poisonous weed location information in the second spectral image into a format recognizable by the neural network model, the preset poisonous weed spectral image feature extraction model simultaneously extracts the spectral image features and the image temporal features, aligning the spectral image features with the image temporal features. The image temporal features are then predicted and calculated using the poisonous weed location detection model to obtain poisonous weed location distribution information, which can be used to visualize the distribution of poisonous weeds on the grassland.

[0044] The method for monitoring the distribution of poisonous weeds in grasslands provided in the embodiments of the present application locates the positions of poisonous weeds in a high-resolution local spectral image using a preset poisonous weed sample information library, and then extracts spectral information from the local spectral image to determine the position distribution of poisonous weeds in a lower-resolution global spectral image. This method combines the multiple acquired spectral images and automatically identifies the distribution of poisonous weeds in the global spectral image, thereby improving the efficiency, comprehensiveness, and accuracy of monitoring the distribution of poisonous weeds.

[0045] Figure 2 The flowchart of the method for monitoring the distribution of poisonous weeds in grasslands provided in the second embodiment of the present application is shown. The difference between the method and the first embodiment is that step S103 specifically includes:

[0046] Step S201 : dividing the first spectral image information into bands to obtain a plurality of initial poisonous weed detection band information.

[0047] In this embodiment, spectral curve data may be extracted from the first spectral image information, and the bands in the first spectral image information may be divided using a sub-interval band selection technique. The multiple spectral intervals obtained by the division are initial poisonous weed detection band information, which is used for subsequent analysis and determination of the distribution status of the poisonous weeds.

[0048] Step S202 : performing band extraction on the first spectral image information according to the initial position information of the poisonous weeds in the first spectral image to obtain band information representing the poisonous weeds.

[0049] In this embodiment, the first spectral image information may be subjected to band extraction using the spectral data of the poisonous weeds reflected in the initial position information of the poisonous weeds in the first spectral image. That is, band data in the first spectral image information that can be used to indicate the location of the poisonous weeds, i.e., poisonous weed characterization band information, may be determined.

[0050] Step S203 , determining whether the band standard deviation of the initial poisonous weed detection band information and the poisonous weed characterization band information is less than a preset band standard deviation threshold; if so, proceeding to step S204 ; if not, not using the initial poisonous weed detection band information as the intermediate poisonous weed detection band information.

[0051] In this embodiment, the band standard deviation of the initial noxious weed detection band information and the noxious weed characterization band information is calculated to quantify the degree of matching between the initial noxious weed detection band information and the noxious weed characterization band information, thereby determining whether the initial noxious weed detection band information can be used as band information representing the location of noxious weeds in the spectral image. The calculation formula for the band standard deviation of the initial noxious weed detection band information and the noxious weed characterization band information is as follows:

[0052] (1)

[0053] In the above formula (1), k Indicates the sample number, from 1 to n , n is the total number of samples, i Indicates the current band number, Indicates the k Samples in the band i The reflectivity on Indicates band i The sample mean of .

[0054] When the band standard deviation of the initial noxious weed detection band information and the noxious weed characterization band information is less than a preset band standard deviation threshold, it indicates that the initial noxious weed detection band information is highly likely to be the band reflection information of noxious weeds in the spectral image and can be used as the intermediate noxious weed detection band information for subsequent calculations. When the band standard deviation of the initial noxious weed detection band information and the noxious weed characterization band information is greater than or equal to the preset band standard deviation threshold, it indicates that the initial noxious weed detection band information is highly likely not the band reflection information of noxious weeds in the spectral image and cannot be used as the intermediate noxious weed detection band information for subsequent calculations.

[0055] In this embodiment, the band standard deviations may be arranged in descending order to screen out the main bands representing poisonous weeds.

[0056] Step S204: Using the initial poisonous weed detection band information as intermediate poisonous weed detection band information.

[0057] In this embodiment, when the band standard deviation of the initial poisonous weed detection band information and the poisonous weed characterization band information is less than the preset band standard deviation threshold, it means that the initial poisonous weed detection band information is highly likely to be the band reflection information of poisonous weeds in the spectral image and can be used as the intermediate poisonous weed detection band information for subsequent calculations.

[0058] Step S205 : Obtaining the target location information of poisonous weeds in the first spectral image according to the intermediate poisonous weed detection band information and a plurality of preset poisonous weed band classification models.

[0059] In this embodiment, the preset noxious weed band classification model can be a support vector machine (SVM), random forest (RF), BP neural network, XGBoost, or other neural network models. The intermediate noxious weed detection band information can be used as input data for the noxious weed band classification model, while the noxious weed band information and normal vegetation band information can be used as output data for the noxious weed band classification model. The noxious weed band information in the output data can be used as the noxious weed target location information in the first spectral image to represent the noxious weed location information in the first spectral image information.

[0060] The method for monitoring the distribution of poisonous weeds in grasslands provided in an embodiment of the present application divides the first spectral image information into bands, and performs a primary screening on the divided band information based on preliminarily determined poisonous weed location information. The band information after the primary screening is matched with the poisonous weed characterization band information to perform a secondary screening on the band information. The band information is then screened a third time using a preset poisonous weed band classification model, thereby ensuring the accuracy of screening the band information used to characterize the location of poisonous weeds in the first spectral image, thereby ensuring the accuracy and effectiveness of automatic monitoring of the distribution of poisonous weeds on grasslands.

[0061] Figure 3 The flowchart of the method for monitoring the distribution of poisonous weeds in grasslands provided in the third embodiment of the present application is shown. The difference between the method and the second embodiment is that the step S205 specifically includes:

[0062] Step S301 : calculating the band correlation and band distance of a plurality of the intermediate poisonous weed detection band information.

[0063] In this embodiment, the band correlation may be calculated to perform correlation analysis on the preliminarily screened intermediate poisonous weed detection band information. The band correlation calculation method may be:

[0064] (2)

[0065] In the above formula (2), x jk Indicates the k Samples in the band j The reflectivity on Indicates band j The sample mean of . In each subinterval, the correlation coefficient is used Measure the correlation between bands i and j;

[0066] You can combine six bands into a group and calculate the Jeffreys-Matusita (JM) distance of the poisonous weed spectrum corresponding to each group of bands. Select JM distance JM(I,j) For band combinations greater than or equal to 1.8, further verification of the selected bands is performed, and the calculation formula is as follows:

[0067] (3)

[0068] In the above formula (3), m i and m j Represents bands i and j The mean of S i and S j is the covariance matrix of the band.

[0069] Step S302 : Filtering the intermediate poisonous weed detection band information according to the band correlation and the band distance to obtain a plurality of poisonous weed detection band combination information.

[0070] In this embodiment, the correlation coefficient can be no higher than 0.3 as a screening criterion to determine the representative bands, and the JM distance is calculated based on the representative bands to select the JM distance. JM (I,j) The band combination greater than or equal to 1.8 is used as the band combination information for poisonous weed detection to ensure that the selected band has the best resolution.

[0071] Step S303 : calculating classification accuracy evaluation indices of multiple poisonous weed band classification models based on the poisonous weed detection band combination information and multiple preset poisonous weed band classification models.

[0072] In this embodiment, the poisonous weed detection band combination information can be used as input data for the poisonous weed band classification model to calculate the classification accuracy evaluation index of the poisonous weed band classification model. The classification accuracy evaluation index includes the overall classification accuracy (OA), the average classification accuracy (AA), and the Kappa coefficient, which are used to verify the effectiveness and reliability of the model in wide-area poisonous weed monitoring. The calculation formulas for the overall classification accuracy (OA), the average classification accuracy (AA), and the Kappa coefficient are as follows:

[0073] (4)

[0074] In the above formula (4), C Indicates the number of correctly classified samples in each category, T i represents the number of correctly classified samples in each category, and N represents the total number of samples.

[0075] (5)

[0076] In the above formula (5), p i Representation category i classification accuracy.

[0077] (6)

[0078] In the above formula (6), p o Indicates the total classification accuracy of the category OA , p i represents the expected classification accuracy, a i Representation category i The number of real samples, b i Indicates that the model predicts the category i The number of .

[0079] Step S304 : screening a plurality of preset poisonous weed band classification models according to the classification accuracy evaluation index to obtain a target poisonous weed band classification model.

[0080] In this embodiment, the best poisonous weed band classification model among the classification accuracy evaluation indicators can be selected as the target poisonous weed band classification model. That is, the poisonous weed band classification model with the highest overall classification accuracy (OA), average classification accuracy (AA), and Kappa coefficient is selected as the target poisonous weed band classification model, which can be used to obtain a small-scale weed distribution map generated based on the UAV hyperspectral image through inversion in the future.

[0081] Step S305 : obtaining target location information of poisonous weeds in the first spectral image according to the poisonous weed detection band combination information and the target poisonous weed band classification model.

[0082] In this embodiment, the poisonous weed detection band combination information can be used as input information of the target poisonous weed band classification model. The target poisonous weed band classification model will automatically distinguish between the poisonous weed band information and the normal vegetation band information, thereby determining whether the target location is a poisonous weed. Based on the target poisonous weed classification model, the first spectral image information is inverted to output the position information of all poisonous weeds in the first spectral image, that is, the first spectral image poisonous weed target position information.

[0083] The method for monitoring the distribution of poisonous weeds in grasslands provided in the embodiments of the present application ensures the accuracy and reliability of determining the characteristic bands of poisonous weeds by calculating the band correlation and band distance of the bands corresponding to the poisonous weeds, and reduces the amount of input data of the poisonous weed position detection model in subsequent calculations to reduce model redundancy. The performance of multiple poisonous weed band classification models is evaluated by calculating a classification accuracy evaluation index, thereby screening out the poisonous weed band classification model with the best classification calculation effect, which is used to accurately locate the location of the poisonous weeds in the first spectral image, thereby improving the accuracy and effectiveness of poisonous weed identification.

[0084] Figure 4 The flowchart of the method for monitoring the distribution of poisonous weeds in grasslands provided in the fourth embodiment of the present application is shown. The difference between the fourth embodiment and the first embodiment is that step S104 specifically includes:

[0085] Step S401 : performing registration processing on the target position information of poisonous weeds in the first spectral image and the second spectral image information, and determining the initial position information of poisonous weeds in the second spectral image according to the target position information of poisonous weeds in the first spectral image.

[0086] In this embodiment, the second spectral image information can first be declouded using the QA60 band, retaining images with less than 30% cloud cover. Atmospheric and geometric corrections are then performed on the image to ensure image quality. An image synthesis strategy is then applied to combine multiple satellite images of the same phenological period into a cloud-free composite image, thereby increasing the coverage of image data during the critical growth period of noxious weeds. Specifically, during the registration process, the local, multi-phenological-period noxious weed invasion distribution map of the first spectral mosaic image information can be resampled using the nearest neighbor interpolation method to achieve the same spatial resolution as the second spectral image information. The second spectral image information can be a Sentinel-2 satellite image. Using the resampled local, multi-phenological-period noxious weed invasion distribution map of the first spectral mosaic image information as a reference, the second spectral image information is geometrically registered using a first-order polynomial method. Ground control points (GCPs) are used for correction during registration to ensure that the correction error is within a single pixel, thereby achieving precise spatial alignment.

[0087] In this embodiment, it is understandable that the second spectral image information may be directly registered with the classification result obtained by classifying the first spectral image information, rather than all the information in the first spectral stitching image information, so as to reduce the amount of calculation and computational complexity while determining the location information of the poisonous weeds in the second spectral image information.

[0088] Step S402: Masking the initial position information of poisonous weeds in the second spectral image according to preset poisonous weed mask pixel information to determine the position information of poisonous weeds in the second spectral image.

[0089] In this embodiment, the preset poisonous weed mask pixel information can be compiled based on poisonous weed sample information in the poisonous weed sample information database and normal vegetation information collected during routine grassland monitoring, or it can be generated based on the classification results obtained by the first spectral image information classification model. Masking can be performed separately on the poisonous weed pixels and non-poisonous weed pixels in the poisonous weed location information of the second spectral image. The number of poisonous weed and non-poisonous weed pixels falling within each second spectral image pixel window is then counted, and the proportion of these pixels to the total number of second spectral image pixel windows is calculated. This distribution characteristic of poisonous weeds and non-poisonous weeds is obtained at the second spectral image pixel scale, which can be represented by the poisonous weed location information of the second spectral image. This distribution characteristic serves as sample label data for subsequent model classification and can be used to train the poisonous weed monitoring model.

[0090] The method for monitoring the distribution of poisonous weeds in grasslands provided in an embodiment of the present application performs registration processing on a high-resolution spectral image and a low-resolution global image, thereby determining the location data of poisonous weeds in a low-resolution global image based on the location information of the poisonous weeds in the high-resolution spectral image. The poisonous weed location data can be used to train a poisonous weed location detection model. The trained poisonous weed location detection model can be used to automatically identify the location distribution information of poisonous weeds in the global image, thereby enabling automatic, rapid, and accurate monitoring of the distribution of poisonous weeds through the global image, helping staff to promptly carry out poisonous weed extermination work.

[0091] Figure 5 The flowchart of the implementation of the method for monitoring the distribution of poisonous weeds in grasslands provided in the fifth embodiment of the present application is shown. The difference between the method and the first embodiment is that the step S105 specifically includes:

[0092] Step S501 : performing format conversion on the poisonous weed position information of the second spectral image to obtain poisonous weed spectral image tensor information.

[0093] In this embodiment, the second spectral image poisonous weed position information may be formatted into a multi-dimensional array in the form of [ N , T , C , H , W ], where N represents the number of samples, T is the time step, Cis the number of spectral bands, H and W are the height and width of the image block, respectively. Then, the location information of poisonous weeds in the two-spectral image is patch segmented. The location information of poisonous weeds in the two-spectral image is divided into 3D patches of equal size. Each patch is subjected to preliminary feature transformation through linear projection to form the spectral image tensor information of poisonous weeds.

[0094] Step S502 : Based on a preset poisonous weed spectral image feature extraction model, spatiotemporal feature extraction is performed on the poisonous weed spectral image tensor information to obtain a plurality of poisonous weed distribution time feature information and a plurality of poisonous weed distribution space feature information.

[0095] In this embodiment, the preset poisonous weed spectral image feature extraction model can be a Transformer model. The temporal and spatial features of the poisonous weed spectral image tensor information are extracted by the Transformer model, and the temporal features and spatial features can be extracted respectively by a convolutional layer and a temporal encoder. Specifically, the feature extraction of each patch can be performed by a convolutional layer. The convolutional layer can adopt a 3×3 kernel size and a step size of 1 to capture the local features of the image. The data of each time step is then input into the temporal encoder. The temporal encoder uses a multi-head attention mechanism to capture the relationship between different time steps. The extracted time step information is used as the temporal feature information of the poisonous weed distribution, and the extracted spatial information is used as the spatial feature information of the poisonous weed distribution.

[0096] Step S503: performing position coding processing on the time characteristic information of the poisonous weed distribution to obtain time series information of the poisonous weed distribution.

[0097] In this embodiment, each time step data is position-encoded and then enters the encoder to preserve the order of the time series. The time series information output by the time encoder can be expressed as time series information of the poisonous weed distribution, which enables the poisonous weed position detection model to understand the changing trend of the poisonous weeds in the time dimension and the association between adjacent time steps.

[0098] Step S504 : alternately splicing the poisonous weed distribution time series information and the poisonous weed distribution spatial feature information to obtain a poisonous weed distribution fusion feature tensor.

[0099] In this embodiment, the extracted spatial features are denoted as Vs and the temporal features are denoted as Vt. First, Vs and Vt are divided into several feature vectors. For each feature vector, elements can be alternately selected from Vs and Vt for fusion. Specifically, the i-th feature is extracted from Vs and placed at the 2i+1 position of the fusion vector; the i-th feature is extracted from Vt and placed at the 2i position, and the fused global spatiotemporal feature tensor is output. X g, which can be represented as a fused feature tensor for the distribution of poisonous weeds. After the alternating fusion process, the spatial and temporal features are evenly distributed in the new feature vector, ensuring that the feature quantities alternate evenly across the spatial and temporal dimensions. This fusion strategy enables the poisonous weed location detection model to retain and utilize information from both space and time in subsequent calculations, making the output feature vector more representative and balanced. Therefore, this alternating fusion strategy ensures that spatial and temporal features are evenly alternating in the fused feature vector, helping the poisonous weed location detection model better extract the distribution characteristics of poisonous weeds and achieve more stable and accurate classification results.

[0100] Step S505 : Based on a preset poisonous weed location detection model, the poisonous weed distribution fusion feature tensor is analyzed and calculated to generate poisonous weed location distribution information.

[0101] In this embodiment, the pre-set noxious weed location detection model can be an LSTM model. This model can extract time information from the noxious weed distribution fusion feature tensor and use this time information as input data for the LSTM model. The LSTM uses its memory cells and gating mechanism to capture long-term and short-term dependencies between time steps, generating noxious weed location distribution information as output data. Specifically, a multi-layer LSTM structure can be used to extract deep time series features.

[0102] The method for monitoring the distribution of poisonous weeds in grasslands provided by the embodiments of the present application extracts the spectral features of poisonous weeds in the poisonous weed position information of the second spectral image through a poisonous weed spectral image feature extraction model, and alternately fuses the extracted temporal features and spatial features to make the output feature vector more representative and balanced, ensuring that the temporal features and spatial features are evenly alternated in the fused feature vector, which helps the poisonous weed position detection model to better extract the distribution characteristics of poisonous weeds within the global field of view, achieving a more stable and accurate classification effect, thereby improving the recognition accuracy and comprehensiveness of poisonous weeds over a wide area, and helping staff to timely and accurately understand the distribution status of poisonous weeds so as to take control measures to control the spread of poisonous weeds.

[0103] Figure 6 The flowchart of the method for monitoring the distribution of poisonous weeds in grasslands provided in the sixth embodiment of the present application is shown. The difference between the method and the fifth embodiment is that the step S505 specifically includes:

[0104] Step S601: performing inverse residual calculation on the poisonous weed distribution fusion feature tensor to obtain a poisonous weed distribution optimization feature tensor.

[0105] In this embodiment, the inverse residual calculation can be performed on the poisonous weed distribution fusion feature tensor through the inverse residual module in the LSTM model, and the calculation result is used as the poisonous weed distribution optimization feature tensor to improve the parameter efficiency of the model. The fusion feature is optimized through the inverse residual module, and the consistency of the feature dimension is ensured. In addition, the feature expression ability in the poisonous weed location detection model is enhanced in subsequent calculations.

[0106] Step S602: performing time series extraction on the optimized feature tensor of the poisonous weed distribution to obtain a time series feature tensor of the poisonous weed distribution.

[0107] In this embodiment, the time series classification module in the LSTM model can be used to further enhance the extraction of time series information. The extracted feature information can be represented as a time series feature tensor of the poisonous weed distribution, which is used as input data for subsequent analysis and calculation of the poisonous weed location detection model.

[0108] Step S603 : Based on a preset poisonous weed location detection model, feature mapping processing and nonlinear transformation are performed on the poisonous weed distribution time series feature tensor to obtain poisonous weed location distribution time series information corresponding to each poisonous weed distribution spatial feature information.

[0109] In this embodiment, the preset noxious weed location detection model can be an LSTM model. This model can process the noxious weed distribution time series feature tensor through a fully connected layer to perform feature mapping. The mapped feature tensor then undergoes a nonlinear transformation using an activation function to obtain the time series information corresponding to the spatial feature information of each noxious weed distribution. This information can be used to identify the time points at which noxious weed characteristic bands are likely to appear relative to each band in the spectral image. Conversely, it can also be used to characterize the spatial distribution of noxious weeds along the temporal dimension in the spectral image to assess the extent of noxious weed spread. The activation function can be a Softmax function.

[0110] Step S604: generating poisonous weed location distribution information based on the poisonous weed distribution spatial characteristic information and the poisonous weed location distribution time sequence information.

[0111] In this embodiment, the time sequence information of the poisonous weed location distribution and the spatial characteristic information of the poisonous weed distribution may be spliced ​​together to obtain the spatiotemporal distribution for characterizing the poisonous weeds, thereby generating the poisonous weed location distribution information.

[0112] The method for monitoring the distribution of poisonous weeds in grasslands provided in the embodiments of the present application performs inverse residual calculation on the fused features of the poisonous weed distribution, thereby reducing the feature values ​​while retaining important features to reduce computational complexity. The method analyzes and processes the temporal features in the fused features of the poisonous weed distribution through a poisonous weed position detection model, thereby calculating the temporal sequence information of the poisonous weed position distribution corresponding to the spatial feature information of each poisonous weed distribution, and generating spatiotemporal distribution status information of the poisonous weeds, thereby achieving rapid, accurate, and comprehensive monitoring of the spread of poisonous weeds in a wide area of ​​grassland, and comprehensively improving the effectiveness, efficiency, and precision of poisonous weed monitoring.

[0113] Corresponding to the method of the above embodiment, Figure 7 The structural block diagram of the grassland poisonous weed distribution status monitoring device provided by the embodiment of the present application is shown. For the convenience of explanation, only the parts related to the embodiment of the present application are shown. Figure 7 The exemplary grassland poisonous weed distribution status monitoring device may be an executing body of the grassland poisonous weed distribution status monitoring method provided in the aforementioned first embodiment.

[0114] Reference Figure 7 The grassland poisonous weed distribution monitoring device includes:

[0115] The spectral image information acquisition module 710 is used to acquire a plurality of first spectral image information and a second spectral image information;

[0116] A first spectral image poisonous weed initial position information determining module 720 is configured to determine the first spectral image poisonous weed initial position information based on a preset poisonous weed sample information library and the first spectral image information;

[0117] The first spectral image poisonous weed target position information calculation module 730 is used to calculate the first spectral image poisonous weed target position information based on the first spectral image poisonous weed initial position information and a plurality of preset poisonous weed band classification models;

[0118] A second spectral image poisonous weed position information determining module 740 is configured to determine the second spectral image poisonous weed position information based on the first spectral image poisonous weed target position information and the second spectral image information;

[0119] The poisonous weed location distribution information generating module 750 is configured to generate poisonous weed location distribution information based on the poisonous weed location information in the second spectral image, a preset poisonous weed spectral image feature extraction model, and a preset poisonous weed location detection model.

[0120] The process of each module in the grassland poisonous weed distribution monitoring device provided in the embodiment of the present application realizing its own function can be specifically referred to the aforementioned Figure 1 The description of the first embodiment is omitted here.

[0121] The second spectrum image poisonous weed location information determination module 740 includes:

[0122] a first spectral stitching image information generating unit, configured to stitch together a plurality of the first spectral image information to generate first spectral stitching image information;

[0123] a second spectral image poisonous weed initial position information determining unit, configured to perform registration processing on the first spectral stitching image information and the second spectral image information, and determine the second spectral image poisonous weed initial position information based on the first spectral image poisonous weed target position information;

[0124] The second spectral image poisonous weed position information determining unit is configured to perform mask processing on the initial position information of the poisonous weeds in the second spectral image according to preset poisonous weed mask pixel information to determine the position information of the poisonous weeds in the second spectral image.

[0125] The process of each unit in the second spectral image poisonous weed position information determination module 740 provided in the embodiment of the present application realizing its own function can be specifically referred to the aforementioned Figure 4 The description of the fourth embodiment is omitted here.

[0126] It should be understood that the size of the serial numbers of the steps in the above embodiments does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0127] It should be understood that when used in the present specification and the appended claims, the term "comprising" indicates the presence of described features, integers, steps, operations, elements and / or components, but does not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or collections thereof.

[0128] It will also be understood that the term "and / or" used in this specification and the appended claims refers to and includes any and all possible combinations of one or more of the associated listed items.

[0129] As used in this specification and the appended claims, the term "if" can be interpreted as "when" or "upon" or "in response to determining" or "in response to detecting," depending on the context. Similarly, the phrase "if it is determined" or "if [described condition or event] is detected" can be interpreted as meaning "upon determination" or "in response to determining" or "upon detection of [described condition or event]" or "in response to detecting [described condition or event]," depending on the context.

[0130] In addition, in the description of the present specification and the appended claims, the terms "first", "second", "third", etc. are only used to distinguish descriptions and should not be understood as indicating or implying relative importance. It should also be understood that although the terms "first", "second", etc. are used in the text to describe various elements in some embodiments of the present application, these elements should not be limited by these terms. These terms are only used to distinguish one element from another. For example, a first table can be named a second table, and similarly, a second table can be named a first table without departing from the scope of the various described embodiments. Both the first table and the second table are tables, but they are not the same table.

[0131] References to "one embodiment" or "some embodiments" in this specification mean that a particular feature, structure, or characteristic described in conjunction with that embodiment is included in one or more embodiments of the present application. Thus, phrases such as "in one embodiment," "in some embodiments," "in other embodiments," and "in other embodiments" appearing in various places in this specification do not necessarily refer to the same embodiment, but rather mean "one or more but not all embodiments," unless otherwise specifically emphasized. The terms "including," "comprising," "having," and variations thereof all mean "including but not limited to," unless otherwise specifically emphasized.

[0132] The grassland poisonous weed distribution monitoring method provided in the embodiments of the present application can be applied to terminal devices such as mobile phones, tablet computers, wearable devices, vehicle-mounted devices, augmented reality (AR) / virtual reality (VR) devices, laptop computers, ultra-mobile personal computers (UMPCs), netbooks, and personal digital assistants (PDAs). The embodiments of the present application do not impose any restrictions on the specific type of terminal device.

[0133] For example, the terminal device can be a station (STAION, ST) in a WLAN, a cellular phone, a cordless phone, a Session Initiation Protocol (SIP) phone, a Wireless Local Loop (WLL) station, a Personal Digital Assistant (PDA) device, a handheld device with wireless communication function, a computing device or other processing device connected to a wireless modem, a vehicle-mounted device, a vehicle networking terminal, a computer, a laptop computer, a handheld communication device, a handheld computing device, a satellite wireless device, a wireless modem card, a TV set-top box (STB), customer premise equipment (CPE) and / or other devices for communicating on a wireless system and a next-generation communication system, such as a mobile terminal in a 5G network or a mobile terminal in a future evolved Public Land Mobile Network (PLMN) network.

[0134] As an example and not a limitation, when the terminal device is a wearable device, the wearable device can also be a general term for wearable devices that are intelligently designed and developed using wearable technology for daily wear, such as glasses, gloves, watches, clothing and shoes. A wearable device is a portable device that is worn directly on the body or integrated into the user's clothes or accessories. Wearable devices are not only hardware devices, but also achieve powerful functions through software support, data interaction, and cloud interaction. Broadly speaking, wearable smart devices include those that are full-featured, large in size, and can achieve complete or partial functions without relying on smartphones, such as smart watches or smart glasses, as well as those that only focus on a certain type of application function and need to be used in conjunction with other devices such as smartphones, such as various smart bracelets and smart jewelry for vital sign monitoring.

[0135] Figure 8 This is a schematic diagram of the structure of a terminal device provided by an embodiment of the present application. Figure 8 As shown, the terminal device 8 of this embodiment includes: at least one processor 80 ( Figure 8 Only one is shown), a memory 81, wherein the memory 81 stores a computer program 82 that can be run on the processor 80. When the processor 80 executes the computer program 82, the steps in the above-mentioned grassland poisonous weed distribution monitoring method embodiments are implemented, such as Figure 1 Alternatively, when the processor 80 executes the computer program 82, the functions of the modules / units in the above-mentioned device embodiments are realized, for example, Figure 7Functions of modules 710 to 750 are shown.

[0136] The terminal device 8 can be a computing device such as a desktop computer, a notebook, a PDA, or a cloud server. The terminal device can include, but is not limited to, a processor 80 and a memory 81. Those skilled in the art will understand that Figure 8 It is only an example of the terminal device 8 and does not constitute a limitation of the terminal device 8. It may include more or fewer components than shown in the figure, or a combination of certain components, or different components. For example, the terminal device may also include an input and sending device, a network access device, a bus, etc.

[0137] The processor 80 may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor.

[0138] In some embodiments, the memory 81 may be an internal storage unit of the terminal device 8, such as a hard drive or memory of the terminal device 8. The memory 81 may also be an external storage device of the terminal device 8, such as a plug-in hard drive, a Smart Media Card (SMC), a Secure Digital (SD) card, a flash memory card, etc. equipped on the terminal device 8. Furthermore, the memory 81 may include both an internal storage unit of the terminal device 8 and an external storage device. The memory 81 is used to store an operating system, application programs, a boot loader, data, and other programs, such as the program code of the computer program. The memory 81 may also be used to temporarily store data that has been sent or is about to be sent.

[0139] In addition, the functional units in the various embodiments of the present application may be integrated into a single processing unit, or each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0140] An embodiment of the present application also provides a terminal device, which includes at least one memory, at least one processor, and a computer program stored in the at least one memory and executable on the at least one processor. When the processor executes the computer program, the terminal device implements the steps of any of the above-mentioned method embodiments.

[0141] An embodiment of the present application further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps in the above-mentioned various method embodiments can be implemented.

[0142] An embodiment of the present application provides a computer program product. When the computer program product is run on a terminal device, the terminal device can implement the steps in the above-mentioned method embodiments when executing the computer program product.

[0143] If the integrated module / unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present application can implement all or part of the processes in the above-mentioned embodiment method by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by a processor, it can implement the steps of each of the above-mentioned method embodiments. The computer program includes computer program code, which can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium can include: any entity or device that can carry the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal and software distribution medium.

[0144] In the above embodiments, the description of each embodiment has its own focus. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant description of other embodiments.

[0145] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0146] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0147] The above-described embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present application, and should all be included in the scope of protection of the present application.

Claims

1. A method for monitoring the distribution of poisonous weeds in grasslands, characterized in that: include: Acquire a plurality of first spectral image information and second spectral image information; Determining initial position information of poisonous weeds in the first spectral image according to a preset poisonous weed sample information database and the first spectral image information; Calculating target position information of poisonous weeds in the first spectral image according to the initial position information of poisonous weeds in the first spectral image and a plurality of preset poisonous weed band classification models; determining the poisonous weed position information of the second spectral image according to the target position information of the poisonous weeds in the first spectral image and the second spectral image information; generating poisonous weed location distribution information based on the poisonous weed location information of the second spectral image, a preset poisonous weed spectral image feature extraction model, and a preset poisonous weed location detection model; The step of generating the poisonous weed location distribution information based on the poisonous weed location information of the second spectral image, a preset poisonous weed spectral image feature extraction model, and a preset poisonous weed location detection model specifically includes: performing format conversion on the second spectral image of poisonous weed position information to obtain spectral image tensor information of the poisonous weed; Based on a preset poisonous weed spectral image feature extraction model, performing spatiotemporal feature extraction on the poisonous weed spectral image tensor information to obtain a plurality of poisonous weed distribution time feature information and a plurality of poisonous weed distribution space feature information; performing position coding processing on the time characteristic information of the poisonous weed distribution to obtain time series information of the poisonous weed distribution; Alternately splicing the poisonous weed distribution time series information and the poisonous weed distribution spatial feature information to obtain a poisonous weed distribution fusion feature tensor; Based on a preset poisonous weed location detection model, the poisonous weed distribution fusion feature tensor is analyzed and calculated to generate poisonous weed location distribution information; The step of analyzing and calculating the poisonous weed distribution fusion feature tensor based on a preset poisonous weed location detection model to generate poisonous weed location distribution information specifically includes: Performing inverse residual calculation on the poisonous weed distribution fusion feature tensor to obtain a poisonous weed distribution optimized feature tensor; Performing time series extraction on the optimized feature tensor of the poisonous weed distribution to obtain a time series feature tensor of the poisonous weed distribution; Based on a preset poisonous weed location detection model, feature mapping processing and nonlinear transformation are performed on the poisonous weed distribution time series feature tensor to obtain poisonous weed location distribution time series information corresponding to each poisonous weed distribution spatial feature information; generating poisonous weed location distribution information based on the poisonous weed distribution spatial characteristic information and the poisonous weed location distribution time sequence information; The step of determining the poisonous weed position information of the second spectral image based on the poisonous weed target position information of the first spectral image and the second spectral image information specifically includes: performing registration processing on the target position information of poisonous weeds in the first spectral image and the second spectral image information, and determining the initial position information of poisonous weeds in the second spectral image according to the target position information of poisonous weeds in the first spectral image; performing mask processing on the initial position information of the poisonous weeds in the second spectral image according to the preset poisonous weed mask pixel information to determine the position information of the poisonous weeds in the second spectral image; The masking process is to perform masking on the poisonous weed pixels and non-poisonous weed pixels in the poisonous weed position information of the second spectral image, respectively, and then count the number of poisonous weed and non-poisonous weed pixels falling within each second spectral image pixel window, and calculate the proportion of these pixels to the total number of second spectral image pixel windows, thereby obtaining the distribution characteristics of poisonous weeds and non-poisonous weeds at the second spectral image pixel scale, and representing them using the poisonous weed position information of the second spectral image; The step of obtaining target location information of poisonous weeds in the first spectral image based on the initial location information of poisonous weeds in the first spectral image and a plurality of preset poisonous weed band classification models specifically includes: Dividing the first spectral image information into bands to obtain a plurality of initial poisonous weed detection band information; performing band extraction on the first spectral image information according to the initial position information of the poisonous weeds in the first spectral image to obtain band information representing the poisonous weeds; When the band standard deviation of the initial poisonous weed detection band information and the poisonous weed characterization band information is less than a preset band standard deviation threshold, the initial poisonous weed detection band information is used as the intermediate poisonous weed detection band information; Obtaining target location information of poisonous weeds in the first spectral image according to the intermediate poisonous weed detection band information and a plurality of preset poisonous weed band classification models; The step of obtaining the target location information of poisonous weeds in the first spectral image based on the intermediate poisonous weed detection band information and a plurality of preset poisonous weed band classification models specifically includes: Calculating the band correlation and band distance of the plurality of intermediate poisonous weed detection band information; filtering the intermediate poisonous weed detection band information according to the band correlation and the band distance to obtain a plurality of poisonous weed detection band combination information; Calculating classification accuracy evaluation indicators of multiple poisonous weed band classification models based on the poisonous weed detection band combination information and multiple preset poisonous weed band classification models; screening a plurality of preset poisonous weed band classification models according to the classification accuracy evaluation index to obtain a target poisonous weed band classification model; Obtaining target location information of poisonous weeds in the first spectral image according to the poisonous weed detection band combination information and the target poisonous weed band classification model; The first spectral image information and the second spectral image information respectively represent two different spectral image information, wherein the first spectral image information represents spectral image information with high resolution reflecting the local situation of the grassland, and the second spectral image information represents spectral image information with low resolution reflecting the overall situation of the grassland.

2. A device for monitoring the distribution of poisonous weeds in grasslands, characterized in that: include: A spectral image information acquisition module, configured to acquire a plurality of first spectral image information and a second spectral image information; A first spectral image poisonous weed initial position information determination module, configured to determine the first spectral image poisonous weed initial position information based on a preset poisonous weed sample information library and the first spectral image information; A first spectral image poisonous weed target position information calculation module, configured to calculate the first spectral image poisonous weed target position information based on the first spectral image poisonous weed initial position information and a plurality of preset poisonous weed band classification models; A second spectral image poisonous weed position information determining module, configured to determine the second spectral image poisonous weed position information based on the first spectral image poisonous weed target position information and the second spectral image information; a poisonous weed location distribution information generating module, configured to generate poisonous weed location distribution information based on the poisonous weed location information of the second spectral image, a preset poisonous weed spectral image feature extraction model, and a preset poisonous weed location detection model; The second spectral image poisonous weed location information determination module includes: a first spectral stitching image information generating unit, configured to stitch together a plurality of the first spectral image information to generate first spectral stitching image information; a second spectral image poisonous weed initial position information determining unit, configured to perform registration processing on the first spectral stitching image information and the second spectral image information, and determine the second spectral image poisonous weed initial position information based on the first spectral image poisonous weed target position information; a second spectral image poisonous weed position information determining unit, configured to perform mask processing on the initial poisonous weed position information of the second spectral image according to preset poisonous weed mask pixel information, filter out non-poisonous weed areas, and determine the poisonous weed position information of the second spectral image; The step of analyzing and calculating the poisonous weed distribution fusion feature tensor based on a preset poisonous weed location detection model to generate poisonous weed location distribution information specifically includes: Performing inverse residual calculation on the poisonous weed distribution fusion feature tensor to obtain a poisonous weed distribution optimized feature tensor; Performing time series extraction on the optimized feature tensor of the poisonous weed distribution to obtain a time series feature tensor of the poisonous weed distribution; Based on a preset poisonous weed location detection model, feature mapping processing and nonlinear transformation are performed on the poisonous weed distribution time series feature tensor to obtain poisonous weed location distribution time series information corresponding to each poisonous weed distribution spatial feature information; generating poisonous weed location distribution information based on the poisonous weed distribution spatial characteristic information and the poisonous weed location distribution time sequence information; The step of determining the poisonous weed position information of the second spectral image based on the poisonous weed target position information of the first spectral image and the second spectral image information specifically includes: performing registration processing on the target position information of poisonous weeds in the first spectral image and the second spectral image information, and determining the initial position information of poisonous weeds in the second spectral image according to the target position information of poisonous weeds in the first spectral image; performing mask processing on the initial position information of the poisonous weeds in the second spectral image according to the preset poisonous weed mask pixel information to determine the position information of the poisonous weeds in the second spectral image; The masking process is to perform masking on the poisonous weed pixels and non-poisonous weed pixels in the poisonous weed position information of the second spectral image, respectively, and then count the number of poisonous weed and non-poisonous weed pixels falling within each second spectral image pixel window, and calculate the proportion of these pixels to the total number of second spectral image pixel windows, thereby obtaining the distribution characteristics of poisonous weeds and non-poisonous weeds at the second spectral image pixel scale, and representing them using the poisonous weed position information of the second spectral image; The step of obtaining target location information of poisonous weeds in the first spectral image based on the initial location information of poisonous weeds in the first spectral image and a plurality of preset poisonous weed band classification models specifically includes: Dividing the first spectral image information into bands to obtain a plurality of initial poisonous weed detection band information; performing band extraction on the first spectral image information according to the initial position information of the poisonous weeds in the first spectral image to obtain band information representing the poisonous weeds; When the band standard deviation of the initial poisonous weed detection band information and the poisonous weed characterization band information is less than a preset band standard deviation threshold, the initial poisonous weed detection band information is used as the intermediate poisonous weed detection band information; Obtaining target location information of poisonous weeds in the first spectral image according to the intermediate poisonous weed detection band information and a plurality of preset poisonous weed band classification models; The step of obtaining the target location information of poisonous weeds in the first spectral image based on the intermediate poisonous weed detection band information and a plurality of preset poisonous weed band classification models specifically includes: Calculating the band correlation and band distance of the plurality of intermediate poisonous weed detection band information; filtering the intermediate poisonous weed detection band information according to the band correlation and the band distance to obtain a plurality of poisonous weed detection band combination information; Calculating classification accuracy evaluation indicators of multiple poisonous weed band classification models based on the poisonous weed detection band combination information and multiple preset poisonous weed band classification models; screening a plurality of preset poisonous weed band classification models according to the classification accuracy evaluation index to obtain a target poisonous weed band classification model; Obtaining target location information of poisonous weeds in the first spectral image according to the poisonous weed detection band combination information and the target poisonous weed band classification model; The first spectral image information and the second spectral image information respectively represent two different spectral image information, wherein the first spectral image information represents spectral image information with high resolution reflecting the local situation of the grassland, and the second spectral image information represents spectral image information with low resolution reflecting the overall situation of the grassland.

3. A terminal device, characterized in that: The terminal device includes a memory and a processor, wherein the memory stores a computer program that can be run on the processor, and when the processor executes the computer program, the steps of the method according to claim 1 are implemented.

4. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method according to claim 1 are implemented.

Images