Invasive aquatic plant monitoring method, system, equipment and medium

Abstract

The invention discloses an invasive aquatic plant monitoring method, system and device and a medium, and relates to the field of biosafety invasive aquatic plant monitoring, and the method comprises the steps: obtaining a multispectral image of a to-be-detected water area; the multispectral image is shot by a multispectral imager carried by an unmanned aerial vehicle; according to the multispectral image of the to-be-detected water area and the classification model, determining a segmentation result of pixel points in the multispectral image of the to-be-detected water area; the classification model is obtained by training a random forest model; the segmentation result of the pixel points in the multispectral image comprises the pixel points of the water surface and the pixel points of pistia stratiotes intruding the aquatic plants; according to the segmentation result of the pixel points in the multispectral image of the to-be-detected water area, generating the intrusion hazard level of the pistia stratiotes in the to-be-detected water area; and giving an alarm according to the intrusion hazard level. Investibulators do not need to enter a water area, the monitoring precision is improved, and the monitoring cost is reduced.

Description

Technical Field

[0001] The present application relates to the field of biosafety invasive aquatic plant monitoring, and in particular to an invasive aquatic plant monitoring method, system, equipment and medium. Background Art

[0002] Pistia stratiotes L., as a kind of invasive aquatic plant, is the only species of the genus Pistia in the family Aracoae. It is a perennial aquatic floating herb. In the 1950s, it was widely planted in the waters of southern China as pig feed. After its introduction, no effective management strategy was formulated, resulting in the proliferation of Pistia stratiotes, which seriously threatened the biodiversity in various parts of southern China. As people pay more attention to ecological and environmental security, Pistia stratiotes has been included in the list of key invasive alien species. Pistia stratiotes spreads and reproduces quickly and has strong adaptability to the environment. After adaptation, it can often form very dense communities, thereby invading the growth environment of other organisms, making other aquatic plants unable to survive, and causing serious impacts on the stability of the aquatic ecosystem.

[0003] At present, the large-scale sedge is widely distributed, especially in areas with dense water networks. In recent years, the impact of eutrophication of water bodies and climate change has accelerated the spread of the large-scale sedge, and its distribution range has tended to expand northward. Monitoring the large-scale sedge is of great significance for protecting ecological balance, water environment management, protecting agriculture and aquaculture, and accumulating scientific research data. Since the large-scale sedge mainly grows on the water surface and has floating characteristics, traditional manual survey technology requires taking a boat to conduct sample plot surveys, which is difficult and time-consuming, and the personal safety of investigators cannot be guaranteed. The resolution of high-altitude remote sensing and satellite remote sensing images makes it difficult to identify specific invasive plants in the image, the monitoring accuracy is not high, and the monitoring cost is high.

[0004] Therefore, there is an urgent need to develop a new monitoring technology and alarm system to monitor the hazard level of invasive aquatic plants in water bodies and guide the prevention and control of invasive aquatic plant invasions in water bodies. Summary of the invention

[0005] The purpose of this application is to provide an invasive aquatic plant monitoring method, system, equipment and medium, so that monitoring personnel can monitor and alarm invasive aquatic plants in water areas without entering the water area, thereby improving monitoring accuracy and reducing monitoring costs.

[0006] To achieve the above objectives, this application provides the following solutions:

[0007] In a first aspect, the present application provides a method for monitoring invasive aquatic plants, comprising:

[0008] Acquire a multispectral image of the water area to be tested; the multispectral image is obtained by photographing with a multispectral imager mounted on an unmanned aerial vehicle;

[0009] Determine the segmentation result of the pixel points in the multispectral image of the water area to be tested according to the multispectral image and the classification model; the classification model is obtained by training the random forest model; the segmentation result of the pixel points in the multispectral image includes: the pixel points of the water surface and the pixel points of the invasive aquatic plant, the large sedge;

[0010] Generating the invasion hazard level of the giant hyacinth in the water area to be tested according to the segmentation result of the pixel points in the multispectral image of the water area to be tested;

[0011] An alarm is issued according to the intrusion hazard level.

[0012] Optionally, according to the multispectral image of the water area to be measured and the classification model, determining the segmentation result of the pixel points in the multispectral image of the water area to be measured specifically includes:

[0013] Extracting feature data of the multispectral image of the water area to be measured; the feature data includes: original spectral features, vegetation index features and texture features; the original spectral features include: red band features, green band features, blue band features and near infrared band features;

[0014] The characteristic data of the multispectral image of the water area to be tested is input into the classification model to obtain the segmentation result of the pixel points in the multispectral image of the water area to be tested.

[0015] Optionally, the method for determining the classification model specifically includes:

[0016] Acquire sample data; the sample data includes: a multispectral image of the training waters and a segmentation result of pixels in the multispectral image of the training waters;

[0017] Extract feature data of multispectral images of training waters;

[0018] Determine the characteristic data of the multispectral image of the training waters and the segmentation results of the pixel points in the multispectral image of the training waters as a classification sample set;

[0019] Dividing the classification sample set into a test set and a training set;

[0020] Using the training set to train a random forest model;

[0021] Testing the trained random forest model according to the test set to obtain a test result; the test result is the accuracy of the output result of the trained random forest model;

[0022] If the test result reaches the set overall classification accuracy, the trained random forest model is determined as the classification model;

[0023] If the test result does not reach the set overall classification accuracy, the number of the training sets is increased, and the random forest model is retrained until the test result reaches the set overall classification accuracy to obtain a classification model.

[0024] Optionally, extracting feature data of the multispectral image of the training waters specifically includes:

[0025] Calculating a normalized water index based on the multispectral image of the training waters, and removing the mask from the multispectral image of the training waters based on the normalized water index to obtain a masked multispectral image of the training waters;

[0026] A superpixel segmentation algorithm is used to perform superpixel segmentation on the multispectral image after the training water area mask, so as to generate superpixel clusters; the superpixel clusters have spectral consistency;

[0027] Classifying and labeling the superpixel clusters to obtain a labeled data set;

[0028] The annotated data set is statistically averaged to obtain characteristic data of the multispectral image of the training water area.

[0029] Optionally, the superpixel segmentation algorithm is a simple non-iterative algorithm.

[0030] Optionally, generating the invasion hazard level of the giant hyacinth in the waters to be tested according to the segmentation result of the pixel points in the multispectral image of the waters to be tested specifically includes:

[0031] Generate the coverage of invasive aquatic plants according to the segmentation results of pixel points in the multispectral image of the water area to be tested;

[0032] Generate an invasion hazard rating based on invasive aquatic plant cover.

[0033] Optionally, the expression for the invasive aquatic plant coverage is:

[0034] A=P i / (P i +P a +P w );

[0035] Among them, A represents the coverage of the water area invaded by the invasive aquatic plants, P i represents the number of pixels marked as invasive aquatic plants in the multispectral image of the water area to be tested, P a represents the number of pixels marked as aquatic plants other than P. wRepresents the number of pixels marked as water surface in the multispectral image of the water area to be tested.

[0036] In a second aspect, the present application provides an invasive aquatic plant monitoring system, comprising:

[0037] An image acquisition platform, a computing platform and a monitoring and alarm platform are connected in sequence; the image acquisition platform includes an unmanned aerial vehicle and a multispectral imager; the unmanned aerial vehicle is equipped with the multispectral imager; the multispectral imager is used to photograph the water area to be tested to obtain a multispectral image of the water area to be tested;

[0038] The computing platform comprises:

[0039] An image acquisition module, used to acquire a multispectral image of the water area to be tested;

[0040] A model determination module is used to determine a classification model; the classification model is obtained by training a random forest model;

[0041] The segmentation result determination module is used to determine the segmentation result of the pixel points in the multispectral image of the water area to be tested according to the multispectral image of the water area to be tested and the classification model; the segmentation result of the pixel points in the multispectral image includes: the pixel points of the water surface and the pixel points of the invasive aquatic plant, the large sedge;

[0042] The monitoring and alarm platform includes:

[0043] An invasion hazard level generation module is used to generate the invasion hazard level of the giant hyacinth in the water area to be tested according to the segmentation result of the pixel points in the multispectral image of the water area to be tested;

[0044] The alarm module is used to generate an alarm according to the intrusion hazard level.

[0045] In a third aspect, the present application provides a computer device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of any one of the above-described invasive aquatic plant monitoring methods.

[0046] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of any one of the above-described invasive aquatic plant monitoring methods.

[0047] According to the specific embodiments provided in this application, this application has the following technical effects:

[0048] The present application provides a method, system, equipment and medium for monitoring invasive aquatic plants. The method uses a drone to take images of the area to be monitored, without the need for investigators to enter the water surface for inspection, thereby ensuring the personal safety of investigators. The multispectral images taken by the drone equipped with a multispectral imager can more accurately identify invasive aquatic plants, reducing the high monitoring costs caused by the low resolution of remote sensing images. The method also uses a machine learning algorithm to efficiently and accurately identify aquatic invasive plants, other aquatic plants and water areas, making the monitoring data more accurate and able to prevent the outbreak of giant lichens in a timely manner. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. 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 paying creative work.

[0050] Figure 1 A schematic diagram of the process of the invasive aquatic plant monitoring method provided for this application;

[0051] Figure 2 A schematic diagram of the structure of the invasive aquatic plant monitoring system provided for this application;

[0052] Figure 3 A schematic diagram of the structure of a computer device provided in this application. DETAILED DESCRIPTION

[0053] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0054] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0055] In an exemplary embodiment, Figure 1 As shown, a method for monitoring invasive aquatic plants is provided, comprising:

[0056] Obtain a multispectral image of the water area to be tested; the multispectral image is obtained by photographing an unmanned aerial vehicle equipped with a multispectral imager.

[0057] According to the multispectral image of the water area to be tested and the classification model, the segmentation result of the pixel points in the multispectral image of the water area to be tested is determined; the classification model is obtained by training the random forest model; the segmentation result of the pixel points in the multispectral image includes: the pixel points of the water surface and the pixel points of the large sedge among the invasive aquatic plants.

[0058] The invasion hazard level of the large mulberry tree in the water area to be tested is generated according to the segmentation results of the pixel points in the multispectral image of the water area to be tested.

[0059] Alarm according to the intrusion hazard level.

[0060] In another exemplary embodiment of the present application, determining the segmentation result of the pixel points in the multispectral image of the water area to be tested according to the multispectral image of the water area to be tested and the classification model specifically includes:

[0061] The characteristic data of the multispectral image of the water area to be measured are extracted; the characteristic data include: original spectral characteristics, vegetation index characteristics and texture characteristics; the original spectral characteristics include: red band characteristics, green band characteristics, blue band characteristics and near infrared band characteristics; the vegetation index characteristics include: normalized difference vegetation index (NDVI), green normalized vegetation index (GNDVI), enhanced vegetation index (EVI), visual greenness index (VARI), ratio vegetation index (RVI), red-green ratio index (RGRI), normalized difference blue-green index (NDBGI) and red edge normalized vegetation index (NDRE). Texture characteristics include mean, standard deviation, skewness, kurtosis, contrast, energy, entropy, correlation and uniformity. Specifically, the normalized water index is calculated based on the green band characteristics and near infrared band characteristics, and the vegetation index characteristics and texture characteristics are calculated based on the original spectral characteristics.

[0062] The characteristic data of the multispectral image of the water area to be tested is input into the classification model to obtain the segmentation result of the pixel points in the multispectral image of the water area to be tested.

[0063] In another exemplary embodiment of the present application, a method for determining a classification model specifically includes:

[0064] Obtain sample data; the sample data includes: a multispectral image of the training waters and a segmentation result of pixels in the multispectral image of the training waters.

[0065] Extract feature data from multispectral images of training waters.

[0066] The characteristic data of the multispectral image of the training waters and the segmentation results of the pixel points in the multispectral image of the training waters are determined as a classification sample set.

[0067] The classification sample set is divided into a test set and a training set; the number of pixels of any one of the four types of water plants, water surface and background in the training set is at least 200. 80% of the samples in the classification sample set are used as the training set, and the remaining 20% ​​of the samples are used as the validation set.

[0068] The training set is used to train the random forest model.

[0069] The trained random forest model is tested according to the test set to obtain the test result; the test result is the accuracy of the output result of the trained random forest model.

[0070] If the test results reach the set overall classification accuracy, the trained random forest model will be determined as the classification model.

[0071] If the test result does not reach the set overall classification accuracy, the number of training sets is increased and the random forest model is retrained until the test result reaches the set overall classification accuracy to obtain the classification model.

[0072] In another exemplary embodiment of the present application, extracting feature data of a multispectral image of a training water area specifically includes:

[0073] The normalized water index is calculated based on the multispectral image of the training water area, and the multispectral image of the training water area is masked based on the normalized water index to obtain the masked multispectral image of the training water area. The normalized water index is a remote sensing index used to enhance the detection and identification of water bodies. The threshold for extracting water bodies from this index is 0, that is, when the index is greater than 0, the target object is judged to be a water body.

[0074] The superpixel segmentation algorithm is used to perform superpixel segmentation on the multispectral image after the training water area mask to generate superpixel clusters; the superpixel clusters have spectral consistency. The superpixel clusters are specifically small area units with spectral consistency.

[0075] The superpixel clusters were classified and labeled to obtain a labeled dataset, which contained superpixel clusters of three types of ground features: the invasive aquatic plant Euphorbia cerasifera, other aquatic plants, and background.

[0076] The annotated data set is statistically averaged to obtain the feature data of the multispectral image of the training waters. Specifically, the original spectral features, vegetation index features, and texture features in each cluster of the annotated data set are statistically averaged to give these clusters feature attributes. These clusters with feature attributes are used as classification sample sets.

[0077] This application inputs the training set into the random forest machine learning classifier, builds a trained random forest model to classify the image, and identifies the target invasive plant, the giant lily, other aquatic plants, and the background. Then the trained random forest model is tested with the test set. If the test result meets the expectation, the model can be used to effectively classify and extract the target invasive plant, the giant lily; if the test result does not meet the standard, the superpixel segmentation parameters and the number of samples are adjusted, and the model is retrained until the accuracy of the test set test model classification is greater than or equal to the expected standard. The new model obtained is used for the analysis of multispectral images of drones in this water area.

[0078] In another exemplary embodiment of the present application, the superpixel segmentation algorithm is a simple non-iterative algorithm.

[0079] In another exemplary embodiment of the present application, the invasion hazard level of the giant hyacinth in the water area to be tested is generated according to the segmentation result of the pixel points in the multispectral image of the water area to be tested, specifically including:

[0080] The coverage (area percentage) of invasive aquatic plants is generated based on the segmentation results of pixel points in the multispectral image of the water area to be tested.

[0081] Generate an invasion hazard rating based on invasive aquatic plant cover.

[0082] In another exemplary embodiment of the present application, the expression of invasive aquatic plant coverage is:

[0083] A=P i / (P i +P a +P w ).

[0084] Among them, A represents the coverage (ratio) of the water area invaded by the invasive aquatic plants, P i represents the number of pixels marked as invasive aquatic plants in the multispectral image of the water area to be tested, P a represents the number of pixels marked as aquatic plants other than P. w Represents the number of pixels marked as water surface in the multispectral image of the water area to be tested.

[0085] The present application relates to invasive plant monitoring and alarm, and specifically discloses a monitoring and alarm system for the invasive aquatic plant Eupatorium truncatum. The present application uses an unmanned aerial vehicle equipped with a multispectral imager to capture images of the monitored area, and based on superpixel segmentation and random forest machine learning algorithms, efficiently and accurately identifies the aquatic invasive plant Eupatorium truncatum, other aquatic plants and water area, evaluates the invasion level according to the proportion of the invasive aquatic plant Eupatorium truncatum on the water surface, and uses emails or text messages to alarm the invasion degree of Eupatorium truncatum, which can intervene in time to prevent Eupatorium truncatum outbreaks.

[0086] Based on the same inventive concept, the embodiment of the present application also provides an invasive aquatic plant monitoring system for implementing the invasive aquatic plant monitoring method involved above. The implementation scheme for solving the problem provided by the system is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more invasive aquatic plant monitoring system embodiments provided below can refer to the limitations of the invasive aquatic plant monitoring method above, and will not be repeated here.

[0087] In an exemplary embodiment, Figure 2 As shown, a system for monitoring invasive aquatic plants is provided, comprising:

[0088] The image acquisition platform, computing platform and monitoring and alarm platform are connected in sequence; the image acquisition platform includes a drone and a multispectral imager; the drone is equipped with a multispectral imager; the multispectral imager is used to photograph the water area to be tested to obtain a multispectral image of the water area to be tested. The computing platform is generally a smart phone. The drone is connected to the multispectral imager to shoot multispectral images of the water area over the water area, with the heading overlap and lateral overlap of not less than 60%, and the GPS positioning coordinates, picture shooting time and other information are recorded.

[0089] The computing platform includes:

[0090] The image acquisition module is used to acquire the multispectral image of the water area to be tested.

[0091] The model determination module is used to determine the classification model; the classification model is obtained by training the random forest model.

[0092] The segmentation result determination module is used to determine the segmentation results of pixel points in the multispectral image of the water area to be tested according to the multispectral image of the water area to be tested and the classification model; the segmentation results of pixel points in the multispectral image include: pixel points of the water surface and pixel points of the large sedge in the invasive aquatic plant.

[0093] The monitoring and alarm platform includes:

[0094] The invasion hazard level generation module is used to generate the invasion hazard level of the large mulberry tree in the water area to be tested according to the segmentation result of the pixel points in the multispectral image of the water area to be tested.

[0095] The alarm module is used to generate an alarm based on the intrusion hazard level.

[0096] According to the agricultural industry standard "Technical Procedures for Monitoring Invasive Alien Plants - Psoralea corylifolia (NY / T 3076-2017)", the invasion hazard of Psoralea corylifolia is divided into three levels: mild occurrence (cover <5%), moderate occurrence (cover 5% to 20%) and severe occurrence (cover >20%); when the invasion hazard level is severe, an SMS or email alarm will be sent to the management personnel.

[0097] As an optional implementation, the invasive aquatic plant monitoring system also includes: a (ground) control platform; the (ground) control platform is connected to the image acquisition platform and the computing platform respectively; the drone can be connected to the (ground) control platform via WiFi, and the (ground) control platform is connected to the computing platform via Bluetooth or WiFi.

[0098] The (ground) control platform is used to control the flight of the UAV, the multispectral imager to capture the multispectral images of the water area to be tested, and the multispectral images of the water area to be tested are received.

[0099] As an optional implementation, the drone is also used to record GPS positioning coordinates and image capture time.

[0100] The computing platform also includes:

[0101] The distribution heat map generating unit is used to generate a heat map of the distribution of the large mulberry tree according to the segmentation results of the pixel points in the multispectral image of the water area to be tested.

[0102] The sending unit is used to send the GPS positioning coordinates, image shooting time, segmentation results of pixel points in the multispectral image of the water area to be tested, the multispectral image of the water area to be tested, the coverage of invasive aquatic plants and the heat map of the distribution of the large caltrops to the monitoring and alarm platform.

[0103] The image determination module is specifically used to perform image stitching on the multispectral images of the water area to be tested to obtain a multispectral image of the entire water area to be tested.

[0104] The monitoring and alarm platform is placed on the cloud server, and managers can manually review the information sent by the computing platform. For example, if the hazard level of the giant hyacinth is severe or above, the severe occurrence and severe occurrence data received can be used to send SMS or email alarms to relevant managers through the network.

[0105] The computing platform uses the trained model to analyze the new multispectral images, and sends the multispectral images, distribution heat maps, GPS positioning coordinates, image shooting time, large lily coverage, and large lily invasion hazard level to the monitoring and alarm platform in the cloud.

[0106] Beneficial effects of this application:

[0107] 1. This application provides high-resolution multispectral images, which contain more image information in addition to visible light and can more accurately identify the invasive aquatic plant Macrophylla.

[0108] 2. This application uses a random forest model for machine learning, and the resulting model is smaller. It does not require large-scale computing power, and local multispectral image recognition and segmentation can be achieved using a smartphone.

[0109] 3. This application can calculate the hazard level of the water area invaded by the giant lichen and report it to the monitoring and alarm platform, making it convenient to monitor and manage the invasion of the giant lichen.

[0110] 4. The monitoring system of this application does not require investigators to enter the water surface for detection, which ensures the personal safety of investigators and makes the monitoring data more accurate.

[0111] The implementation of invasive aquatic plant monitoring for this application requires the following steps:

[0112] Set the flight route of the drone (DJI Mavic 3M), with a flight altitude of no less than 20 meters, and a heading overlap and a lateral overlap of no less than 60%. The multispectral imager should include at least four bands: red, green, blue, and near-infrared. According to the planned route, perform fixed-point hovering photography over the water area, obtain a large number of multispectral images of the water area, and record GPS positioning coordinates, picture shooting time and other information. Ensure that the ground sampling distance resolution of the captured image is at least better than 10 cm.

[0113] Image stitching: Obtain high-resolution multispectral images of the entire water area. First, the scale-invariant feature transformation principle is used to detect key points between images. For each detected key point, a feature vector is generated. By calculating the Euclidean distance of the key points between the images, similar key point pairs are found. Based on the matched key point pairs, the geometric transformation matrix between the two images is estimated. Use the calculated geometric transformation matrix to project one image into the coordinate system of another image to achieve image alignment. The overlapping areas are fused (such as weighted averaging and smooth transition) to eliminate seams. Finally, the entire water area image is obtained.

[0114] Based on the original green band and near infrared band of the multispectral image, the normalized water index is calculated by: (green band - near infrared band) / (green band + near infrared band). The threshold is set to 0, and the area greater than the threshold is determined to be a water body. The water body is masked out.

[0115] Vegetation index features and texture features are calculated based on the original bands. The calculated vegetation index features include normalized vegetation index, green normalized vegetation index, enhanced vegetation index, visual greenness index, ratio vegetation index, red-green ratio index, normalized difference blue-green index, and red-edge normalized vegetation index, a total of 8 vegetation indices. Texture features include mean, standard deviation, skewness, kurtosis, entropy, contrast, energy, correlation, and uniformity, a total of 9 types. Because texture features are extracted based on each original band separately, 4 bands generate 4*9 features for a total of 36 features.

[0116] The image after water body masking is subjected to superpixel segmentation to generate superpixel clusters with spectral consistency. The boundary information of the cluster is obtained; the superpixel segmentation uses a simple non-iterative algorithm (SNIC) to segment the original band and obtain a unified cluster center based on spatial heterogeneity and spatial autocorrelation. Specifically, the initial cluster center, namely seed initialization, is first defined, and pixels are assigned to the nearest cluster center based on color similarity and spatial proximity. Then the seed position is refined by iteratively refining the seed position by calculating the average color and spatial coordinates of the pixels assigned to each seed. The parameters involved in superpixel segmentation include superpixel size, compactness, connectivity, and neighborhood size, which are set to 50, 10, 8, and 15, respectively.

[0117] Superpixel clusters were classified and labeled, and at least 500 labeled dataset sample points containing three types of ground features, including the invasive aquatic plant Macrophylla sphaerocephala, other aquatic plants, and background, were randomly selected through visual interpretation and saved as point vector files.

[0118] The characteristic information of the cluster is obtained according to the annotated data set, that is, the original spectral features (4), vegetation index features (10) and texture features (36) in each cluster are statistically averaged, and the characteristic averages are assigned as the characteristic attributes of these clusters. These clusters containing characteristic attributes are used as classification sample sets. 80% of the samples are used as training sets, and the remaining 20% ​​of the samples are used as validation sets.

[0119] The classification accuracy evaluation index is the overall classification accuracy (OverallAccuracy, OA), and the calculation formula is as follows:

[0120]

[0121] n ii is the number of correctly classified samples of the i-th category in the confusion matrix (diagonal elements). k is the total number of classified categories. N is the total number of samples (the sum of all elements in the confusion matrix).

[0122] The random forest machine learning classifier was used as the classification model. The training samples were input into the random forest machine learning classifier, and a classification model was constructed to classify the entire image and identify the target invasive plant, the large sedge, other aquatic plants, and the background. Training model parameters: the number of decision trees ntree = 500, the maximum number of nodes in a single tree maxnodes = 50, and the feature sampling parameter mtry was set to the square root of the number of features (4+10+36), which is

[0123] After the model parameters are set, the trained model is tested using the test set. If the test results meet expectations (overall classification accuracy OA ≥ 75%), the model can be used to effectively classify and extract the target invasive plant, D. serrata. If the test results do not meet the standards, the superpixel segmentation parameters and sample size are adjusted, and the model is retrained until the classification accuracy of the test set test model is greater than or equal to the expected standard. The new model obtained is used for the analysis of multispectral images of drones in this water area.

[0124] Based on the classification results, a heat map of the distribution of P. truncatum was generated, and the coverage (area percentage) of P. truncatum invasion in the water area was calculated using the number of pixels.

[0125] The trained model is used to analyze the new multispectral image using a computing platform (usually a smart phone), and the multispectral image, distribution heat map, GPS positioning coordinates, image shooting time, sedge cover, and sedge invasion hazard level are sent to the monitoring and alarm platform in the cloud.

[0126] The management personnel manually review the information sent by the computing platform, and the received severe occurrence and severe occurrence data can be used to send SMS or email alarms to the management personnel through the network.

[0127] In an exemplary embodiment, a computer device is provided. The computer device may be a server or a terminal. The internal structure diagram thereof may be as follows: Figure 3As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store invasive aquatic plant monitoring data. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a method for monitoring invasive aquatic plants is implemented.

[0128] Those skilled in the art will understand that Figure 3 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components. In an exemplary embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above-mentioned method embodiments when executing the computer program.

[0129] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0130] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0131] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0132] The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. The non-relational database may include a distributed database based on blockchain, etc., but is not limited thereto. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but is not limited thereto.

[0133] The technical features of the above embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0134] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. At the same time, for those skilled in the art, according to the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A method for monitoring invasive aquatic plants, characterized in that: The invasive aquatic plant monitoring method comprises: Acquire a multispectral image of the water area to be tested; the multispectral image is obtained by photographing with a multispectral imager mounted on an unmanned aerial vehicle; Determine the segmentation result of the pixel points in the multispectral image of the water area to be tested according to the multispectral image and the classification model; the classification model is obtained by training the random forest model; the segmentation result of the pixel points in the multispectral image includes: the pixel points of the water surface and the pixel points of the invasive aquatic plant, the large sedge; Generating the invasion hazard level of the giant hyacinth in the water area to be tested according to the segmentation result of the pixel points in the multispectral image of the water area to be tested; An alarm is issued according to the intrusion hazard level.

2. The method for monitoring invasive aquatic plants according to claim 1, characterized in that: According to the multispectral image of the water area to be tested and the classification model, the segmentation result of the pixel points in the multispectral image of the water area to be tested is determined, specifically including: Extracting feature data of the multispectral image of the water area to be measured; the feature data includes: original spectral features, vegetation index features and texture features; the original spectral features include: red band features, green band features, blue band features and near infrared band features; The characteristic data of the multispectral image of the water area to be tested is input into the classification model to obtain the segmentation result of the pixel points in the multispectral image of the water area to be tested.

3. The method for monitoring invasive aquatic plants according to claim 1, characterized in that: The method for determining the classification model specifically includes: Acquire sample data; the sample data includes: a multispectral image of the training waters and a segmentation result of pixels in the multispectral image of the training waters; Extract feature data of multispectral images of training waters; Determine the characteristic data of the multispectral image of the training waters and the segmentation results of the pixel points in the multispectral image of the training waters as a classification sample set; Dividing the classification sample set into a test set and a training set; Using the training set to train a random forest model; Testing the trained random forest model according to the test set to obtain a test result; the test result is the accuracy of the output result of the trained random forest model; If the test result reaches the set overall classification accuracy, the trained random forest model is determined as the classification model; If the test result does not reach the set overall classification accuracy, the number of the training sets is increased, and the random forest model is retrained until the test result reaches the set overall classification accuracy to obtain a classification model.

4. The method for monitoring invasive aquatic plants according to claim 3, characterized in that: Extract feature data of multispectral images of training waters, including: Calculating a normalized water index based on the multispectral image of the training waters, and removing the mask from the multispectral image of the training waters based on the normalized water index to obtain a masked multispectral image of the training waters; A superpixel segmentation algorithm is used to perform superpixel segmentation on the multispectral image after the training water area mask, so as to generate superpixel clusters; the superpixel clusters have spectral consistency; Classifying and labeling the superpixel clusters to obtain a labeled data set; The annotated data set is statistically averaged to obtain characteristic data of the multispectral image of the training water area.

5. The method for monitoring invasive aquatic plants according to claim 4, characterized in that: The superpixel segmentation algorithm is a simple non-iterative algorithm.

6. The method for monitoring invasive aquatic plants according to claim 1, characterized in that: The invasion hazard level of the giant hyacinth in the waters to be tested is generated according to the segmentation results of the pixel points in the multispectral image of the waters to be tested, specifically including: Generate the coverage of invasive aquatic plants according to the segmentation results of pixel points in the multispectral image of the water area to be tested; Generate an invasion hazard rating based on invasive aquatic plant cover.

7. The method for monitoring invasive aquatic plants according to claim 6, characterized in that: The expression of the invasive aquatic plant cover is: A=P i / (P i +P a +P w ); Among them, A represents the coverage of the water area invaded by the invasive aquatic plants, P i represents the number of pixels marked as invasive aquatic plants in the multispectral image of the water area to be tested, P a represents the number of pixels marked as aquatic plants other than P. w Represents the number of pixels marked as water surface in the multispectral image of the water area to be tested.

8. An invasive aquatic plant monitoring system, characterized in that: The invasive aquatic plant monitoring system comprises: an image acquisition platform, a computing platform and a monitoring alarm platform connected in sequence; the image acquisition platform comprises an unmanned aerial vehicle and a multispectral imager; the unmanned aerial vehicle is equipped with the multispectral imager; the multispectral imager is used to photograph the water area to be tested to obtain a multispectral image of the water area to be tested; The computing platform comprises: An image acquisition module, used to acquire a multispectral image of the water area to be tested; A model determination module is used to determine a classification model; the classification model is obtained by training a random forest model; The segmentation result determination module is used to determine the segmentation result of the pixel points in the multispectral image of the water area to be tested according to the multispectral image of the water area to be tested and the classification model; the segmentation result of the pixel points in the multispectral image includes: the pixel points of the water surface and the pixel points of the invasive aquatic plant, the large sedge; The monitoring and alarm platform includes: An invasion hazard level generation module is used to generate the invasion hazard level of the giant hyacinth in the water area to be tested according to the segmentation result of the pixel points in the multispectral image of the water area to be tested; The alarm module is used to generate an alarm according to the intrusion hazard level.

9. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the invasive aquatic plant monitoring method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the invasive aquatic plant monitoring method described in any one of claims 1 to 7 is implemented.

Images