Methods, equipment, storage media and devices for tree identification by drones

Abstract

This invention discloses a method, device, storage medium, and apparatus for tree identification using a drone. The invention preprocesses remote sensing image information acquired by the drone, segments the remote sensing tree images based on an improved Mask-R-CNN model, and identifies tree species based on a preset tree species model and tree image information, obtaining image sets corresponding to each tree species type. Based on a preset morphological model and the image sets of each tree species type, the invention predicts the growth trend of trees within a target area, and determines the perceived growth status of trees in the target area based on the prediction results. This invention segments the preprocessed remote sensing tree images using an improved Mask-R-CNN model and identifies tree species based on a preset tree species model, thereby determining the growth status of each type of tree. This invention achieves accurate vegetation identification and precise prediction of tree growth status using a drone, improving data accuracy.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, and in particular to a method, device, storage medium and apparatus for tree recognition by unmanned aerial vehicles (UAVs). Background Technology

[0002] With global warming or other reasons, it is necessary to determine the parameters of natural forests and planted trees (such as determining the carbon emissions of trees through vegetation area and the number of trees). However, due to terrain or other reasons, trees in many places cannot be identified manually to collect tree-related data. Therefore, drones are often used to collect tree images for subsequent identification. Although drone vegetation monitoring can provide great convenience, existing drone tree identification methods have problems such as incomplete identification processes and low accuracy, resulting in inaccurate data and affecting subsequent data analysis.

[0003] The above content is only used to help understand the technical solution of the present invention and does not represent an admission that the above content is prior art. Summary of the Invention

[0004] The main objective of this invention is to provide a method, device, storage medium, and apparatus for tree identification using unmanned aerial vehicles (UAVs), aiming to solve the technical problem of poor accuracy in tree identification results due to complex terrain and limitations in image processing when using UAVs in the prior art.

[0005] To achieve the above objectives, the present invention provides a method for identifying trees using a drone, the method comprising the following steps:

[0006] Preprocessing of remote sensing image information collected by UAVs yields processed remote sensing tree images;

[0007] The remote sensing tree image is segmented based on the improved Mask-R-CNN model to obtain the segmented tree image information;

[0008] Based on the preset tree species model and the tree image information, the tree species types are identified to obtain the image set corresponding to each tree species type;

[0009] Based on a preset morphological model and image sets of various tree species, the growth trend of trees in the target area is predicted, and the growth status of trees in the target area is determined according to the prediction results.

[0010] Optionally, the step of preprocessing the remote sensing image information collected by the UAV to obtain processed remote sensing tree images includes:

[0011] The remote sensing image information collected by the UAV is segmented based on the SLIC algorithm to obtain a set of segmented remote sensing images.

[0012] The SLIC algorithm is used to extract color and texture features from the remote sensing image set, and the remote sensing tree images in the remote sensing image set are determined based on the color and texture features.

[0013] Optionally, the step of segmenting the remote-sensed tree image based on the improved Mask-R-CNN model to obtain segmented tree image information includes:

[0014] The remote sensing tree image is segmented and identified based on the improved Mask-R-CNN model and the preset cross-entropy loss function to obtain the contour parameters corresponding to the target tree crown.

[0015] The centroid coordinates of the target tree canopy are determined based on a preset boundary tracking algorithm and the contour parameters.

[0016] The segmented tree image information is determined based on the contour parameters and the centroid coordinates.

[0017] Optionally, the step of identifying tree species types based on a preset tree species model and the tree image information to obtain an image set corresponding to each tree species type includes:

[0018] The tree crown type is identified based on the preset tree species model, the outline parameters, and the centroid coordinates, and the tree species type and quantity are determined based on the identification results;

[0019] The remote sensing tree images are classified according to the tree species type and quantity to obtain the image set corresponding to each tree species type.

[0020] Optionally, the step of predicting the growth trend of trees in the target area based on a preset morphological model and image sets of various tree species, and determining the tree growth status perception corresponding to the target area based on the prediction results, includes:

[0021] Based on a pre-defined morphological model, the morphological features of trees in image sets of various tree species are compared, and the growth age of trees in the target area is determined based on the comparison results.

[0022] Based on the growth age, the growth trend of trees in the target area is predicted to obtain the prediction results;

[0023] Based on the prediction results and the preset Markov model, the tree growth status perception corresponding to the target area is determined.

[0024] Optionally, the step of determining the tree growth status perception corresponding to the target area based on the prediction result and the preset Markov model includes:

[0025] The growth rate of the target region within a preset period is determined based on the prediction results and the preset Markov model.

[0026] The predicted area is obtained by predicting the area occupied by trees in the target area based on the growth rate.

[0027] The tree growth status is determined based on the predicted area and a preset clustering algorithm.

[0028] Optionally, before the step of preprocessing the remote sensing image information collected by the UAV to obtain the processed remote sensing tree image, the method further includes:

[0029] Obtain growth sample data for each type of tree species;

[0030] The growth sample data is input into the initial Mask-R-CNN model for training, and the training results are labeled according to the preset labeling tool to obtain the labeled sample dataset;

[0031] The initial Mask-R-CNN model is iteratively trained based on the labeled sample dataset until the output training result meets the preset conditions, and the trained Mask-R-CNN model is used as the improved Mask-R-CNN model.

[0032] Furthermore, to achieve the above objectives, the present invention also proposes a drone tree recognition device, which includes a memory, a processor, and a drone tree recognition program stored in the memory and executable on the processor. The drone tree recognition program is configured to implement the drone tree recognition steps described above.

[0033] In addition, to achieve the above objectives, the present invention also proposes a storage medium storing a drone tree recognition program, which, when executed by a processor, implements the steps of the drone tree recognition method described above.

[0034] Furthermore, to achieve the above objectives, the present invention also proposes a drone tree identification device, the drone tree identification device comprising:

[0035] The image preprocessing module is used to preprocess the remote sensing image information collected by the UAV to obtain the processed remote sensing tree image;

[0036] The image segmentation module is used to segment the remotely sensed tree image based on the improved Mask-R-CNN model to obtain segmented tree image information;

[0037] The tree species identification module is used to identify tree species types based on a preset tree species model and the tree image information, and obtain an image set corresponding to each tree species type;

[0038] The situation awareness module is used to predict the growth trend of trees in a target area based on a preset morphological model and image sets of various tree species, and to determine the tree growth situation awareness corresponding to the target area based on the prediction results.

[0039] This invention preprocesses remote sensing image information collected by a drone to obtain processed remote sensing tree images; segments the remote sensing tree images based on an improved Mask-R-CNN model to obtain segmented tree image information; identifies tree species based on a preset tree species model and the tree image information to obtain image sets corresponding to each tree species type; and predicts the growth trend of trees in a target area based on a preset morphological model and the image sets of each tree species type, determining the perceived growth status of trees in the target area based on the prediction results. Because this invention segments the preprocessed remote sensing tree images using the improved Mask-R-CNN model and identifies tree species based on a preset tree species model, thereby further determining the growth status of each type of tree, it significantly improves data accuracy compared to existing drone tree identification methods which suffer from incomplete identification processes and low accuracy, leading to inaccurate data. This invention achieves accurate vegetation identification and precise prediction of tree growth status simultaneously through drones, thus enhancing data accuracy. Attached Figure Description

[0040] Figure 1 This is a schematic diagram of the structure of a drone tree recognition device in the hardware operating environment of the embodiment of the present invention;

[0041] Figure 2 This is a flowchart illustrating the first embodiment of the UAV tree identification method of the present invention;

[0042] Figure 3 This is a flowchart illustrating the second embodiment of the UAV tree identification method of the present invention;

[0043] Figure 4 This is a flowchart illustrating the third embodiment of the UAV tree identification method of the present invention;

[0044] Figure 5 This is a structural block diagram of the first embodiment of the tree identification device for unmanned aerial vehicles (UAVs) of the present invention.

[0045] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0046] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0047] Reference Figure 1 , Figure 1 This is a schematic diagram of the structure of a drone tree recognition device in the hardware operating environment of an embodiment of the present invention.

[0048] like Figure 1 As shown, the UAV tree recognition device may include: a processor 1001, such as a central processing unit (CPU), a communication bus 1002, a user interface 1003, a network interface 1004, and a memory 1005. The communication bus 1002 is used to enable communication between these components. The user interface 1003 may include a display screen, and optionally, it may also include a standard wired interface or a wireless interface. In this invention, the wired interface of the user interface 1003 may be a USB interface. The network interface 1004 may optionally include a standard wired interface or a wireless interface (such as a Wireless-Fidelity (Wi-Fi) interface). The memory 1005 may be a high-speed random access memory (RAM) or a non-volatile memory (NVM), such as a disk storage device. The memory 1005 may also optionally be a storage device independent of the aforementioned processor 1001.

[0049] Those skilled in the art will understand that Figure 1 The structure shown does not constitute a limitation on the drone tree recognition device and may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0050] like Figure 1 As shown, the memory 1005, which is identified as a computer storage medium, may include an operating system, a network communication module, a user interface module, and a UAV tree recognition program.

[0051] exist Figure 1In the drone tree recognition device shown, the network interface 1004 is mainly used to connect to the backend server and communicate data with the backend server; the user interface 1003 is mainly used to connect to the user equipment; the drone tree recognition device calls the drone tree recognition program stored in the memory 1005 through the processor 1001 and executes the drone tree recognition method provided in the embodiment of the present invention.

[0052] Based on the above hardware structure, an embodiment of the UAV tree recognition method of the present invention is proposed.

[0053] Reference Figure 2 , Figure 2 This is a flowchart illustrating the first embodiment of the UAV tree identification method of the present invention, which presents the first embodiment of the UAV tree identification method of the present invention.

[0054] In this embodiment, the drone tree identification method includes the following steps:

[0055] Step S10: Preprocess the remote sensing image information collected by the UAV to obtain processed remote sensing tree images.

[0056] It should be noted that the executing entity in this embodiment can be a device with drone tree recognition function, such as a drone, computer, laptop, PC, or tablet, or other drone tree recognition devices that can achieve the same or similar functions. This embodiment does not limit this. The above-mentioned computer will be used as an example to describe this embodiment and the following embodiments.

[0057] Understandably, drones can be used for tree identification in various scenarios. However, in existing drone-based identification of complex and steep terrain, issues arise due to the complex terrain effects: large spatial variability across regions, ground heterogeneity, incomplete ground sampling, and image distortion caused by flight altitude. These issues lead to a lack of data integrity. Furthermore, field surveys have extremely high resolution, distinguishing dead leaves and vegetation, while aerial imagery can only distinguish vegetation patches, resulting in drone-measured vegetation cover generally being higher than that obtained through field surveys. Therefore, to overcome these problems, it is necessary to preprocess the remote sensing images acquired by drones to reduce the incompleteness or distortion of drone images caused by these issues, which could prevent accurate tree identification for subsequent image recognition.

[0058] It should be understood that before preprocessing the acquired remote sensing images, the camera configured on the UAV needs to be calibrated. The calibration process involves determining the displacement of the image point captured by the UAV based on parameters such as the UAV's ground speed, camera exposure time, focal length, and flight altitude, and then calibrating the camera based on this displacement. Calibrating the UAV camera in this way can improve mapping accuracy. The preprocessing can involve segmenting the remote sensing image information acquired by the UAV using the SLIC algorithm to obtain processed remote sensing tree images. The remote sensing image information can be video information acquired by the UAV, containing multiple frames of remote sensing images. To avoid duplicate images, images with a similarity rate greater than a preset similarity threshold between adjacent frames are deleted. The filtered remote sensing image information is then segmented using the SLIC algorithm to obtain processed remote sensing tree images. The remote sensing tree images can be images containing trees identified from the filtered remote sensing image information using the SLIC algorithm.

[0059] In practice, the SLIC algorithm is used to preprocess the remote sensing image information collected by the UAV to obtain the processed remote sensing tree image.

[0060] Further, step S10 includes: segmenting the remote sensing image information collected by the UAV based on the SLIC algorithm to obtain a segmented remote sensing image set; extracting color features and texture features from the remote sensing image set according to the SLIC algorithm; and identifying remote sensing tree images in the remote sensing image set based on the color features and texture features.

[0061] It should be noted that the SLIC (Simple linear iterative clustering) algorithm is a superpixel segmentation algorithm based on the k-means algorithm. The SLIC algorithm divides the original pixel-level image into region-level images, that is, it further refines the segmented image, thereby achieving more accurate image segmentation for subsequent recognition and processing.

[0062] The SLIC algorithm initializes cluster centers by uniformly distributing them across the image according to a predetermined number of superpixels. Assuming an image has A pixels, pre-segmented into B superpixels of the same size, each superpixel is A / B in size, and the distance between adjacent cluster centers is approximately S = sqrt(A / B). Then, cluster centers are reselected within the n*n neighborhood of each cluster center (typically n = 3). Specifically, the gradient values ​​of all pixels within this neighborhood are calculated, and the cluster center is moved to the location with the smallest gradient within that neighborhood. This is done to avoid cluster centers falling on contour boundaries with large gradients, which could affect subsequent clustering results. Finally, a class label (i.e., indicating which cluster center a pixel belongs to) is assigned to each pixel within its neighborhood. This differs from the basic k-means search method across the entire image; SLIC's search range is limited to 2S*2S, which accelerates algorithm convergence, as shown in the figure below. Note that the desired superpixel size is SS, but the search range is 2S*2S. For each found pixel, the distance between it and its cluster center is calculated. Specifically, the final distance metric is calculated using color distance, spatial distance, and distances to adjacent cluster centers. Since each pixel is searched by multiple cluster centers, each pixel has a distance to its surrounding cluster centers. The cluster center corresponding to the minimum distance is taken as the pixel's cluster center. A preset iterative algorithm is then used for optimization until convergence, achieving the desired segmentation effect.

[0063] Understandably, the remote sensing image information acquired by the UAV is segmented based on the SLIC algorithm to obtain a set of remote sensing images corresponding to a preset number of irregular pixel blocks. The SLIC algorithm is then used to extract remote sensing images corresponding to pixel blocks with color and texture features similar to tree features from the remote sensing image set, generating a corresponding set. Finally, remote sensing tree images are extracted from this set.

[0064] Step S20: Segment the remote sensing tree image based on the improved Mask-R-CNN model to obtain segmented tree image information.

[0065] It should be noted that the improved Mask-R-CNN model can be an improved version of the Faster R-CNN model. The Mask-R-CNN structure includes a backbone network layer, region network layers, RoI alignment layers, bounding boxes, classification, and masks. The improved Mask-R-CNN model can employ a bottom-up approach to achieve feature fusion between upper and lower layers of the network. This Mask-R-CNN model is used to process high-resolution UAV images of large-scale forest areas with mixed species, thereby simultaneously solving the goals of individual tree canopy segmentation, species classification, and counting detection. The model modifies the top-down feature fusion feature of the FPN network to reduce the feature fusion path between lower and upper layers. This is added to the cross-entropy loss function through a boundary-weighted loss module as an improvement to the prediction algorithm at target boundaries.

[0066] Understandably, the remotely sensed tree images are segmented based on the improved Mask-R-CNN model to obtain segmented tree image information.

[0067] Furthermore, before step S10, the method further includes: acquiring growth sample data corresponding to each type of tree species; inputting the growth sample data into the initial Mask-R-CNN model for training, and labeling the training results according to a preset labeling tool to obtain a labeled sample dataset; iteratively training the initial Mask-R-CNN model based on the labeled sample dataset until the output training results meet the preset conditions, and using the trained Mask-R-CNN model as the improved Mask-R-CNN model.

[0068] It should be noted that growth sample data corresponding to various types of tree species is obtained. The growth sample data includes growth sample datasets corresponding to various types of trees at different times. The growth sample datasets are divided into training sets, validation sets, and test sets. The initial Mask-R-CNN model is trained, and the training results are labeled according to a preset labeling tool to obtain the labeled sample dataset. The preset labeling tool can be a pre-set tool for assigning values ​​to feature points.

[0069] Understandably, the initial Mask-R-CNN model is iteratively trained based on the labeled sample dataset until the output training result meets the preset conditions, and the trained Mask-R-CNN model is used as the improved Mask-R-CNN model. To improve the completeness of the dataset, the images in the training and validation sets are expanded by translation, rotation, and inversion to fully extract feature points, achieve the best training effect, and output the optimal model.

[0070] In practice, the initial model can be trained on the growth sample data corresponding to a single tree, and then the number of trees can be added up sequentially until the training result meets the preset conditions, and the Mask-R-CNN model is output.

[0071] Step S30: Identify tree species types based on the preset tree species model and the tree image information to obtain the image set corresponding to each tree species type.

[0072] It should be noted that the preset tree species model can be a pre-set model for identifying tree species, which is based on the segmented tree image information.

[0073] Step S40: Based on the preset morphological model and the image set of each tree species, predict the growth trend of trees in the target area, and determine the tree growth status perception corresponding to the target area according to the prediction results.

[0074] It should be noted that the preset morphological model can be a pre-set model for identifying tree morphology based on morphological algorithms. The model can take into account both the growth model of tree growth mechanism and the morphological model of plant appearance.

[0075] This embodiment preprocesses remote sensing image information collected by a UAV to obtain processed remote sensing tree images; it then segments the remote sensing tree images based on an improved Mask-R-CNN model to obtain segmented tree image information; based on a preset tree species model and the tree image information, it identifies tree species types to obtain image sets corresponding to each tree species type; and based on a preset morphological model and the image sets of each tree species type, it predicts the growth trend of trees in the target area, and determines the perceived growth status of trees in the target area based on the prediction results. Because this embodiment segments the preprocessed remote sensing tree images using the improved Mask-R-CNN model and identifies tree species based on a preset tree species model, thereby further determining the growth status of each type of tree, compared to existing UAV tree identification methods which suffer from incomplete identification processes and low accuracy, leading to inaccurate data, this embodiment achieves accurate prediction of tree growth status while accurately identifying vegetation using a UAV, thus improving data accuracy.

[0076] Reference Figure 3 , Figure 3 This is a flowchart illustrating the second embodiment of the UAV tree identification method of the present invention, based on the above. Figure 2 The first embodiment shown presents a second embodiment of the UAV tree identification method of the present invention.

[0077] In this embodiment, step S20 includes:

[0078] Step S201: Based on the improved Mask-R-CNN model and the preset cross-entropy loss function, the remote sensing tree image is segmented and identified to obtain the contour parameters corresponding to the target tree crown.

[0079] It should be noted that the preset cross-entropy loss function can be a pre-set function used to optimize prediction accuracy at target boundaries. This cross-entropy loss function can be a loss function used for classification; the lower the cross-entropy, the more similar the probabilities. Cross-entropy can include information entropy and relative entropy. In the loss calculation of Mask-R-CNN, each RoI alignment layer output has a corresponding binary mask, and the mask loss is part of the overall network loss. To obtain the number of categories and image size, the mask branch encodes an output matrix of size Km² for each RoI, where K is the number of categories. Combining this with the sigmoid function applied to each individual pixel, the mask loss is defined as the average binary cross-entropy loss function.

[0080] It should be understood that due to differences in hydrothermal conditions, vegetation characteristics, and soil factors in different study areas, the models determined by UAV data also differ, resulting in low accuracy and poor stability in remote sensing monitoring and estimation of grassland vegetation. Furthermore, the development of image recognition software is problematic: some target information in grassland images acquired by UAVs is obscured, requiring researchers to observe it visually, significantly reducing work efficiency and limiting the acquisition of other information from the images. Therefore, the application of UAVs in vegetation resource monitoring is limited. To avoid these problems and improve the accuracy of boundary segmentation, a boundary-weighted loss function is added to a preset cross-entropy loss function to optimize the prediction accuracy at target boundaries.

[0081] In the specific implementation, the remote sensing tree image is segmented and identified based on the improved Mask-R-CNN model and the preset cross-entropy loss function to obtain the contour parameters corresponding to the target tree crown. The contour parameters include contour boundary starting point information and contour line information.

[0082] Step S202: Determine the centroid coordinates corresponding to the target tree crown according to the preset boundary tracking algorithm and the contour parameters.

[0083] It should be noted that the preset boundary tracking algorithm can be a pre-set algorithm for tracking contour edges and performing boundary segmentation. The algorithm can be to find a boundary point in the image, and then start from this boundary point to find the next boundary point according to a certain strategy, thereby tracking out the target boundary.

[0084] Understandably, a hollow boundary can be generated by using a boundary tracking algorithm based on the starting boundary point, and its color can be restored before grayscale to generate each solid surface contour. The centroid coordinates of the graphic corresponding to each solid surface contour can be determined according to a preset matrix formula.

[0085] Step S203: Determine the segmented tree image information based on the contour parameters and the centroid coordinates.

[0086] It should be noted that the segmented tree image information is determined based on the contour parameters and the centroid coordinates.

[0087] In this embodiment, step S30 further includes:

[0088] Step S301: Identify the tree crown type based on the preset tree species model, the outline parameters, and the centroid coordinates, and determine the tree species type and quantity based on the identification results.

[0089] It should be noted that different tree species have different crowns, and therefore different outlines. The segmented tree image information is determined by a preset tree species model, outline parameters, and centroid coordinates. The tree image information may include the tree species type, outline, and centroid coordinates of the tree in the image to be identified.

[0090] Understandably, the accuracy of tree counts largely determines the biomass assessment of the entire forest area. Therefore, it is necessary to first determine the tree species and their quantities to inform the subsequent biomass assessment of the entire forest area.

[0091] Step S302: Classify the remote sensing tree images according to the tree species type and quantity to obtain the image set corresponding to each tree species type.

[0092] It should be noted that remote sensing tree images can be accurately classified by tree species type and quantity, obtaining image sets corresponding to each tree species type, which facilitates the prediction of the later growth status of each type of tree species.

[0093] Understandably, each type of tree species corresponds to a set. Since different tree species have different growth parameters, in order to facilitate subsequent data analysis, remote sensing tree images are pre-classified according to tree species to obtain image sets corresponding to each tree species type.

[0094] This embodiment preprocesses remote sensing image information collected by UAVs to obtain processed remote sensing tree images; it segments and identifies the remote sensing tree images based on an improved Mask-R-CNN model and a preset cross-entropy loss function to obtain contour parameters corresponding to the target tree crown; it determines the centroid coordinates corresponding to the target tree crown according to a preset boundary tracking algorithm and the contour parameters; it determines the segmented tree image information based on the contour parameters and the centroid coordinates; it identifies the tree crown type based on a preset tree species model, the contour parameters, and the centroid coordinates, and determines the tree species type and quantity based on the identification results; it classifies the remote sensing tree images according to the tree species type and quantity to obtain image sets corresponding to each tree species type; it predicts the growth trend of trees in the target area based on a preset morphological model and the image sets of each tree species type, and determines the tree growth status perception corresponding to the target area based on the prediction results. Because this embodiment segments the pre-processed remote sensing tree images using an improved Mask-R-CNN model and identifies tree species based on a preset tree species model, thereby further determining the growth status of various types of trees, compared to the incomplete identification process and low accuracy of existing UAV tree identification methods, which lead to inaccurate data, this embodiment achieves accurate prediction of tree growth status while accurately identifying vegetation using UAVs, thus improving data accuracy.

[0095] Reference Figure 4 , Figure 4 This is a flowchart illustrating the third embodiment of the UAV tree identification method of the present invention, based on the above. Figure 2 The first embodiment shown presents a third embodiment of the UAV tree identification method of the present invention.

[0096] In this embodiment, step S40 further includes:

[0097] Step S401: Based on the preset morphological model, the morphological features of trees in the image set of each tree species are compared, and the growth age of trees in the target area is determined according to the comparison results.

[0098] It should be noted that the tree morphology will be different at different growth ages. Therefore, by comparing the morphological features of trees in the image set of each tree species using a preset morphological model, the growth age of trees in the target area can be determined.

[0099] Step S402: Based on the growth age, predict the growth trend of the trees in the target area and obtain the prediction results.

[0100] It should be noted that the growth cycle of trees gradually slows down after a certain period of time. Therefore, the growth trend of trees in the target area can be predicted by their growth age to obtain the prediction results.

[0101] Step S403: Determine the tree growth status perception corresponding to the target area based on the prediction results and the preset Markov model.

[0102] It should be noted that the preset Markov model can be a pre-set tree growth status perception model used to predict the tree growth status of the target area. The prediction of tree growth status perception can refer to the prediction of the probability of changes in tree growth status. For example, if the probability is greater than the preset probability value, the tree growth status perception is determined to be good; if the probability is lower than the preset probability value, the tree growth status perception is determined to be poor. By determining the tree growth status perception, the trees in the target area can be maintained accordingly to achieve a more effective environmental protection effect.

[0103] Understandably, a pre-defined Markov model can determine the tree growth state, starting from the prediction objective, and consider the decision-making needs to classify the state of the target building. It calculates initial probabilities and uses state probabilities from historical data analysis as initial probabilities; it then calculates state transition probabilities and uses these probabilities to perceive and predict the tree growth trend in the target area.

[0104] Furthermore, step S403 further includes: determining the growth rate of the target area within a preset period based on the prediction result and a preset Markov model; predicting the area occupied by trees in the target area based on the growth rate to obtain the predicted area; and determining the tree growth status perception based on the predicted area and a preset clustering algorithm.

[0105] It should be noted that the preset period can refer to a future time period, such as one year, five years, or ten years, without specific limitations; the growth rate can refer to the changes in tree height and area occupied within a certain time range. The area occupied by trees in the target area is predicted based on the growth rate, resulting in a predicted area. This predicted area, along with a preset clustering algorithm, allows for the perception of tree growth trends. The preset clustering algorithm can be one that pre-defines algorithms for clustering and analyzing tree growth trends among trees of the same type.

[0106] This embodiment preprocesses remote sensing image information collected by a UAV to obtain processed remote sensing tree images; segments the remote sensing tree images based on an improved Mask-R-CNN model to obtain segmented tree image information; identifies tree species based on a preset tree species model and the tree image information to obtain image sets corresponding to each tree species type; compares the morphological features of trees in the image sets of each tree species type based on a preset morphological model, and determines the growth age of trees in the target area based on the comparison results; predicts the growth trend of trees in the target area based on the growth age to obtain prediction results; and determines the perceived growth status of trees in the target area based on the prediction results and a preset Markov model. Because this embodiment segments the preprocessed remote sensing tree images using the improved Mask-R-CNN model and identifies tree species based on a preset tree species model, thereby further determining the growth status of each type of tree, compared to existing UAV tree identification methods which suffer from incomplete identification processes and low accuracy leading to inaccurate data, this embodiment achieves accurate prediction of tree growth status while accurately identifying vegetation using a UAV, thus improving data accuracy.

[0107] Furthermore, to achieve the above objectives, the present invention also proposes a drone tree recognition device, which includes a memory, a processor, and a drone tree recognition program stored in the memory and executable on the processor. The drone tree recognition program is configured to implement the drone tree recognition steps described above.

[0108] In addition, to achieve the above objectives, the present invention also proposes a storage medium storing a drone tree recognition program, which, when executed by a processor, implements the steps of the drone tree recognition method described above.

[0109] Reference Figure 5 , Figure 5 This is a structural block diagram of the first embodiment of the tree identification device for unmanned aerial vehicles (UAVs) of the present invention.

[0110] like Figure 5 As shown, the drone tree identification device proposed in this embodiment of the invention includes:

[0111] Image preprocessing module 10 is used to preprocess the remote sensing image information collected by the UAV to obtain processed remote sensing tree images;

[0112] Image segmentation module 20 is used to segment the remote sensing tree image based on the improved Mask-R-CNN model to obtain segmented tree image information;

[0113] Tree species identification module 30 is used to identify tree species types based on a preset tree species model and the tree image information, and obtain an image set corresponding to each tree species type;

[0114] The situation awareness module 40 is used to predict the growth trend of trees in the target area based on a preset morphological model and image sets of various tree species, and to determine the tree growth situation awareness corresponding to the target area based on the prediction results.

[0115] This embodiment preprocesses remote sensing image information collected by a UAV to obtain processed remote sensing tree images; it then segments the remote sensing tree images based on an improved Mask-R-CNN model to obtain segmented tree image information; based on a preset tree species model and the tree image information, it identifies tree species types to obtain image sets corresponding to each tree species type; and based on a preset morphological model and the image sets of each tree species type, it predicts the growth trend of trees in the target area, and determines the perceived growth status of trees in the target area based on the prediction results. Because this embodiment segments the preprocessed remote sensing tree images using the improved Mask-R-CNN model and identifies tree species based on a preset tree species model, thereby further determining the growth status of each type of tree, compared to existing UAV tree identification methods which suffer from incomplete identification processes and low accuracy, leading to inaccurate data, this embodiment achieves accurate prediction of tree growth status while accurately identifying vegetation using a UAV, thus improving data accuracy.

[0116] Furthermore, the image preprocessing module 10 is also used to segment the remote sensing image information collected by the UAV based on the SLIC algorithm to obtain a segmented remote sensing image set; extract color features and texture features from the remote sensing image set according to the SLIC algorithm; and determine the remote sensing tree images in the remote sensing image set according to the color features and the texture features.

[0117] Furthermore, the image segmentation module 20 is also used to segment and identify the remote sensing tree image based on the improved Mask-R-CNN model and the preset cross-entropy loss function to obtain the contour parameters corresponding to the target tree crown; determine the centroid coordinates corresponding to the target tree crown according to the preset boundary tracking algorithm and the contour parameters; and determine the segmented tree image information according to the contour parameters and the centroid coordinates.

[0118] Furthermore, the tree species identification module 30 is also used to identify the tree crown type based on the preset tree species model, the contour parameters and the centroid coordinates, and determine the tree species type and quantity according to the identification results; classify the remote sensing tree images according to the tree species type and quantity, and obtain the image set corresponding to each tree species type.

[0119] Furthermore, the situation awareness module 40 is also used to perform feature comparison of tree morphological features in the image set of each tree species based on a preset morphological model, determine the growth age of trees in the target area based on the comparison results; predict the growth trend of trees in the target area based on the growth age to obtain the prediction result; and determine the tree growth situation awareness corresponding to the target area based on the prediction result and the preset Markov model.

[0120] Furthermore, the situation awareness module 40 is also used to determine the growth rate of the target area within a preset period based on the prediction result and a preset Markov model; to predict the area occupied by trees in the target area based on the growth rate to obtain the predicted area; and to determine the tree growth situation awareness based on the predicted area and a preset clustering algorithm.

[0121] Furthermore, the UAV tree recognition device also includes a model training module, which is used to acquire growth sample data corresponding to various types of tree species; input the growth sample data into an initial Mask-R-CNN model for training, and annotate the training results according to a preset annotation tool to obtain an annotated sample dataset; iteratively train the initial Mask-R-CNN model based on the annotated sample dataset until the output training results meet preset conditions, and use the trained Mask-R-CNN model as an improved Mask-R-CNN model.

[0122] It should be understood that the above are merely illustrative examples and do not constitute any limitation on the technical solutions of the present invention. In specific applications, those skilled in the art can make settings as needed, and the present invention does not impose any restrictions on this.

[0123] It should be noted that the workflow described above is merely illustrative and does not limit the scope of protection of this invention. In practical applications, those skilled in the art can select some or all of the workflow to achieve the purpose of this embodiment according to actual needs, and no restrictions are imposed here.

[0124] In addition, for technical details not described in detail in this embodiment, please refer to the UAV tree identification method provided in any embodiment of the present invention, which will not be repeated here.

[0125] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or system that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or system. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or system that includes that element.

[0126] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments. In the unit claims listing several devices, several of these devices may be embodied by the same hardware item. The use of the terms first, second, and third, etc., does not indicate any order and can be interpreted as names.

[0127] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as a read-only memory image (ROM) / random access memory (RAM), magnetic disk, optical disk), and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0128] The above are merely preferred embodiments of the present invention and do not limit the scope of the patent. Any equivalent structural or procedural transformations made based on the description and drawings of the present invention, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of the present invention.

Claims

1. A method for identifying a tree by a UAV, characterized in that, The unmanned aerial vehicle tree identification method comprises the following steps: The remote sensing image information collected by the unmanned aerial vehicle is preprocessed to obtain a processed remote sensing tree image; The remote sensing tree image is segmented based on the improved Mask-R-CNN model to obtain segmented tree image information; The tree species type is identified based on a preset tree species model and the tree image information to obtain an image set corresponding to each tree species type; The growth trend of the trees in the target region is predicted based on a preset morphological model and the image set of each tree species type, and the tree growth situation awareness corresponding to the target region is determined according to the prediction result. The step of segmenting the remote sensing tree image based on the improved Mask-R-CNN model to obtain segmented tree image information comprises: The remote sensing tree image is segmented and identified based on the improved Mask-R-CNN model and a preset cross-entropy loss function to obtain the contour parameters corresponding to the target tree crown; The center of gravity coordinates corresponding to the target tree crown are determined according to a preset boundary tracking algorithm and the contour parameters; The segmented tree image information is determined according to the contour parameters and the center of gravity coordinates. The step of identifying the tree species type based on the preset tree species model and the tree image information to obtain an image set corresponding to each tree species type comprises: The tree crown type is identified based on the preset tree species model, the contour parameters and the center of gravity coordinates, and the tree species type and the number are determined according to the identification result; The remote sensing tree image is classified according to the tree species type and the number to obtain an image set corresponding to each tree species type. 2.The unmanned aerial vehicle tree identification method of claim 1, wherein, The step of preprocessing the remote sensing image information collected by the unmanned aerial vehicle to obtain a processed remote sensing tree image comprises: The remote sensing image information collected by the unmanned aerial vehicle is segmented based on the SLIC algorithm to obtain a segmented remote sensing image set; The color features and the texture features are extracted from the remote sensing image set according to the SLIC algorithm, and the remote sensing tree image in the remote sensing image set is determined according to the color features and the texture features. 3.The unmanned aerial vehicle tree identification method of claim 1, wherein, The step of predicting the growth trend of the trees in the target region based on the preset morphological model and the image set of each tree species type, and determining the tree growth situation awareness corresponding to the target region according to the prediction result comprises: The growth age of the trees in the target region is determined according to the feature comparison of the tree morphological features in the image set of each tree species type based on the preset morphological model; The growth trend of the trees in the target region is predicted according to the growth age to obtain a prediction result; The tree growth situation awareness corresponding to the target region is determined according to the prediction result and a preset Markov model. 4.The unmanned aerial vehicle tree identification method of claim 3, wherein, The step of determining the tree growth situation awareness corresponding to the target region according to the prediction result and a preset Markov model comprises: The growth speed corresponding to the target region in a preset period is determined according to the prediction result and a preset Markov model; The area occupied by the trees in the target region is predicted according to the growth speed to obtain a predicted area. Determine tree growth trend awareness according to the predicted area and a preset clustering algorithm. 5.The unmanned aerial vehicle tree identification method of claim 1, wherein, Before the step of preprocessing the remote sensing image information collected by the unmanned aerial vehicle to obtain the processed remote sensing tree image, the method further comprises: Obtaining growth sample data corresponding to each type of tree species; Inputting the growth sample data into an initial Mask-R-CNN model for training, and labeling the training result according to a preset labeling tool to obtain a labeled sample data set; Iteratively training the initial Mask-R-CNN model based on the labeled sample data set until the training result output satisfies a preset condition, and taking the trained Mask-R-CNN model as an improved Mask-R-CNN model. 6.A drone tree recognition device, characterized in that, The unmanned aerial vehicle tree identification device comprises a memory, a processor, and an unmanned aerial vehicle tree identification program stored on the memory and executable on the processor, and the unmanned aerial vehicle tree identification program implements the unmanned aerial vehicle tree identification method of any one of claims 1 to 5 when executed by the processor.

7. A storage medium, characterized by The storage medium stores an unmanned aerial vehicle tree identification program, and the unmanned aerial vehicle tree identification program implements the unmanned aerial vehicle tree identification method of any one of claims 1 to 5 when executed by the processor.

8. A drone tree recognition device, comprising: The unmanned aerial vehicle tree identification device comprises: An image preprocessing module configured to preprocess remote sensing image information collected by the unmanned aerial vehicle to obtain a processed remote sensing tree image; An image segmentation module configured to segment the remote sensing tree image based on the improved Mask-R-CNN model to obtain segmented tree image information; A tree species identification module configured to identify tree species types based on a preset tree species model and the tree image information to obtain an image set corresponding to each tree species type; A trend awareness module configured to predict the growth trend of trees in a target region based on a preset morphological model and the image set of each tree species type, and determine the tree growth trend awareness corresponding to the target region according to the prediction result. The step of segmenting the remote sensing tree image based on the improved Mask-R-CNN model to obtain segmented tree image information comprises: Segmenting and identifying the remote sensing tree image based on the improved Mask-R-CNN model and a preset cross-entropy loss function to obtain contour parameters corresponding to a target tree crown; Determining the barycentric coordinates corresponding to the target tree crown according to a preset boundary tracking algorithm and the contour parameters; Determining the segmented tree image information according to the contour parameters and the barycentric coordinates; The step of identifying tree species types based on a preset tree species model and the tree image information to obtain an image set corresponding to each tree species type comprises: Identifying tree crown types based on a preset tree species model, the contour parameters, and the barycentric coordinates, and determining tree species types and quantities according to the identification result; Classifying the remote sensing tree image according to the tree species types and quantities to obtain an image set corresponding to each tree species type.

Images