Image recognition system and method for real-time monitoring of forest tending operations

Abstract

This invention relates to the field of real-time monitoring technology for forest tending operations, specifically to an image recognition system and method for real-time monitoring of forest tending operations. The recognition system includes: a communication module, a positioning module, an image acquisition module, an image processing module, a real-time monitoring module, and a storage management module. An intelligent monitoring method for forest tending operations includes: real-time operation location positioning; real-time acquisition of images of the forest tending operation site; identification of target objects in images of the same location using a pre-trained image recognition model; acquisition of relevant parameters of the target objects; and obtaining a judgment result. This invention can judge the types of target objects such as trees, weeds, soil, and fertilizer, greatly improving the efficiency and accuracy of forest tending operations; and effectively solving problems such as the inability to ensure that forest tending operations are carried out according to specifications, the inability to re-inspect the quality of planting pit fertilization tending operations, and the low comprehensiveness of sampling inspections.

Description

Technical Field

[0001] This invention relates to the field of real-time monitoring technology for forest tending operations, and specifically to an image recognition system and method for real-time monitoring of forest tending operations. Background Technology

[0002] Forest tending includes weeding, loosening the soil, fertilizing, and pest and disease control. Currently, forest tending is mainly done manually. Due to the large area of ​​forest work, on-site workers often work alone, making it impossible to monitor in real time whether the work process is standardized and whether the quality of work is guaranteed. During the acceptance inspection, due to the large area and the difficulty of climbing mountains again, the inspectors often neglect to check, creating significant loopholes in the supervision of the work process and the quality of tending. It is impossible to track the tending of each tree, ultimately resulting in low overall forest quality and an inability to fulfill its due ecological functions.

[0003] The aforementioned manual forest tending operations, which cannot be monitored in real time, will affect the overall quality of forests and consequently impact green and sustainable development. Therefore, there is an urgent need to provide a system or method capable of monitoring the manual forest tending operations in real time, accurately tracking the tending status of each tree, and thereby increasing the overall forest stock. Summary of the Invention

[0004] This invention provides an image recognition system and method for real-time monitoring of forest tending operations, in order to solve the technical problems in the background art.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: An image recognition system for real-time monitoring of forest tending operations includes: The communication module is used to connect the various processing modules and enable communication coupling between the equipment and the central control console; The positioning module is used to effectively track the location of the trees being worked on. The image acquisition module is used to collect visual data of the forest tending operation site in real time through a camera; The image processing module is used to receive visual data output by the image acquisition module and location information transmitted by the positioning module, integrate and process the visual data of the same location, identify various target objects such as trees and / or weeds and / or soil and / or fertilizer in the visual data and extract corresponding feature parameters, judge the compliance status of various target objects based on the feature parameters, and obtain the judgment result. The real-time monitoring module is used to display the location of the target object and its related parameters, so as to monitor in real time whether the operation process of the forest tending personnel is standardized and display the judgment results; The storage management module is used to store the location, relevant parameters, and judgment results of the target object throughout the entire operation process.

[0006] Furthermore, the image processing module includes: a recognition unit, a determination unit, and an output unit; The recognition unit is used to identify target objects in forest tending operation site images at the same location using a pre-trained image recognition model, and output forest tending operation site images with the target objects annotated. The judgment unit is used to obtain the location and related parameter information of the target object from the storage management module, perform compliance judgment on the target object based on the specified parameters, and obtain the judgment result; The output unit is used to send the location of the target object, relevant parameters and judgment results to the real-time monitoring module for display, and to send the judgment results to the storage management module for storage.

[0007] Furthermore, the recognition unit is based on the YOLOv5 target detection and recognition algorithm, combined with multi-scale feature enhancement method detection, and uses a pre-trained standard model of trees and / or weeds and / or soil and / or fertilizer to recognize the forest tending operation site image. Based on the predicted score of the recognition, the target object recognition result is confirmed by comparing it with the target threshold, and the target object is labeled.

[0008] Furthermore, the YOLOv5 algorithm is used to introduce the CBAM attention mechanism, which consists of two parts: Channel attention module: Extracts channel importance through global max pooling and average pooling, and then generates weights through shared MLP; Spatial attention module: Max / average pooling is performed on the feature map after channel compression, and a spatial attention map is generated by combining it with 7×7 convolution; The channel attention module and the spatial attention module are connected sequentially, meaning that channel attention is performed first, followed by spatial attention; the specific operation is shown in the formula: , In the formula: F is the original input feature map; F' is the feature map after channel attention weighting; F'' is the channel attention weight (output weight of the channel attention module); F'' is the final feature map after channel + spatial dual attention enhancement.

[0009] Furthermore, in YOLOv5, the network structure mainly includes a backbone, feature fusion, and detection layers; several key residual modules of the backbone are then embedded with CBAM modules to enhance the ability of mid-to-low-level features to represent targets.

[0010] Furthermore, the BiFPN_Concat structure is adopted to optimize multi-scale feature fusion. The BiFPN module adopts the improved BiFPN_Concat structure, which replaces the traditional weighted sum feature fusion method with channel concatenation and performs channel compression through a subsequent 1×1 convolution. The concatenation operation fully preserves the information of features at each scale, which helps to improve the detection accuracy of the model on multi-scale targets, especially small targets.

[0011] Furthermore, to enhance the expressive power of the detection head, a Four-Head Detection module, or four-head detection structure, is constructed based on the traditional three-head detection structure. This structure includes four branches: a category prediction branch, a bounding box regression branch, a confidence branch, and an auxiliary task branch such as orientation angle prediction or IoU-aware prediction.

[0012] Furthermore, the determination unit is specifically used for: Obtain the location and related parameters of the target object from the storage management module; the target objects include: trees and / or weeds and / or soil and / or fertilizer; The relevant parameters of the same target object in different images are stitched together, and the stitched relevant parameters are sent to the output unit; The compliance of the target object is determined based on the parameters, and the determination result is obtained; The determination results include: tree type and / or tree age and / or tree condition and / or tree pest and disease type and / or weed type and / or weed condition and / or soil color and / or planting hole marking and / or fertilizer placement status.

[0013] A method for intelligent monitoring of forest tending operations includes: Real-time operation location positioning; Real-time acquisition of images of forest tending operations; Use a pre-trained image recognition model to identify target objects in images of forest tending operations at the same location; Obtain relevant parameters of the target object; Displays the location of the target object and its related parameters, performs compliance judgment on the target object based on the specified parameters, and obtains the judgment result.

[0014] Furthermore, a pre-trained image recognition model is used to identify target objects in images of forest tending operations, specifically: Based on the YOLOv5 target detection and recognition algorithm, combined with multi-scale feature enhancement method detection, a pre-trained standard model of trees and / or weeds and / or soil and / or fertilizer is used to identify images of forest tending operation sites. The identification results of target objects are confirmed by comparing the predicted scores with the target threshold and the target objects are labeled.

[0015] The beneficial effects of this invention compared to the prior art are as follows: This invention marks several preset target objects in real-time location-based images of forest tending operations. It can identify target object types such as trees, weeds, soil, and fertilizer, compare them with the system, display relevant parameters, and then judge the operation results. Simultaneously, it records the entire location trajectory and operation process and transmits it back to the backend in real time, facilitating supervision by supervisors. This system greatly improves the efficiency and accuracy of forest tending operations and effectively solves problems such as the inability to ensure that forest tending operations are carried out in accordance with regulations, the inability to re-inspect the quality of planting pit fertilization and tending operations, and the low comprehensiveness of sampling inspections. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the various modules of the identification system of the present invention; Figure 2 This is a flowchart illustrating the intelligent monitoring method of the present invention; Figure 3 Here is an example image of a successfully identified tree; Figure 4 This is a successful example image of standard operating procedure identification for fertilizer pits, fertilizer, and seedlings. Figures 5-7 This is an example image showing a recognition failure caused by an unqualified captured image. Detailed Implementation

[0017] The above description is a detailed explanation of this patent in conjunction with specific embodiments, and it should not be construed that the specific embodiments of this patent are limited to the above description. For those skilled in the art, any substitutions or modifications made to the described embodiments without departing from the concept of this patent should be considered within the scope of protection of this patent. In this specification, specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification and the features of different embodiments or examples. Although the embodiments of the present invention and their advantages have been described in detail, it should be understood that various changes, substitutions, and modifications can be made herein without departing from the scope of protection of the patent application.

[0018] like Figure 1As shown, an image recognition system for real-time monitoring of forest tending operations is disclosed. The applicant has produced a physical prototype of this system and conducted scene application trials in forest areas with good results. The following embodiment, using this physical product as an example, includes: a communication module 10, a positioning module 11, an image acquisition module 12, an image processing module 13, a real-time monitoring module 14, and a storage management module 15. During processing, a combination of a fertilizer backpack and a handheld fertilizer hopper is used. The fertilizer backpack is made of rainproof canvas and contains an inner bag specifically for fertilizer, a bag for AI analysis equipment, and a bag for lithium batteries. The fertilizer in the inner bag is transported through a flexible tube to the storage bin of the handheld fertilizer hopper, which then dispenses the fertilizer into the fertilizer pit. The relevant modules of this system are described below: The communication module 10 connects the various processing modules by setting a communication chip (in this embodiment, the communication chip used is Quectel EG800AK-CN) on the manual wearable work device, thereby realizing the communication coupling between the device and the central control console and ensuring the real-time transmission and feedback of various command information. The positioning module 11 is equipped with a positioning chip (in this embodiment, the positioning chip used is SIMCom sim68vb) on the manual operation device to effectively track the location of the trees being operated on, and transmits the location information of the trees being operated on to the image processing module 13 through the communication module 10. Image acquisition module 12: This module acquires images of the forest tending operation site in real time by setting a camera on the work equipment worn by the operator. In this embodiment, it includes a front camera and a bottom camera, which are respectively set on the front and bottom sides of the handheld fertilizer hopper. The front camera is used to take pictures of the site in front, and the bottom camera is used to take pictures of the area below.

[0019] Image processing module 13: This module performs feature extraction and recognition on the forest tending operation site images collected by image acquisition module 12, combined with fixed location information transmitted by positioning module 11, to identify target objects at the same location, organize all image information collected at the same location, identify target objects such as trees, weeds, soil, and fertilizer at that location, obtain relevant parameters, and determine the compliance of the target objects based on the parameters to obtain the determination result. After identifying the target object, the image processing module 13 retrieves the target object's location and corresponding related parameter information from the storage management module 15. These related parameters refer to the target object's normal state and normal parameters. The image processing module 13 transmits the target object's related parameters and the judgment result output based on the specified parameters for compliance assessment to the target object to the real-time monitoring module 14 for display via the communication module 10. Simultaneously, the image processing module 13 transmits the judgment result to the storage management module 15, which records the relevant judgment results to facilitate supervision of the afforestation workers' operations. In this embodiment, the judgment result includes, but is not limited to: tree type, tree age, tree condition, type of tree diseases and pests, weed type, weed condition, soil color, planting hole markings, and fertilizer placement status.

[0020] The real-time monitoring module 14 displays the location of the target object and its related parameters to monitor in real time whether the on-site forest tending operations are in place and standardized, and displays the judgment results.

[0021] Storage management module 15 stores and records the location, relevant parameters, and judgment results of the target object to achieve full-process tracking and management of the operation.

[0022] In this embodiment, the storage management module 15 can be understood as a background database system that stores relevant data of the target object and records the judgment results.

[0023] This invention marks preset target objects in real-time location-based images of forest tending operations, determines the type of the target objects, compares them with the system to display relevant parameters, and then judges the operation results. At the same time, it records the entire location trajectory and operation process and transmits it back to the background in real time, which is a system that facilitates supervision by supervisors. It greatly improves the efficiency and accuracy of forest tending operations and can also be used as a detection method for AI verification.

[0024] In this embodiment, the image processing module 13 further includes: a recognition unit, a determination unit, and an output unit. The recognition unit is used to identify target objects in forest tending operation site images at the same location using a pre-trained image recognition model, and outputs forest tending operation site images labeled with the target objects; In this embodiment, the image recognition model can use the YOLOv5 target detection and recognition algorithm, combined with multi-scale feature enhancement method detection, and use pre-trained standard models such as trees, weeds, soil, and fertilizer to identify images of forest tending operation sites. Based on the predicted score of the identification, the target threshold is compared to confirm the identification result of the target object, and the target object is labeled.

[0025] For example, the YOLOv5 algorithm is used to introduce the CBAM attention mechanism. CBAM consists of two parts: Channel Attention Module: Extracts channel importance through global max pooling and average pooling, and then generates weights through a shared MLP; Spatial Attention Module: Performs max / average pooling on the channel-compressed feature map and combines it with 7×7 convolution to generate a spatial attention map.

[0026] These two submodules are connected sequentially, meaning channel attention is performed first, followed by spatial attention. The specific operation is shown in the formula: , In the formula: for:; for: for In YOLOv5, the network structure mainly includes a Backbone, a Neck (feature fusion), and a Head (detection layer). This embodiment embeds a CBAM module after several key residual modules (CSP Bottleneck) in the backbone network to enhance the representation ability of mid-to-low-level features on the target. Hyperparameter settings: Channel compression ratio is set to 16, and spatially, a 7×7 convolutional kernel is used.

[0027] Furthermore, the BiFPN_Concat structure can be used to optimize multi-scale feature fusion. The BiFPN module employs an improved BiFPN_Concat structure. This structure replaces the traditional weighted summation feature fusion method with channel concatenation (Concat), and performs channel compression through a subsequent 1×1 convolution. The concatenation operation effectively preserves information from features at each scale, helping to improve the model's detection accuracy on multi-scale targets, especially small targets.

[0028] To enhance the expressive power of the detection head, this embodiment constructs a Four-Head Detection module based on the traditional three-head detection structure. This structure includes four branches: a class prediction branch, a bounding box regression branch, a confidence branch, and an auxiliary task branch (such as orientation angle prediction or IoU-aware prediction). The four-head structure allows the network to maintain detection accuracy while possessing stronger task scalability and versatility, making it particularly suitable for fine-grained target modeling in complex scenes. The data used comes partly from collected data and partly from published public datasets.

[0029] The judgment unit is used to obtain the location and related parameter information of the target object from the storage management module, perform compliance judgment on the target object based on the specified parameters, and obtain the judgment result; Specifically, the judgment unit determines the compliance of the target object based on the output of the identification unit, such as whether the tree type is correct, the tree age is appropriate, the weed type is correct, and the fertilizer is applied to the planting hole. The judgment unit needs to obtain the location of the target object and its related parameters from the storage management module 15, and then send the obtained relevant parameters of the target object to the output unit. The output unit then sends the relevant parameters of the target object to the real-time monitoring module 14 for display.

[0030] The image acquisition module 12 continues to acquire images of the nursery operation site. The recognition unit identifies the target object, and the judgment unit can obtain multiple images labeled with the target object. The labeled images are compared with the standard images corresponding to the target object to obtain the compliance judgment result of the nursery operation site.

[0031] Since the image processing module 13 needs to identify the target object and obtain relevant parameters of the target object for compliance determination, the image processing module 13 needs powerful image processing and computing capabilities. Therefore, in this embodiment, the image processing module is located on the backend server.

[0032] In another embodiment, the determination unit is specifically used for: Obtain the location and related parameters of the target object from the storage management module 15; The relevant parameters of the same target object in different images are stitched together, and the stitched relevant parameters are sent to the output unit; The compliance of the target object is determined based on the parameters, and the determination result is obtained.

[0033] Specifically, relevant parameters corresponding to the target object are displayed to regulatory personnel through image annotation to determine the compliance of the target object in on-site forest tending operations.

[0034] The judgment unit identifies whether the target object in each image is consistent with the standard library. If the predicted score is greater than the threshold, it is confirmed as "completed". Otherwise, "incomplete" is displayed in the real-time monitoring module 14. The recognition result is retained, and the recognition results of the same target object are stitched together and annotated images are returned to the real-time monitoring module 14 and the storage management module 15.

[0035] The output unit is used to send the location of the target object, relevant parameters and judgment results to the real-time monitoring module 14 for display, and to send the judgment results to the storage management module 15 for storage.

[0036] This invention can be widely used in scenarios such as garden planting, plant fertilization, and pesticide spraying of agricultural and forestry crops.

[0037] like Figure 2 The diagram shown is a flowchart illustrating the intelligent monitoring method for forest tending operations provided in an embodiment of the present invention. It employs an image recognition system for real-time monitoring of forest tending operations as described above, comprising: Step 21: Real-time operation location positioning; Specifically, the location information includes latitude and longitude coordinates.

[0038] Step 22: Real-time acquisition of images of forest tending operations; Specifically, images of forest tending operations include images of planting, fertilizing, spraying pesticides, and other on-site activities.

[0039] Step 23: Use a pre-trained image recognition model to identify target objects in images of forest tending operations at the same location; Specifically, a pre-trained image recognition model is used to extract and identify target features from images of forest tending operations at the same location, resulting in forest tending operation images labeled with the target objects. Examples of standard fertilization operation images include... Figure 3 As shown.

[0040] In some embodiments, based on the YOLOv5 target detection and recognition algorithm, combined with multi-scale feature enhancement methods, pre-trained standard models for trees, weeds, soil, and fertilizer are used to identify images of forest tending operations at the same location frame by frame. The predicted scores are compared with target thresholds to confirm the identification results of the target objects, and the target objects are then labeled. The target objects in forest tending operations can be divided into the following four categories: trees, weeds, soil, and fertilizer.

[0041] Step 24: Obtain the relevant parameters of the target object; Specifically, relevant parameters are retrieved based on the target object. These relevant parameters refer to the target object's normal status, normal parameters, or operational data.

[0042] Step 25: Display the location of the target object and its related parameters, perform a compliance judgment on the target object based on the specified parameters, and obtain the judgment result; Specifically, the location of the target object and its related parameters are displayed, and a similarity comparison is performed on the target object standard based on the specified parameters to obtain the judgment result.

[0043] Step 26: Display the judgment result; Specifically, the judgment result is displayed to indicate whether the task is completed or not.

[0044] For example, the identification and judgment results of fertilization operations are as follows: First, if we obtain such Figure 3As shown, this indicates successful tree recognition: the tree is displayed in the front camera view. Hint: Eucalyptus tree.

[0045] Then obtain such as Figure 4 As shown, this indicates successful operation status recognition: the bottom camera clearly shows fertilizer pits, fertilizer, and saplings. The message reads: Operation complete. This judgment is based on a combination of the images of trees shown in the front camera and the bottom camera. Figure 4 The result of the judgment.

[0046] If you obtain such Figure 5 As shown, this indicates recognition failure 1: the front camera did not capture the tree. The system will prompt: Please point the camera at the tree, and ensure the lens is clean and well-lit until the tree is clearly visible in the frame.

[0047] If you obtain such Figure 6 or Figure 7 As shown, this indicates recognition failure 2: The bottom camera needs to be able to clearly identify and display the geometric boundaries of the fertilizer pit and the distribution outline of the fertilizer. If the current image is blurry, reflective, or obstructed, the system will prompt: Please adjust the camera angle, clean the lens, and optimize the lighting conditions to ensure that the edge features of the fertilizer pit and fertilizer are distinguishable in the image.

[0048] The intelligent monitoring method for forest tending operations described in this application marks preset target objects in real-time location images of forest tending operations, determines the type of the target objects, compares them with the system, displays their relevant parameters, and then judges the operation results. At the same time, the entire location trajectory and operation process are recorded and transmitted back to the background in real time, making it a system that facilitates supervision by supervisors. This greatly improves the efficiency and accuracy of forest tending operations and can also be used as a detection method for AI verification.

[0049] The embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0050] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the contribution to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, optical disc, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

Claims

1. An image recognition system for real-time monitoring of forest tending operations, characterized in that, include: The communication module is used to connect the various processing modules and enable communication coupling between the equipment and the central control console; The positioning module is used to effectively track the location of the trees being worked on. The image acquisition module is used to collect visual data of the forest tending operation site in real time through a camera; The image processing module is used to receive visual data output by the image acquisition module and location information transmitted by the positioning module, integrate and process the visual data of the same location, identify various target objects such as trees and / or weeds and / or soil and / or fertilizer in the visual data and extract corresponding feature parameters, judge the compliance status of various target objects based on the feature parameters, and obtain the judgment result. The real-time monitoring module is used to display the location of the target object and its related parameters, so as to monitor in real time whether the operation process of the forest tending personnel is standardized and display the judgment result; The storage management module is used to store the location, relevant parameters, and judgment results of the target object throughout the entire operation process.

2. The image recognition system for real-time monitoring of forest tending operations according to claim 1, characterized in that: The image processing module includes: a recognition unit, a determination unit, and an output unit; The recognition unit is used to identify target objects in forest tending operation site images at the same location using a pre-trained image recognition model, and output forest tending operation site images with the target objects annotated. The judgment unit is used to obtain the location and related parameter information of the target object from the storage management module, perform compliance judgment on the target object based on the specified parameters, and obtain the judgment result; The output unit is used to send the location of the target object, relevant parameters and judgment results to the real-time monitoring module for display, and to send the judgment results to the storage management module for storage.

3. The image recognition system for real-time monitoring of forest tending operations according to claim 2, characterized in that: The recognition unit is based on the YOLOv5 target detection and recognition algorithm, combined with multi-scale feature enhancement method detection. It uses a pre-trained standard model of trees and / or weeds and / or soil and / or fertilizer to recognize the forest tending operation site image. Based on the predicted score of the recognition, it compares with the target threshold to confirm the recognition result of the target object and marks the target object.

4. The image recognition system for real-time monitoring of forest tending operations according to claim 3, characterized in that: The YOLOv5 algorithm is used to introduce the CBAM attention mechanism, which consists of two parts: Channel attention module: Extracts channel importance through global max pooling and average pooling, and then generates weights through shared MLP; Spatial attention module: Max / average pooling is performed on the feature map after channel compression, and a spatial attention map is generated by combining it with 7×7 convolution; The channel attention module and the spatial attention module are connected sequentially, meaning that channel attention is performed first, followed by spatial attention; the specific operation is shown in the formula: , In the formula: F is the original input feature map; F' is the feature map after channel attention weighting; F'' is the channel attention weight (output weight of the channel attention module); F'' is the final feature map after channel + spatial dual attention enhancement.

5. The image recognition system for real-time monitoring of forest tending operations according to claim 4, characterized in that: In YOLOv5, the network structure mainly includes a backbone, feature fusion, and detection layers; several key residual modules of the backbone are then embedded with CBAM modules to enhance the ability of mid-to-low-level features to represent targets.

6. The image recognition system for real-time monitoring of forest tending operations according to claim 4, characterized in that: The BiFPN_Concat structure is used to optimize multi-scale feature fusion. The BiFPN module adopts the improved BiFPN_Concat structure, which replaces the traditional weighted sum feature fusion method with channel concatenation and performs channel compression through a subsequent 1×1 convolution. The concatenation operation fully preserves the information of features at each scale, which helps to improve the detection accuracy of the model on multi-scale targets, especially small targets.

7. The image recognition system for real-time monitoring of forest tending operations according to claim 4, characterized in that: To enhance the expressive power of the detection head, a Four-Head Detection module, or four-head detection structure, is constructed based on the traditional three-head detection structure. This structure includes four branches: a category prediction branch, a bounding box regression branch, a confidence branch, and an auxiliary task branch such as orientation angle prediction or IoU-aware prediction.

8. The image recognition system for real-time monitoring of forest tending operations according to claim 2, characterized in that: The determination unit is specifically used for: Obtain the location and related parameters of the target object from the storage management module; the target objects include: trees and / or weeds and / or soil and / or fertilizer; The relevant parameters of the same target object in different images are stitched together, and the stitched relevant parameters are sent to the output unit; The compliance of the target object is determined based on the parameters, and the determination result is obtained; The determination results include: tree type and / or tree age and / or tree condition and / or tree pest and disease type and / or weed type and / or weed condition and / or soil color and / or planting hole marking and / or fertilizer placement status.

9. A method for intelligent monitoring of forest tending operations, characterized in that, include: Real-time operation location positioning; Real-time acquisition of images of forest tending operations; Use a pre-trained image recognition model to identify target objects in images of forest tending operations at the same location; Obtain relevant parameters of the target object; Displays the location of the target object and its related parameters, performs compliance judgment on the target object based on the specified parameters, and obtains the judgment result.

10. The intelligent monitoring method for forest tending operations according to claim 9, characterized in that, The target objects in forest tending operation site images are identified using a pre-trained image recognition model. Specifically, the target detection and recognition algorithm is based on YOLOv5, combined with multi-scale feature enhancement methods. A pre-trained standard model of trees and / or weeds and / or soil and / or fertilizer is used to identify the forest tending operation site images. The identification results of the target objects are confirmed by comparing the predicted scores with the target threshold, and the target objects are labeled.

Images