Poppy recognition method and device based on inverse thinking, equipment and storage medium

Abstract

The invention provides a poppy recognition method and device based on inverse thinking, equipment and a storage medium, and relates to the technical field of image recognition, and the method comprises the steps: obtaining a plant image set; analyzing and processing the plant image set by using the trained image analysis model, and identifying at least one possible poppy planting area and the planting probability of each possible poppy planting area; the image analysis model is constructed based on a body confrontation theory; the highly suspected area is a possible poppy planting area of which the planting probability is greater than a preset probability threshold; dividing each possible poppy planting area according to the planting probability of each possible poppy planting area to obtain a highly suspected area for planting poppy, and identifying poppy in the execution area based on the highly suspected area through reverse thinking. The method can effectively overcome the influence of shielding and other factors, improves the accuracy of poppy recognition, and can remarkably improve the detection efficiency of a large-area region.

Description

Technical Field

[0001] This invention relates to the field of image recognition technology, and in particular to a poppy identification method, apparatus, device, and storage medium based on reverse thinking. Background Technology

[0002] With technological advancements, using drones for plant identification and monitoring has become an important research direction. Timely identification and eradication of illegally planted poppies are of great significance. Traditional poppy identification relies primarily on manual patrols, which is costly in terms of manpower and resources and inefficient. With the development of remote sensing technology and computer vision, drone-based poppy identification methods are gradually becoming a research hotspot.

[0003] Currently, drone-based poppy identification methods mainly fall into the following categories: first, spectral feature-based identification methods, which identify poppies by analyzing their spectral reflectance characteristics; second, texture feature-based identification methods, which utilize the texture features of poppy leaves and flowers for identification; and third, deep learning-based identification methods, which employ algorithms such as convolutional neural networks to extract features and classify poppy images. These methods can achieve good identification results under ideal conditions.

[0004] However, several difficulties and challenges remain in practical applications: First, there's the issue of obstruction; growers often employ concealment methods, making it difficult for drones to directly photograph poppy plants. Second, environmental interference; the presence of numerous similar plants in the natural environment can easily lead to misidentification. Third, there are image quality issues; affected by factors such as lighting and weather, acquired images may suffer from blurriness, noise, and other quality problems. These factors severely impact the accuracy and practicality of existing poppy identification methods. Especially in complex terrain environments such as mountainous areas, growers often choose secluded locations and employ various obstruction methods, making methods based on direct poppy plant identification ineffective. Due to these issues, existing drone-based poppy identification methods have low accuracy in complex terrain environments. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a poppy identification method, apparatus, device, and storage medium based on reverse thinking, which effectively overcomes the influence of factors such as obstruction, improves the accuracy of poppy identification, and significantly enhances the detection efficiency over large areas.

[0006] In a first aspect, the present invention provides a poppy identification method based on reverse thinking, the method comprising the following steps: The drone was used to acquire multi-channel images of the target area, resulting in a collection of plant images. The trained image analysis model is used to analyze and process the plant image set to identify at least one possible poppy planting area and the planting probability of each possible poppy planting area; wherein, the trained image analysis model is constructed based on the embodied adversarial theory; Based on the planting probability of each of the possible poppy planting areas, the possible poppy planting areas are divided to obtain highly suspected poppy planting areas; the highly suspected areas are possible poppy planting areas where the planting probability is greater than a preset probability threshold. By reversing the thinking process, poppies in the execution area are identified based on the highly suspicious area.

[0007] According to the present invention, a poppy identification method based on reverse thinking is provided. The trained image analysis model includes an environmental adaptability assessment module, a metacognitive identity control module, a perspective switching module, and a strategy motivation modeling module. The step of using the trained image analysis model to analyze and process the plant image set to identify at least one possible poppy planting area and the planting probability of each possible poppy planting area includes: For any plant image in the plant image set, the environmental features of the plant image are extracted by the environmental adaptability assessment module to obtain the environmental features and environmental fitness weights of the region where the plant image is located. Based on the environmental features corresponding to the region where the plant image is located and the image features of the plant image, the cross-attention mechanism in the metacognitive identity control module is used to verify the identity consistency of the plant in the plant image with that of the poppy, and obtain the identity consistency confidence of the plant in the plant image with the poppy. The perspective conversion module determines the perspective-converted feature map of the plant image based on the environmental features corresponding to the region where the plant image is located; the perspective-converted feature map represents the spatial distribution of the suspected region. The strategy motivation modeling module performs strategy modeling based on the plant image to obtain the strategy motivation weight corresponding to the plant image; the strategy motivation weight represents the risk coefficient corresponding to the potential adversarial behavior in the plant image. Based on the environmental features corresponding to the region where the plant image is located, the confidence level of the consistency between the plant and the poppy in the plant image, the feature map after the viewpoint transformation of the plant image, and the strategy motivation weight corresponding to the plant image, feature fusion and probability calculation are performed to obtain each possible poppy planting area and the planting probability of each possible poppy planting area.

[0008] According to the poppy identification method based on reverse thinking provided by the present invention, before analyzing and processing the plant image set using a trained image analysis model to identify at least one possible poppy planting area and the planting probability of each possible poppy planting area, the method includes: The plant image set is preprocessed to obtain a target image set; the preprocessing operation includes at least one of the following: contrast increase, noise suppression, sharpening, and color enhancement. The step of analyzing and processing the plant image set using a trained image analysis model to identify at least one possible poppy planting area and the planting probability of each possible poppy planting area includes: The trained image analysis model is used to analyze and process the target image set to identify each possible poppy planting area and the planting probability of each possible poppy planting area.

[0009] According to a poppy identification method based on reverse thinking provided by the present invention, the step of dividing each possible poppy planting area according to the planting probability of each possible poppy planting area to obtain a highly suspicious area for poppy planting includes: For any of the poppy-growing potential areas, the planting probability of the poppy-growing potential area is compared with a preset probability threshold; If the probability of poppy cultivation in the potential poppy-growing area is greater than or equal to the preset probability threshold, the potential poppy-growing area is identified as the highly suspicious area.

[0010] According to the poppy identification method based on reverse thinking provided by the present invention, the training steps of the trained image analysis model include: The initial image analysis model is trained using the Embodied Loss composite loss function and the Adam optimizer through a three-stage progressive training strategy to obtain the trained image analysis model. The three-stage progressive training strategy includes, in sequence, a basic detector pre-training stage, a modular training stage, and an end-to-end fine-tuning stage.

[0011] According to the present invention, a poppy identification method based on reverse thinking is provided, the method further includes: Based on the highly suspicious area, a clearance order is triggered; In response to the clearance order, the target personnel for the clearance of the highly suspicious area are obtained by matching the status information of multiple law enforcement personnel; the status information includes: real-time location, workload, and professional capabilities. The coordinate information of the highly suspicious area is sent to the terminal device corresponding to the target investigation personnel.

[0012] Secondly, the present invention also provides a poppy identification device based on reverse thinking, the device comprising the following modules: The image acquisition module is used to acquire multi-channel images of the execution area using a drone to obtain a set of plant images; The identification module is used to analyze and process the plant image set using a trained image analysis model to identify at least one possible poppy planting area and the planting probability of each possible poppy planting area; wherein, the trained image analysis model is constructed based on the embodied adversarial theory. Based on the planting probability of each of the possible poppy planting areas, the possible poppy planting areas are divided to obtain highly suspected poppy planting areas; the highly suspected areas are possible poppy planting areas where the planting probability is greater than a preset probability threshold. By reversing the thinking process, poppies in the execution area are identified based on the highly suspicious area.

[0013] Thirdly, the present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the poppy identification method based on reverse thinking as described above.

[0014] Fourthly, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the poppy identification method based on reverse thinking as described above.

[0015] Fifthly, the present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the poppy identification method based on reverse thinking as described above.

[0016] The present invention provides a poppy identification method, apparatus, device, and storage medium based on reverse thinking. First, a drone is used to acquire multi-channel images of the execution area, obtaining a set of plant images. Then, a trained image analysis model is used to analyze and process the plant image set to identify at least one possible poppy planting area and the planting probability of each possible poppy planting area. The trained image analysis model is constructed based on embodied adversarial theory. Furthermore, based on the planting probability of each possible poppy planting area, the possible poppy planting areas are divided into highly suspicious areas for poppy cultivation. Highly suspicious areas are those with a planting probability greater than a preset probability threshold. Through reverse thinking, poppies in the execution area are identified based on these highly suspicious areas.

[0017] This invention constructs a trained image analysis model based on the theory of embodied adversarial behavior. The trained image analysis model is used to analyze and process the set of plant images to identify at least one possible poppy planting area and the planting probability of each possible poppy planting area. This allows for the determination of highly suspicious areas for poppy planting. By analyzing the growth environment characteristics of poppies, this invention identifies and classifies possible poppy planting areas. This method can effectively overcome the influence of factors such as occlusion, improve the accuracy of poppy identification, and significantly improve the detection efficiency of large areas. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0019] Figure 1 This is a flowchart illustrating the poppy identification method based on reverse thinking provided by the present invention.

[0020] Figure 2 This is a schematic diagram illustrating the principle of the embodied adversarial theory provided by this invention.

[0021] Figure 3 This is a schematic diagram illustrating the principle of the poppy identification method based on reverse thinking provided by the present invention.

[0022] Figure 4 This is a schematic diagram of the structure of the poppy identification device based on reverse thinking provided by the present invention.

[0023] Figure 5 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0024] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0025] To more clearly understand the various embodiments provided by the present invention, the technical content involved in the present invention will first be described as follows: Existing drone-based poppy identification methods mostly focus on improving the accuracy of identifying the poppy plants themselves, but research on identifying poppies in obscured areas is relatively limited. Therefore, developing new methods that can effectively identify poppies in obscured areas is of significant theoretical and practical value, as it breaks through traditional thinking.

[0026] This invention transforms the traditional approach of direct identification by establishing a new image analysis model based on the characteristics of the poppy's growing environment, providing a new solution for improving the accuracy and efficiency of poppy identification.

[0027] The following is combined with Figures 1-5 The present invention describes a poppy identification method, apparatus, device, and storage medium based on reverse thinking.

[0028] Figure 1 This is a flowchart illustrating the poppy identification method based on reverse thinking provided by the present invention, as shown below. Figure 1 As shown, the method includes the following: Step 101: Use a drone to collect multi-channel images of the execution area to obtain a set of plant images.

[0029] Specifically, it should be noted that the execution subject of this embodiment of the invention is an electronic device, used to realize poppy identification based on reverse thinking, and to identify and classify possible poppy planting areas by analyzing the characteristics of the poppy's growth environment.

[0030] First, a set of original plant images is acquired. The multi-channel positioning and imaging module of the drone is used to acquire multi-channel images of the execution area; that is, the plants in the execution area are photographed, forming a set of plant images.

[0031] The acquired image types include at least one of the following: red-green-blue (RGB) images, multispectral images, thermal images, and terahertz imaging. Utilizing the penetrating power of terahertz waves can help determine the distribution of poppies under vegetation cover, creating more possibilities for improving monitoring accuracy and environmental adaptability in complex scenarios.

[0032] More specifically, during detection, if the area to be detected is large, it can be divided into several medium-sized areas. For each area, drones are used for patrols to collect images of suspicious areas using the drones' fast and efficient features. Then, poppy cultivation areas are identified and classified by region.

[0033] Step 102: Analyze and process the plant image set using the trained image analysis model to identify at least one possible poppy planting area and the planting probability of each possible poppy planting area; wherein, the trained image analysis model is constructed based on embodied adversarial theory.

[0034] Specifically, after obtaining the set of plant images, the plant images are analyzed and processed by the trained image analysis model to identify possible planting areas and the planting probability of each possible poppy planting area.

[0035] The trained image analysis model is constructed based on embodied adversarial theory. Figure 2 This is a schematic diagram illustrating the principle of the embodied adversarial theory provided by the present invention, such as... Figure 2 As shown in the figure, this diagram intuitively illustrates the adversarial dynamic process that exists in the complex identification scenarios targeted by this invention (such as the investigation of concealed poppy cultivation).

[0036] As shown on the left side of the attached diagram, under normal conditions, the detection system embodies the identity of the subject. When interacting normally with the environment (e.g., a natural scene containing various plants), it can stably perceive targets (such as poppies) based on a preset recognition model. This process is represented by the yellow arrow in the diagram, symbolizing the system's smooth detection and recognition flow.

[0037] As shown on the right side of the attached diagram, in an adversarial state, an adversary identity is introduced into the environment, intending to evade detection. This adversarial identity alters or disrupts the original state of the environment by actively employing interference strategies (e.g., camouflaging, intercropping, or selectively planting poppies in obscured areas), resulting in a significant decrease in the recognition efficiency of the subject's identity based on normal interaction patterns. This interference process is represented by the purple arrow in the diagram, symbolizing the conflict and disruption caused by the adversarial behavior to the subject's identity cognition.

[0038] The core of the solution described in this invention lies in enabling the image analysis model to dynamically switch perspectives between two identity states. Specifically, the image analysis model in this application simulates and internalizes this adversarial principle, particularly the Perspective Inversion Module (PIM) and the Strategy Motivation Modeling Module (SMMM). The model not only identifies target features from the perspective of the subject (detector), but more importantly, it can switch to the perspective of the adversary (hider). This dual cognition is maintained through the Meta-cognitive Identity Control Module (MICM), and the Environmental Suitability Assessment Module (ESAM) analyzes interfering factors in the environment to predict potential hiding strategies. Ultimately, through this "reverse thinking" training based on embodied adversarial theory, the model can more accurately identify potentially hidden poppy planting areas that are deliberately concealed in adversarial states, thereby significantly improving the accuracy and robustness of detection in real-world complex scenarios.

[0039] The trained image analysis model in this invention mainly comprises the four modules mentioned above: Environmental Adaptability Assessment Module (ESAM), Perspective Transition Module (PIM), Metacognitive Identity Control Module (MICM), and Strategy Motivation Modeling Module (SMMM). Among them, the Environmental Adaptability Assessment Module (ESAM) focuses on extracting and analyzing the environmental features of the area captured by the drone. It uses depthwise separable convolution technology to evaluate features such as terrain, lighting, and vegetation cover to determine the "suitability" of a specific environment for poppy growth and its concealment, thus providing crucial environmental context information for identifying potential planting areas. The core function of the Perspective Transition Module (PIM) is to realize the system's transformation from the observer's perspective to the hider's perspective. Through an identity consistency module and consistency gating mechanism based on a Transformer encoder, it can simulate the thought process of potential illegal growers choosing concealed locations, thereby inversely inferring possible poppy planting areas that are easily obscured or disguised. The Metacognitive Identity Control Module (MICM) aims to solve the conflict problem of target identity recognition in complex scenarios. This module uses a cross-attention-based identity conflict detector to compare plant images captured by drones with a prior poppy feature library (such as flower morphology and leaf texture) in real time. It also utilizes a multi-head self-attention mechanism in the metacognitive layer to dynamically analyze recognition biases caused by environmental interference (such as similar plants like corn poppies or occlusions), ensuring the system maintains task consistency during perspective transitions. The Strategy Motivation Modeling Module (SMMM) is responsible for predicting and modeling adversarial strategies. Based on a CNN environmental encoder and a multilayer perceptron, it analyzes environmental features and historical data to simulate the decision-making process of potential adversaries (i.e., illegal growers), understanding and predicting their possible adversarial strategies such as camouflage and intercropping. This provides dynamic risk assessment weights for the recognition model to adjust the final planting probability.

[0040] Step 103: Divide each of the possible poppy planting areas according to the planting probability of each of the possible poppy planting areas to obtain highly suspected poppy planting areas; the highly suspected areas are poppy planting areas with a planting probability greater than a preset probability threshold. By reversing the thinking process, poppies in the execution area are identified based on the highly suspicious area.

[0041] Specifically, after identifying the planting probability of each possible poppy planting area through the trained image analysis model, the possible poppy planting areas are further divided based on the planting probability to obtain highly suspicious areas.

[0042] For example, if 10 potential poppy-growing areas are identified, the probability of poppy cultivation in these 10 areas is as follows: Region 1: 10%; Region 2: 20%; Region 3: 30%; Region 4: 50%; Region 5: 60%; Region 6: 70%; Region 7: 80%; Region 8: 90%; Region 9: 55%; Region 10: 68%.

[0043] Areas with a planting probability greater than 70% (Areas 6, 7, and 8) were designated as highly suspicious areas. These highly suspicious areas are poppy cultivation areas that require immediate investigation. This invention, by shifting from the traditional direct identification approach to establishing a new image analysis model based on the characteristics of the poppy's growth environment, provides a new solution for improving the accuracy and efficiency of poppy identification.

[0044] Step 104: Using reverse thinking, identify the poppies in the execution area based on the highly suspicious area.

[0045] The direct poppy identification process is transformed into first identifying highly suspicious areas where the planting probability is higher than a preset probability threshold, and then identifying the poppies in those highly suspicious areas.

[0046] The method provided in this embodiment first acquires a set of plant images by using a drone to collect multi-channel images of the execution area. Then, it analyzes and processes the plant image set using a trained image analysis model to identify at least one possible poppy planting area and the planting probability of each possible poppy planting area. The trained image analysis model is constructed based on embodied adversarial theory and includes an environmental adaptability assessment module, a metacognitive identity control module, a perspective switching module, and a strategy motivation modeling module. Furthermore, the possible poppy planting areas are divided according to their planting probabilities to obtain highly suspicious areas for poppy planting. Highly suspicious areas are those with a planting probability greater than a preset probability threshold. By using reverse thinking, the poppies in the execution area are identified based on the highly suspicious areas.

[0047] This invention constructs a trained image analysis model based on the theory of embodied adversarial behavior. The trained image analysis model is used to analyze and process the set of plant images to identify at least one possible poppy planting area and the planting probability of each possible poppy planting area. This allows for the determination of highly suspicious areas for poppy planting. By analyzing the growth environment characteristics of poppies, this invention identifies and classifies possible poppy planting areas. This method can effectively overcome the influence of factors such as occlusion, improve the accuracy of poppy identification, and significantly improve the detection efficiency of large areas.

[0048] According to the present invention, a poppy identification method based on reverse thinking is provided, wherein the method utilizes a trained image analysis model to analyze and process the plant image set to identify at least one possible poppy planting area and the planting probability of each possible poppy planting area, including: For any plant image in the plant image set, the environmental features of the plant image are extracted by the environmental adaptability assessment module to obtain the environmental features and environmental fitness weights of the region where the plant image is located. Based on the environmental features corresponding to the region where the plant image is located and the image features of the plant image, the cross-attention mechanism in the metacognitive identity control module is used to verify the identity consistency of the plant in the plant image with that of the poppy, and obtain the identity consistency confidence of the plant in the plant image with the poppy. The perspective conversion module determines the perspective-converted feature map of the plant image based on the environmental features corresponding to the region where the plant image is located; the perspective-converted feature map represents the spatial distribution of the suspected region. The strategy motivation modeling module performs strategy modeling based on the plant image to obtain the strategy motivation weight corresponding to the plant image; the strategy motivation weight represents the risk coefficient corresponding to the potential adversarial behavior in the plant image. Based on the environmental features corresponding to the region where the plant image is located, the confidence level of the consistency between the plant and the poppy in the plant image, the feature map after the viewpoint transformation of the plant image, and the strategy motivation weight corresponding to the plant image, feature fusion and probability calculation are performed to obtain each possible poppy planting area and the planting probability of each possible poppy planting area.

[0049] Specifically, in some embodiments, the identification of potential poppy planting areas and planting probabilities in step 102 is achieved through an Environmental Adaptability Assessment Module (ESAM), a Metacognitive Identity Control Module (MICM), a Perspective Shift Module (PIM), and a Strategy Motivation Modeling Module (SMMM). The specific implementation of each module is as follows: (1) Environmental Adaptability Assessment Module (ESAM) performs environmental feature extraction: For any of the plant images in the plant image set, the Environmental Adaptability Assessment (ESAM) module extracts environmental features from the plant image, including the environmental features and environmental fitness weights corresponding to the region where the plant image is located. The environmental fitness weights, for example, are scores from 0 to 1, used to assess the "fitness" of a specific environment for poppy growth and concealment. Higher weight values ​​indicate that the environmental features of the region are more suitable for poppy growth.

[0050] Specifically, the Environmental Adaptability Assessment (ESAM) module performs preliminary processing on the input plant images. The ESAM module uses depthwise separable convolution technology. This module extracts and analyzes environmental or terrain features such as terrain, illumination, and vegetation cover in the image through a network structure containing components such as convolutional layers (Conv2d) and linear layers (Linear). It calculates the environmental fitness weights through multi-scale perception (such as spatial pyramid operations).

[0051] (2) The Metacognitive Identity Control Module (MICM) performs identity conflict detection: The MICM module receives environmental feature output from ESAM and combines it with image data to enhance detection accuracy. Based on a cross-attention mechanism, the MICM module compares the input image with a prior poppy feature library to detect interference from non-target plants (such as corn poppies). A multi-head self-attention layer maintains task awareness and outputs identity consistency confidence, ensuring that the identified target remains within acceptable limits.

[0052] Specifically, the system receives environmental features corresponding to the region where the plant image is located and image features of the plant image. It then uses the cross-attention mechanism in the Metacognitive Identity Control Module (MICM) to verify the identity consistency of the plant in the image, outputting the confidence score of the identity consistency between the plant in the image and the poppy. This module is used to filter false detection regions. The MICM includes structures such as a self-attention mechanism and LayerNorm to perform identity consistency detection, ensuring that the system maintains its core mission objective without drifting in complex adversarial scenarios and effectively distinguishes target plants from non-target interference.

[0053] (3) The Perspective Transition Module (PIM) realizes perspective transition and feature fusion: The Perspective Switching Module (PIM) relies on the Environmental Adaptability Assessment Module (ESAM) to perform perspective switching based on environmental characteristics.

[0054] Specifically, the input to the perspective transformation module (PIM) is the environmental features corresponding to the region where the plant image is located. Based on a Transformer encoder, PIM uses fully connected layers and a sigmoid activation function to re-analyze the image information, switching the system's perspective from the detector to the hider. For example, it simulates how a grower selects occluded areas, thereby fusing multi-scale features to generate a transformed feature map. The final output is the perspective-transformed feature map corresponding to the plant image. Compared to the original feature map, the perspective-transformed feature map highlights the spatial distribution of the suspected region.

[0055] (4) Strategy Motivation Modeling Module (SMMM) predicts adversarial strategies: The Strategy Motivation Modeling Module (SMMM) is based on a CNN environment encoder (including convolutional and pooling layers) and a multilayer perceptron. It analyzes historical planting patterns to predict potential adversarial behaviors (such as intercropping camouflage). It outputs strategy motivation weights to adjust recognition sensitivity. SMMM receives plant images and combines them with environmental data for strategy modeling.

[0056] Specifically, the Strategy Motivation Modeling (SMMM) module performs strategy modeling based on plant images to obtain the strategy motivation weights corresponding to the plant images, which are also the features of the Strategy Motivation Modeling (SMMM) module. The strategy motivation weights characterize the risk coefficients corresponding to potential adversarial behaviors in the plant images and are used to optimize probability calculations.

[0057] Finally, the outputs of all the above modules are converged to the decision layer (such as the feature fusion module or the adaptive perspective switching module). Through weighted fusion (e.g., 30% for fitness in ESAM and 40% for policy in SMMM) and a Softmax classifier, the planting probability of each possible planting area is calculated. For example, based on feature fusion and probability calculation of all module outputs, each possible poppy planting area and its planting probability are obtained.

[0058] For example, Figure 3 This is a schematic diagram illustrating the principle of the poppy identification method based on reverse thinking provided by this invention, and a flowchart illustrating the operation of the four modules, as shown below. Figure 3 As shown, the method includes: For any given plant image, the environmental features are first extracted using the environmental adaptability assessment module to obtain the environmental features and environmental fitness weights corresponding to the region where the plant image is located. For example, ... Figure 3 As shown, the process involves multiple 2D convolutions and linear rectified functions, followed by adaptive average pooling (2D), linear layers, and a sigmoid activation function. Finally, environmental features and environmental fitness weights (based on environmental features) are obtained.

[0059] Furthermore, based on the environmental features corresponding to the region where the plant image is located and the image features of the plant image, the cross-attention mechanism in the metacognitive identity control module is used to verify the identity consistency of the plant in the plant image with that of the poppy, thus obtaining the identity consistency confidence score of the plant in the plant image with the poppy (i.e., the consistency factor of the metacognitive identity control module). Figure 3 As shown, the environmental features corresponding to the area where the plant image is located and the image features of the plant image are processed by multiple two-dimensional convolutions and linear rectified functions, followed by adaptive average pooling (two-dimensional). Then, after flattening (operation), linear layer, linear rectified function, linear layer and sigmoid activation function, the consistency factor of the metacognitive identity control module is finally generated.

[0060] The viewpoint transformation module determines the viewpoint-transformed feature map of the plant image based on the environmental features corresponding to the region where the plant image is located. For example... Figure 3 As shown, firstly, the plant image is processed by an embodied advanced detector, which includes a UAV vision detector, a multimodal perception module, and an environment adaptation module (combining environmental features). After passing through the embodied advanced detector, the global and basic features of the plant image are obtained. For the global features, the current global features are extracted sequentially, passing through a linear layer, detector representation, and a sequential module (including a linear layer, layer normalization, linear rectified function, linear layer, layer normalization), an identity gate, and a hybrid identifier to obtain the feature-wise linear modulation parameters. For the basic features, the feature-wise linear modulation module is used for processing. The current features are passed through the sequential module (including a linear layer, layer normalization, linear rectified function, linear layer) to obtain the current global features, and then through a linear layer to obtain the viewpoint transformation module features (i.e., the viewpoint-transformed feature map corresponding to the plant image).

[0061] Furthermore, through the policy motivation modeling module, policy modeling is performed based on plant images to obtain the policy motivation weights corresponding to the plant images. For example... Figure 3 As shown, the strategic motivation modeling module features (i.e., the strategic motivation weights corresponding to the plant images) are obtained after passing through the UAV visual detector. The UAV visual detector includes a visual transformer, a feature fusion module, a spatial pyramid attention mechanism, a feature enhancement module, a multimodal perception module, and an environment adaptation module (combining environmental features).

[0062] Finally, the outputs of the above modules (consistency factor from the metacognitive identity control module, features from the perspective switching module, features from the strategy motivation modeling module, and environmental features) are input into the adaptive perspective switching module to obtain the possible poppy planting areas and the planting probability of each possible poppy planting area. In the figure, the areas marked with red and yellow boxes are the possible poppy planting areas, and the points in the marked boxes represent the planting probability. The more points in the figure, the higher the planting probability. For example, the planting probability of the area marked with red boxes is greater than that of the area marked with yellow boxes.

[0063] The beneficial effects of the image analysis module in identifying poppies are specifically manifested in the following ways: (1) Accuracy: Multi-dimensional feature fusion significantly reduces false alarm rate (such as interference from similar plants) and false negative rate (such as occluded areas).

[0064] (2) Adaptability: Environmental features and adversarial strategies modeling make the model applicable to complex scenarios such as mountainous areas and forests.

[0065] (3) Initiative: By reversing thinking, we can achieve a leap from passive identification to proactive prediction, which meets the forward-looking needs of drug control work.

[0066] The method provided in this embodiment integrates environmental analysis, identity verification, viewpoint transformation, and strategy prediction into a cohesive whole through a modular pipeline design, thereby improving the accuracy and efficiency of poppy identification in occluded areas.

[0067] According to the poppy identification method based on reverse thinking provided by the present invention, before analyzing and processing the plant image set using a trained image analysis model to identify at least one possible poppy planting area and the planting probability of each possible poppy planting area, the method includes: The plant image set is preprocessed to obtain a target image set; the preprocessing operation includes at least one of the following: contrast increase, noise suppression, sharpening, and color enhancement. The step of analyzing and processing the plant image set using a trained image analysis model to identify at least one possible poppy planting area and the planting probability of each possible poppy planting area includes: The trained image analysis model is used to analyze and process the target image set to identify each possible poppy planting area and the planting probability of each possible poppy planting area.

[0068] Specifically, in some embodiments, step 102 is preceded by: filtering the original set of plant images, for example, preprocessing the set of plant images to obtain a target set of images.

[0069] For example, all images in the plant image set are first filtered to retain those whose image quality meets the preset requirements. Then, each image that meets the requirements is preprocessed to obtain a preprocessed image. The target image set can be determined based on the preprocessed images.

[0070] Image preprocessing is a crucial step in poppy identification methods, aiming to improve the quality of images acquired by drones so that subsequent analysis models can more accurately identify suspected areas. Preprocessing operations include at least one of the following: contrast enhancement, noise suppression, sharpening, and color enhancement. These operations belong to image enhancement techniques, used to improve the visual effect of images and eliminate environmental interference (such as uneven lighting and weather effects), thereby providing reliable input for identifying potential poppy cultivation areas. The definition, purpose, and general implementation method of each processing operation are described below: 1. Increased contrast (Contrast Enhancement) Definition and Purpose: Contrast enhancement refers to widening the difference between bright and dark areas in an image, making bright areas brighter and dark areas darker, thereby enhancing the overall sense of depth in the image. In poppy recognition, this operation helps to highlight the boundary between the plant and the background, reduce blurring caused by insufficient lighting or shadows, and improve the model's sensitivity to poppy morphological features.

[0071] How to do it: Usually achieved by adjusting the distribution of pixel values ​​in the image. For example: (1) Histogram Equalization: Redistributes the brightness values ​​of image pixels to make the histogram (pixel brightness distribution map) more uniform, thereby enhancing contrast. This method is simple and efficient and suitable for natural scene images captured by drones. (2) Gamma Correction: Adjusts pixel values ​​using power-law transformation, and customizes the contrast level by setting gamma parameters (e.g., γ<1 enhances dark areas, γ>1 enhances bright areas). In document applications, this can be optimized for complex lighting conditions in mountainous areas.

[0072] 2. Noise Suppression Definition and Purpose: Noise suppression aims to reduce random interference points or graininess in images (such as sensor noise or weather-induced snowflake effects), improving image smoothness and signal-to-noise ratio. For poppy identification, noise may originate from drone flight vibrations or rainy weather; noise suppression can prevent false detections of similar plants.

[0073] How to do it: Use filtering algorithms to smooth the image: (1) Gaussian filtering: Convolve the image using a Gaussian function to blur noise points while preserving edge information. Suitable for suppressing slight noise in drone images. (2) Median filtering: Replace the current pixel value with the median of the pixel neighborhood to effectively remove impulse noise (such as salt and pepper noise) without significantly damaging the texture. This is very useful in the rural environment described in the document, where vegetated areas are easily disturbed by dust.

[0074] 3. Sharpening Definition and Purpose: Sharpening enhances the clarity of image edges and details, making object outlines more distinct. In poppy identification, it enhances leaf texture and flower shape, helping to distinguish poppies from similar plants such as corn poppies.

[0075] How to do it: Highlight high frequency components based on gradient or high-pass filtering: (1) Laplacian Operator: Apply second-order differential operators to detect edges, and achieve sharpening by superimposing the original image with edge information. This method is simple and fast, and is suitable for real-time processing of UAV images. (2) Unsharp Masking: First, perform Gaussian blur on the image to generate a mask, and then subtract the mask from the original image to enhance the edges. This method has strong controllability and the sharpening intensity can be adjusted according to the poppy characteristics.

[0076] 4. Color Enhancement Definition and Purpose: Color enhancement makes colors more vibrant or more realistic by adjusting the saturation, brightness, and hue of an image. For poppy identification, the red color of poppy flowers may fade in natural environments; color enhancement can improve the matching accuracy of a color feature library.

[0077] How to do it: Focus on color space transformation: (1) HSV space adjustment: Convert the image from RGB to HSV (hue, saturation, lightness) space, enhance the saturation (S) and lightness (V) components separately, and then convert it back to RGB space. This can specifically enhance the salience of poppy red. (2) Histogram Matching: Align the color histogram of the input image with the reference image (such as a standard poppy color template) to achieve color normalization. The "poppy color feature library" mentioned in the document can be optimized based on this method.

[0078] In the poppy identification process described in the document, these four preprocessing operations are often applied sequentially or in parallel (e.g., noise suppression followed by contrast enhancement) to ultimately generate a high-quality set of target images. By improving image quality, the reliability of the image analysis model is indirectly enhanced.

[0079] Subsequently, the trained image analysis model is used to analyze the target image set (and the processed image set) to identify each possible poppy planting area and the planting probability of each possible poppy planting area. The steps for analyzing the target image set (and the processed image set) using the trained image analysis model are similar to the steps for analyzing the plant image set using the trained image analysis model, and will not be repeated here.

[0080] The method provided in this embodiment preprocesses the plant image set before using the image analysis model for identification, thereby obtaining a high-quality target image set. The preprocessing operations include at least one of the following: contrast enhancement, noise suppression, sharpening, and color enhancement. By improving image quality, the reliability of the image analysis model is indirectly strengthened, and the accuracy of poppy identification is improved.

[0081] According to a poppy identification method based on reverse thinking provided by the present invention, the step of dividing each possible poppy planting area according to the planting probability of each possible poppy planting area to obtain a highly suspicious area for poppy planting includes: For any of the poppy-growing potential areas, the planting probability of the poppy-growing potential area is compared with a preset probability threshold; If the probability of poppy cultivation in the potential poppy-growing area is greater than or equal to the preset probability threshold, the potential poppy-growing area is identified as the highly suspicious area.

[0082] Specifically, in some embodiments, step 103 is implemented in the following manner, including: A preset probability threshold is set. The probability threshold indicates that the possible planting area corresponding to the probability value exceeding the threshold is a highly suspicious area. The specific division process is as follows: For any possible poppy-growing area, the planting probability of the possible poppy-growing area is compared with a preset probability threshold, and highly suspicious areas are determined based on the comparison results.

[0083] For example, if the preset probability threshold is 70%, then if the probability of poppy cultivation in a possible poppy-growing area is greater than or equal to 70%, that area will be identified as a highly suspicious area, thereby improving the accuracy and efficiency of poppy identification.

[0084] The method provided in this embodiment, after obtaining the recognition results of the image analysis model, divides highly suspicious areas according to the planting probability corresponding to each potential area (the area where poppies may be planted), thereby achieving efficient and accurate identification of poppies.

[0085] According to the poppy identification method based on reverse thinking provided by the present invention, the training steps of the trained image analysis model include: The initial image analysis model is trained using the Embodied Loss composite loss function and the Adam optimizer through a three-stage progressive training strategy to obtain the trained image analysis model. The three-stage progressive training strategy includes, in sequence, a basic detector pre-training stage, a modular training stage, and an end-to-end fine-tuning stage.

[0086] Specifically, in some embodiments, the trained image analysis model is obtained by training an initial image analysis model, and the training steps include the following: The training strategy employs a three-stage progressive approach: a pre-training stage for the base detector, a modular training stage, and an end-to-end fine-tuning stage. The specific implementation of this training strategy is described below, stage by stage, in conjunction with the modular functions of the image analysis model: 1. Basic detector pre-training phase (2000 epochs) Objective: To build the basic feature extraction capability of the model and to learn the general features of poppies (such as color and shape).

[0087] Training content: Train a shared feature extraction backbone network using a dataset of poppy images collected by drones (such as the UAV-Poppy dataset).

[0088] Technical details: The Adam optimizer was used, with an initial learning rate of 0.001 and a cross-entropy loss function.

[0089] Data augmentation techniques (such as rotation and scaling) improve generalization and avoid overfitting.

[0090] Compared with existing strategies: Unlike traditional pre-training that relies on general datasets, this stage directly uses poppy-specific data to ensure task relevance.

[0091] 2. Modular Training Phase Objective: To enhance the model's adaptability in adversarial scenarios by providing modular and specialized training on perspective switching and policy prediction capabilities.

[0092] Training content: Training for the following modules: Environmental Adaptability Assessment Module, Metacognitive Identity Control Module, Perspective Shifting Module, and Strategy Motivation Modeling Module. Input: Feature maps extracted in the basic stage + environmental adaptation data. Training focus: Using a Transformer encoder to learn the perspective transition from detector to hider, simulating the planter's camouflage strategy (e.g., generating adversarial examples through occlusion regions). Output: Feature maps after perspective transition, highlighting suspicious areas.

[0093] Technical details: The Embodied Loss composite loss function is adopted, which is based on consistency loss (viewpoint shift) and policy prediction loss.

[0094] The batch size is fixed at 128, and mixed precision training accelerates the process.

[0095] Module collaboration: This stage enables each module to acquire specialized capabilities, laying the foundation for end-to-end collaboration.

[0096] 3. End-to-end fine-tuning phase (2000 epochs) Objective: To integrate all modules, optimize overall performance, and improve recognition accuracy in real-world, complex scenarios.

[0097] Training content: The four modules (MICM, ESAM, PIM, SMMM) are connected in series to fine-tune the model in an end-to-end manner.

[0098] Input complete image data and output the probability of poppy cultivation. The focus is on optimizing feature fusion between modules (such as the environmental weight of ESAM and the policy weight of SMMM).

[0099] Technical details: Use a dynamic learning rate to avoid oscillations. Regularization techniques (such as Dropout) enhance robustness to cope with complex mountainous environments.

[0100] Evaluation and Iteration: Performance is evaluated by average accuracy and recall on the test set, and modular training is re-executed when the expected results are not achieved (i.e., the "relearning" mechanism).

[0101] The correspondence between each training stage and module function is shown in Table 1 below: Table 1

[0102] The Embodied Loss composite loss function in this method is as follows: L_total=L_detection+active_perception_scale×L_uncertainty +0.01×L_action+0.01×L_attention+iiet_scale×(L_KL+L_cf+L_id) +0.01×L_gate Where L_total is the Embodied Loss composite loss function, L_detection is the detection loss, L_uncertainty is the uncertainty loss, active_perception_scale is the weight coefficient corresponding to the uncertainty loss, L_action is the action prediction loss, L_attention is the attention loss, iiet_scale is the weight coefficient corresponding to the IIET-specific identity-related loss group, L_KL is the KL divergence loss of the IIET-specific identity-related loss group, L_cf is the counterfactual realism loss of the IIET-specific identity-related loss group, L_id is the identity adversarial loss of the IIET-specific identity-related loss group, and L_gate is the identity gating loss.

[0103] Embedded Loss employs a multi-component weighted summation strategy to balance the training objectives of different cognitive mechanisms. Its total loss function is a linear combination of the losses of each component. Specifically, the detection loss L_detection, as the basic component, is not given additional weight coefficients to ensure that the model's core object detection capability is fully optimized. Internally, this component has amplified the coordinate loss with coord_scale=5.0 to enhance localization accuracy, and reduced the penalty for background regions with noobj_scale=0.5 to reduce false positives. The uncertainty loss L_uncertainty related to active perception is configured with a weight coefficient of active_perception_scale=0.1. This moderate weight ensures that the model does not deviate excessively from the main detection task while focusing on uncertainty estimation. The action prediction loss L_action and the attention loss L_attention both use small weight coefficients of 0.01. This is because these two components mainly play an auxiliary regularization role, preventing the model from generating excessively large action prediction values ​​or overly scattered attention distributions. Excessive weights would interfere with the main target. IIET's unique identity-related loss set (KL divergence loss L_KL, counterfactual verifiability loss L_cf, and identity adversarial loss L_id) uses a weight coefficient of iiet_scale=0.1. This coefficient is carefully tuned to balance the relationship between identity switching mechanisms and detection performance, ensuring that the model can learn meaningful identity representations and perspective switching capabilities while avoiding excessive weakening of basic detection accuracy due to the complexity of identity modeling. The identity gating loss L_gate also uses a small weight of 0.01, serving only as a mild regularization term to guide the gating value towards a balanced state (around 0.5), without forcing a strict balance.

[0104] The method provided in this embodiment avoids the instability of one-time end-to-end training by decomposing complex tasks into three stages. Through staged training, the model gradually progresses from basic feature learning to adversarial strategy understanding, achieving efficient training of the image analysis model and ultimately realizing accurate identification of possible poppy planting areas.

[0105] According to the present invention, a poppy identification method based on reverse thinking is provided, the method further includes: Based on the highly suspicious area, a clearance order is triggered; In response to the clearance order, the target personnel for the clearance of the highly suspicious area are obtained by matching the status information of multiple law enforcement personnel; the status information includes: real-time location, workload, and professional capabilities. The coordinate information of the highly suspicious area is sent to the terminal device corresponding to the target investigation personnel.

[0106] Specifically, in some embodiments, the method further includes the investigation of highly suspicious areas, and the specific implementation process is as follows: First, a clearance command is triggered based on the highly suspicious areas. For example, when the model identifies a highly suspicious area (e.g., planting probability ≥ 70%), a confirmation signal is automatically generated. This signal is a structured data packet containing: the geographic coordinates of the suspicious area, such as a Global Positioning System (GPS) boundary point; the planting probability value (e.g., 75% or 90%); and a summary of the area's environmental features (e.g., occlusion type, terrain complexity). The confirmation signal is monitored in real time by the matching end. Once the signal is received, the data integrity is immediately verified (e.g., whether the coordinates are valid, whether the probability meets the standard). If the verification passes, a clearance command is automatically triggered. It is important to emphasize that the command triggering is instantaneous and requires no manual intervention.

[0107] Furthermore, in response to the aforementioned investigation order, the personnel status information of multiple law enforcement officers is monitored through a matching system. This personnel status information includes, for example: ① Real-time location: The current location of law enforcement officers is obtained via GPS or base station positioning to ensure priority matching of personnel closer to the suspected area. ② Task load: The system checks whether law enforcement officers are currently idle or already assigned tasks to avoid over-allocation. ③ Professional competence: The system assesses the personnel's experience in poppy investigations (e.g., novice or expert) based on historical task records.

[0108] Furthermore, the target personnel for the high-suspect area are identified by matching the status information of multiple law enforcement officers. Specifically, an optimization algorithm (such as a greedy algorithm or a weighted scoring model) is used at the matching end to calculate the "matching score" for each law enforcement officer. For example: Score = Distance weight × (1 / distance) + Load weight × (1 / number of tasks) + Capability weight × Experience value. The officer with the highest score is matched as the target personnel for the investigation, and the system automatically sends task instructions to the terminal (such as a smartphone or drone controller) of the officer with the highest score. The task instructions include detailed action guidelines, including at least the coordinate information of the high-suspect area, such as: coordinates of the suspect area and navigation path; expected actions (such as on-site investigation, sample collection); and emergency contact information.

[0109] The method provided in this embodiment generates a deterministic signal based on the probability results output by an image model, and then triggers a clearance command through a matching terminal. It matches target personnel based on their status, and the target personnel perform the clearance and provide feedback for optimization. This embodiment of the invention solves the problem of poppy identification in obscured areas, seamlessly integrating technical identification with clearance operations through a reverse thinking approach, thereby improving the efficiency of the clearance work.

[0110] The poppy identification device based on reverse thinking provided by the present invention will be described below. The poppy identification device based on reverse thinking described below can be referred to in correspondence with the poppy identification method based on reverse thinking described above.

[0111] Figure 4 This is a schematic diagram of the poppy identification device based on reverse thinking provided by the present invention, as shown below. Figure 4 As shown, the poppy identification device 400 based on reverse thinking includes the following modules: Image acquisition module 410 is used to acquire multi-channel images of the execution area using a drone to obtain a set of plant images; The identification module 420 is used to analyze and process the plant image set using a trained image analysis model to identify at least one possible poppy planting area and the planting probability of each possible poppy planting area; wherein, the trained image analysis model is constructed based on the embodied adversarial theory. Based on the planting probability of each of the possible poppy planting areas, the possible poppy planting areas are divided to obtain highly suspected poppy planting areas; the highly suspected areas are possible poppy planting areas where the planting probability is greater than a preset probability threshold. By reversing the thinking process, poppies in the execution area are identified based on the highly suspicious area.

[0112] The apparatus provided in this embodiment of the invention includes an image acquisition module 410, used to acquire multi-channel images of an execution area using a drone to obtain a set of plant images; and an identification module 420, used to analyze and process the set of plant images using a trained image analysis model to identify at least one possible poppy planting area and the planting probability of each possible poppy planting area; wherein, the trained image analysis model is constructed based on embodied adversarial theory; furthermore, each possible poppy planting area is divided according to the planting probability of each possible poppy planting area to obtain highly suspicious areas for poppy planting; highly suspicious areas are poppy planting areas with a planting probability greater than a preset probability threshold; and poppies in the execution area are identified based on highly suspicious areas using reverse thinking.

[0113] This invention constructs a trained image analysis model based on the theory of embodied adversarial behavior. The trained image analysis model is used to analyze and process the set of plant images to identify at least one possible poppy planting area and the planting probability of each possible poppy planting area. This allows for the determination of highly suspicious areas for poppy planting. By analyzing the growth environment characteristics of poppies, this invention identifies and classifies possible poppy planting areas. This method can effectively overcome the influence of factors such as occlusion, improve the accuracy of poppy identification, and significantly improve the detection efficiency of large areas.

[0114] According to the present invention, a poppy identification device 400 based on reverse thinking is provided, wherein the trained image analysis model includes an environmental adaptability assessment module, a metacognitive identity control module, a perspective switching module, and a strategy motivation modeling module; the identification module 420 is specifically used for: For any plant image in the plant image set, the environmental features of the plant image are extracted by the environmental adaptability assessment module to obtain the environmental features and environmental fitness weights of the region where the plant image is located. Based on the environmental features corresponding to the region where the plant image is located and the image features of the plant image, the cross-attention mechanism in the metacognitive identity control module is used to verify the identity consistency of the plant in the plant image with that of the poppy, and obtain the identity consistency confidence of the plant in the plant image with the poppy. The perspective conversion module determines the perspective-converted feature map of the plant image based on the environmental features corresponding to the region where the plant image is located; the perspective-converted feature map represents the spatial distribution of the suspected region. The strategy motivation modeling module performs strategy modeling based on the plant image to obtain the strategy motivation weights corresponding to the plant image; the strategy motivation weights characterize the risk coefficients corresponding to potential adversarial behaviors in the plant image. Based on the environmental features corresponding to the region where the plant image is located, the confidence level of the consistency between the plant and the poppy in the plant image, the feature map after the viewpoint transformation of the plant image, and the strategy motivation weight corresponding to the plant image, feature fusion and probability calculation are performed to obtain each possible poppy planting area and the planting probability of each possible poppy planting area.

[0115] According to the present invention, a poppy identification device 400 based on reverse thinking is provided, wherein the identification module 420 is further used for: The plant image set is preprocessed to obtain a target image set; the preprocessing operation includes at least one of the following: contrast increase, noise suppression, sharpening, and color enhancement. The trained image analysis model is used to analyze and process the target image set to identify each possible poppy planting area and the planting probability of each possible poppy planting area.

[0116] According to the present invention, a poppy identification device 400 based on reverse thinking is provided, wherein the identification module 420 is further used for: For any of the poppy-growing potential areas, the planting probability of the poppy-growing potential area is compared with a preset probability threshold; If the probability of poppy cultivation in the potential poppy-growing area is greater than or equal to the preset probability threshold, the potential poppy-growing area is identified as the highly suspicious area.

[0117] According to the present invention, a poppy identification device 400 based on reverse thinking is provided, wherein the model further includes a training module; The training module is used for: The initial image analysis model is trained using the Embodied Loss composite loss function and the Adam optimizer through a three-stage progressive training strategy to obtain the trained image analysis model. The three-stage progressive training strategy includes, in sequence, a basic detector pre-training stage, a modular training stage, and an end-to-end fine-tuning stage.

[0118] According to the present invention, a poppy identification device 400 based on reverse thinking is provided, the device further includes a personnel matching module; The personnel matching module is used for: Based on the highly suspicious area, a clearance order is triggered; In response to the clearance order, the target personnel for the clearance of the highly suspicious area are obtained by matching the status information of multiple law enforcement personnel; the status information includes: real-time location, workload, and professional capabilities. The coordinate information of the highly suspicious area is sent to the terminal device corresponding to the target investigation personnel.

[0119] Figure 5 This is a schematic diagram of the structure of the electronic device provided by the present invention, such as... Figure 5 As shown, the electronic device may include: a processor 510, a communications interface 520, a memory 530, and a communication bus 540, wherein the processor 510, the communications interface 520, and the memory 530 communicate with each other via the communication bus 540. The processor 510 can call logical instructions in the memory 530 to execute a poppy detection method based on reverse thinking, which includes: The drone was used to acquire multi-channel images of the target area, resulting in a collection of plant images. The trained image analysis model is used to analyze and process the plant image set to identify at least one possible poppy planting area and the planting probability of each possible poppy planting area; wherein, the trained image analysis model is constructed based on the embodied adversarial theory; Based on the planting probability of each of the possible poppy planting areas, the possible poppy planting areas are divided to obtain highly suspected poppy planting areas; the highly suspected areas are possible poppy planting areas where the planting probability is greater than a preset probability threshold. By reversing the thinking process, poppies in the execution area are identified based on the highly suspicious area.

[0120] Furthermore, the logical instructions in the aforementioned memory 530 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0121] On the other hand, the present invention also provides a computer program product, the computer program product comprising a computer program that can be stored on a non-transitory computer-readable storage medium, wherein when the computer program is executed by a processor, the computer is able to execute the poppy identification method based on reverse thinking provided by the above methods, the method comprising: The drone was used to acquire multi-channel images of the target area, resulting in a collection of plant images. The trained image analysis model is used to analyze and process the plant image set to identify at least one possible poppy planting area and the planting probability of each possible poppy planting area; wherein, the trained image analysis model is constructed based on the embodied adversarial theory; Based on the planting probability of each of the possible poppy planting areas, the possible poppy planting areas are divided to obtain highly suspected poppy planting areas; the highly suspected areas are possible poppy planting areas where the planting probability is greater than a preset probability threshold. By reversing the thinking process, poppies in the execution area are identified based on the highly suspicious area.

[0122] In another aspect, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to perform the poppy identification method based on reverse thinking provided by the above methods, the method comprising: The drone was used to acquire multi-channel images of the target area, resulting in a collection of plant images. The trained image analysis model is used to analyze and process the plant image set to identify at least one possible poppy planting area and the planting probability of each possible poppy planting area; wherein, the trained image analysis model is constructed based on the embodied adversarial theory; Based on the planting probability of each of the possible poppy planting areas, the possible poppy planting areas are divided to obtain highly suspected poppy planting areas; the highly suspected areas are possible poppy planting areas where the planting probability is greater than a preset probability threshold. By reversing the thinking process, poppies in the execution area are identified based on the highly suspicious area.

[0123] The device 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.

[0124] 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 part that contributes 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, magnetic disk, optical disk, 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.

[0125] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A poppy identification method based on reverse thinking, characterized in that, include: The drone was used to acquire multi-channel images of the target area, resulting in a collection of plant images. The trained image analysis model is used to analyze and process the plant image set to identify at least one possible poppy planting area and the planting probability of each possible poppy planting area; wherein, the trained image analysis model is constructed based on the embodied adversarial theory; Based on the planting probability of each of the possible poppy planting areas, the possible poppy planting areas are divided to obtain highly suspected poppy planting areas; the highly suspected areas are possible poppy planting areas where the planting probability is greater than a preset probability threshold. By reversing the thinking process, poppies in the execution area are identified based on the highly suspicious area.

2. The poppy identification method based on reverse thinking according to claim 1, characterized in that, The trained image analysis model includes an environmental adaptability assessment module, a metacognitive identity control module, a perspective switching module, and a strategy motivation modeling module. The process of analyzing and processing the plant image set using the trained image analysis model to identify at least one possible poppy planting area and the planting probability of each possible poppy planting area includes: For any plant image in the plant image set, the environmental features of the plant image are extracted by the environmental adaptability assessment module to obtain the environmental features and environmental fitness weights of the region where the plant image is located. Based on the environmental features corresponding to the region where the plant image is located and the image features of the plant image, the cross-attention mechanism in the metacognitive identity control module is used to verify the identity consistency of the plant in the plant image with that of the poppy, and obtain the identity consistency confidence of the plant in the plant image with the poppy. The perspective conversion module determines the perspective-converted feature map of the plant image based on the environmental features corresponding to the region where the plant image is located; the perspective-converted feature map represents the spatial distribution of the suspected region. The strategy motivation modeling module performs strategy modeling based on the plant image to obtain the strategy motivation weight corresponding to the plant image; the strategy motivation weight represents the risk coefficient corresponding to the potential adversarial behavior in the plant image. Based on the environmental features corresponding to the region where the plant image is located, the confidence level of the consistency between the plant and the poppy in the plant image, the feature map after the viewpoint transformation of the plant image, and the strategy motivation weight corresponding to the plant image, feature fusion and probability calculation are performed to obtain each possible poppy planting area and the planting probability of each possible poppy planting area.

3. The poppy identification method based on reverse thinking according to claim 1, characterized in that, Before analyzing and processing the plant image set using the trained image analysis model to identify at least one possible poppy planting area and the planting probability of each possible poppy planting area, the process includes: The plant image set is preprocessed to obtain a target image set; the preprocessing operation includes at least one of the following: contrast increase, noise suppression, sharpening, and color enhancement. The step of analyzing and processing the plant image set using a trained image analysis model to identify at least one possible poppy planting area and the planting probability of each possible poppy planting area includes: The trained image analysis model is used to analyze and process the target image set to identify each possible poppy planting area and the planting probability of each possible poppy planting area.

4. The poppy identification method based on reverse thinking according to claim 1, characterized in that, The step of dividing the potential poppy-growing areas according to the planting probability of each potential poppy-growing area to obtain highly suspected areas for poppy cultivation includes: For any of the poppy-growing potential areas, the planting probability of the poppy-growing potential area is compared with a preset probability threshold; If the probability of poppy cultivation in the potential poppy-growing area is greater than or equal to the preset probability threshold, the potential poppy-growing area is identified as the highly suspicious area.

5. The poppy identification method based on reverse thinking according to any one of claims 1-4, characterized in that, The training steps of the trained image analysis model include: The initial image analysis model is trained using the Embodied Loss composite loss function and the Adam optimizer through a three-stage progressive training strategy to obtain the trained image analysis model. The three-stage progressive training strategy includes, in sequence, a basic detector pre-training stage, a modular training stage, and an end-to-end fine-tuning stage.

6. The poppy identification method based on reverse thinking according to any one of claims 1-4, characterized in that, The method further includes: Based on the highly suspicious area, a clearance order is triggered; In response to the clearance order, the target personnel for the clearance of the highly suspicious area are obtained by matching the status information of multiple law enforcement personnel; the status information includes: real-time location, workload, and professional capabilities. The coordinate information of the highly suspicious area is sent to the terminal device corresponding to the target investigation personnel.

7. A poppy identification device based on reverse thinking, characterized in that, include: The image acquisition module is used to acquire multi-channel images of the execution area using a drone to obtain a set of plant images; The identification module is used to analyze and process the plant image set using a trained image analysis model to identify at least one possible poppy planting area and the planting probability of each possible poppy planting area; wherein, the trained image analysis model is constructed based on the embodied adversarial theory. Based on the planting probability of each of the possible poppy planting areas, the possible poppy planting areas are divided to obtain highly suspected poppy planting areas; the highly suspected areas are possible poppy planting areas where the planting probability is greater than a preset probability threshold. By reversing the thinking process, poppies in the execution area are identified based on the highly suspicious area.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the poppy identification method based on reverse thinking as described in any one of claims 1 to 6.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the poppy identification method based on reverse thinking as described in any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the poppy identification method based on reverse thinking as described in any one of claims 1 to 6.

Images