A method and system for constructing a database of forestry pest species

Abstract

The present application relates to the technical field of forestry pest image data processing and database construction, and specifically discloses a forestry pest species database construction method and system. The present application collects forestry pest images and multi-dimensional information such as location, time and illumination, performs species labeling and morphological feature point labeling, divides regional image sets according to the collection location, further subdivides species image sub-sets according to the species label, and then clusters and divides insect state image sub-sets according to morphological feature labels such as the head, chest and wing vein, constructs a three-level directory structure of the regional layer, species layer and insect state layer, and establishes a multi-dimensional index relationship, thereby solving the problems of insufficient sample coverage, lack of labeling standardization and poor scene adaptability of the existing forestry pest database, and improving the structured storage and fine management level of forestry pest data.

Description

Technical Field

[0001] This invention relates to the field of forest pest image data processing and database construction technology, and in particular to a method and system for constructing a database of forest pest species. Background Technology

[0002] In the field of forest pest control, accurate identification of pest species is the prerequisite and foundation for implementing precision control. Currently, the identification of forest pests mainly relies on manual morphological comparison and expert judgment. This method is not only inefficient but also requires a high level of professional expertise from practitioners, making it difficult to meet the needs of large-scale, high-frequency monitoring. In recent years, with the development of image recognition and deep learning technologies, some studies have attempted to apply convolutional neural networks to pest image classification. However, existing methods are mostly designed for agricultural or storage pests, lacking data support specifically for forestry ecological scenarios. Forest pests are characterized by their diverse species, varied life stages, and complex habitats. Their image acquisition is easily affected by factors such as lighting conditions, background interference, and shooting angle, resulting in significant deficiencies in the sample coverage, labeling standardization, and scene adaptability of existing general databases, making it difficult to support the training and deployment of high-precision identification models. Therefore, it is urgent to construct a pest species database that is structurally standardized and accurately labeled for forestry application scenarios, laying the foundation for the practical application of intelligent monitoring technologies.

[0003] Therefore, there is an urgent need to provide a technical solution to address the above problems. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention provides a method and system for constructing a database of forest pest species.

[0005] Firstly, the present invention provides a method for constructing a database of forest pest species, the technical solution of which is as follows:

[0006] Acquire images of forest pests, including the acquisition location, acquisition time, and lighting conditions of the images;

[0007] The forest pest images are labeled with pest species to generate species labels, and the individual pests in the forest pest images are labeled with morphological feature points to generate morphological feature labels.

[0008] Based on the collection location, the forestry pest images are divided into corresponding regional image sets;

[0009] In each of the regional image sets, the forestry pest images are divided into corresponding category image subsets according to the category labels;

[0010] In each of the aforementioned image subsets, the forestry pest images are clustered according to the morphological feature labels to divide the insect life image subsets. The morphological feature labels include at least head feature labels, thorax feature labels, and wing vein feature labels.

[0011] A three-level directory structure is constructed, comprising a regional layer, a species layer, and an insect life stage layer. The regional image set is associated with the regional layer, the species image subset is associated with the species layer, and the insect life stage image subset is associated with the insect life stage layer.

[0012] Establish an index relationship between the forest pest images and the collection location, collection time, lighting conditions, species label, and morphological feature label, and store the forest pest images according to the three-level directory structure to form a forest pest database.

[0013] Secondly, this invention provides a system for constructing a database of forest pest species, the technical solution of which is as follows:

[0014] The acquisition module is used to acquire images of forest pests, as well as the acquisition location, acquisition time, and lighting conditions of the forest pest images;

[0015] The annotation module is used to annotate the forest pest images with pest species and generate species labels, and to annotate the individual pests in the forest pest images with morphological feature points and generate morphological feature labels.

[0016] The classification module is used to classify the forestry pest images into corresponding regional image sets based on the collection location;

[0017] The segmentation module is used to segment the forestry pest images into corresponding category image subsets in each of the regional image sets according to the category labels.

[0018] The clustering module is used to cluster the forestry pest images in each of the image subsets according to the morphological feature labels, and divide the insect life image subsets. The morphological feature labels include at least head feature labels, thorax feature labels, and wing vein feature labels.

[0019] The association module is used to construct a three-level directory structure that includes a regional layer, a species layer, and an insect life layer. It associates the set of regional images with the regional layer, the subset of species images with the species layer, and the subset of insect life images with the insect life layer.

[0020] The generation module is used to establish an index relationship between the forest pest images and the collection location, collection time, lighting conditions, species labels, and morphological feature labels, and to store the forest pest images according to the three-level directory structure to form a forest pest database.

[0021] The technical solution of this invention collects images of forest pests and their multi-dimensional information such as location, time, and illumination, performs species labeling and morphological feature point labeling, divides the regional image set according to the collection location, further subdivides the category image subsets according to the category label, and then clusters the insect life image subsets according to morphological feature labels such as head, thorax, and wing veins. It constructs a three-level directory structure of regional layer, category layer, and insect life layer and establishes a multi-dimensional index relationship, which solves the problems of insufficient sample coverage, lack of labeling standardization, and poor scene adaptability of existing forest pest databases, and improves the level of structured storage and refined management of forest pest data.

[0022] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it in accordance with the contents of the specification, and in order to make the above and other objects, features and advantages of the present invention more apparent and understandable, specific embodiments of the present invention are described below. Attached Figure Description

[0023] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0024] The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings:

[0025] Figure 1 This is a flowchart illustrating an embodiment of a method for constructing a forestry pest species database according to the present invention.

[0026] Figure 2 This is a schematic diagram of an embodiment of a forestry pest species database construction system according to the present invention. Detailed Implementation

[0027] Exemplary embodiments of the invention will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the invention are shown in the drawings, it should be understood that the invention can be implemented in various forms and should not be limited to the embodiments set forth herein.

[0028] Figure 1This diagram illustrates a flowchart of an embodiment of a method for constructing a forestry pest species database provided by the present invention, executed by a control terminal. Figure 1 As shown, it includes the following steps:

[0029] S1. Acquire images of forest pests, including the location, time, and lighting conditions at which the images were collected.

[0030] Among them, forest pest images refer to digital pictures containing individual images of pests acquired through image acquisition equipment in the forest ecological environment; for example, in Forest Farm C, City B, Province A, at 10:23 on January 1, 2026, an image of an adult forest pest A with a resolution of 12 million pixels is taken using an image acquisition device, and this image is a forest pest image.

[0031] The acquisition location refers to the spatial information of the geographical location at which the forest pest image was acquired; for example, the acquisition location of the image of adult forest pest A is Forest Farm C, City B, Province A, specifically latitude X degrees North and longitude Y degrees East. The acquisition time refers to the time information recorded when the forest pest image was acquired; for example, the acquisition time of the image of adult forest pest A is 10:23 AM on January 1, 2026. The lighting conditions refer to the quantitative representation of the ambient lighting conditions at the time the forest pest image was acquired; for example, the lighting conditions for the image of adult forest pest A are a sunny day with a natural light intensity of 8500 lx.

[0032] S2. Label the forest pest images with pest species to generate species labels, and label the individual pests in the forest pest images with morphological feature points to generate morphological feature labels.

[0033] Among them, the category label refers to the category identifier assigned to the image of forest pest based on the pest species identification results; for example, an image of an adult forest pest A is assigned the category label "forest pest A" after being identified by insect classification experts.

[0034] In this context, "individual pest" refers to a single pest organism presented in an image of forest pests; for example, an intact adult forest pest A located on the bark surface in an image of an adult forest pest A is an individual pest. Morphological feature point annotation refers to the operation of identifying and marking key morphological structural points of individual pests in an image of forest pests; for example, in an image of an adult forest pest A, the annotator annotates the head outline point set along the outer edge of the head, the pronotum outline point set along the edge of the pronotum, and the forewing vein intersection point set at the junction of the forewing veins. This operation is morphological feature point annotation. Morphological feature labels refer to data identifiers used to characterize the morphological and structural features of individual pests; for example, the coordinate data of the head outline point set, the coordinate data of the pronotum outline point set, and the coordinate data of the forewing vein intersection point set generated after morphological feature point annotation of an image of an adult forest pest A together constitute the morphological feature label.

[0035] S3. Based on the collection location, divide the forest pest images into the corresponding regional image sets.

[0036] Among them, a regional image set refers to a collection of forest pest images that share the same geographical origin and are divided according to the collection location; for example, all forest pest images collected from forest farm C in city B of province A constitute a regional image set.

[0037] S4. In each of the regional image sets, the forestry pest images are divided into corresponding category image subsets according to the category labels.

[0038] Among them, the category image subset refers to the subset of forestry pest images that have the same pest species, which are further divided according to the category label within the regional image set; for example, in the regional image set of Forest Farm C in City B of Province A, all forestry pest images with the category label "Forestry Pest A" together constitute a category image subset.

[0039] S5. In each of the aforementioned image subsets, the forestry pest images are clustered according to the morphological feature labels to divide the insect life image subsets. The morphological feature labels include at least head feature labels, thorax feature labels, and wing vein feature labels.

[0040] Among them, the insect life image subset refers to the forest pest image subset that corresponds to the same developmental stage, which is divided into clusters based on morphological feature labels within the species image subset; for example, in the forest pest A species image subset, all forest pest A images in the adult stage are classified into the same insect life image subset by clustering based on morphological feature labels.

[0041] Among them, the head feature label refers to a data label generated through morphological feature point annotation, used to characterize the head outline morphology of an individual pest; for example, the coordinate data of the head outline point set annotated in the image of an adult forest pest A is the head feature label. The thorax feature label refers to a data label generated through morphological feature point annotation, used to characterize the thorax tergite morphology of an individual pest; for example, the coordinate data of the pronotum tergite outline point set annotated in the image of an adult forest pest A is the thorax feature label. The wing vein feature label refers to a data label generated through morphological feature point annotation, used to characterize the wing vein structure morphology of an individual pest; for example, the coordinate data of the forewing vein intersection point set annotated in the image of an adult forest pest A is the wing vein feature label.

[0042] S6. Construct a three-level directory structure including a regional layer, a species layer, and an insect state layer. Associate the set of regional images with the regional layer, associate the subset of species images with the species layer, and associate the subset of insect state images with the insect state layer.

[0043] The three-level directory structure is divided into three layers: **Regional Layer:** The first layer organizes forest pest images according to their collection location. For example, in this structure, "Province A, City B, Forest Farm C" serves as a regional layer node, linking all forest pest images corresponding to that location. **Specialty Layer:** The second layer organizes forest pest images according to pest species. For example, under the regional layer node "Province A, City B, Forest Farm C," "Forest Pest A" serves as a species layer node, linking all forest pest images labeled "Forest Pest A" at that location. **Insect Development Layer:** The third layer organizes forest pest images according to their developmental stage. For example, under the species layer node "Forest Pest A," "Adult Insect" serves as an insect development layer node, linking all images of forest pests in the adult stage within that category.

[0044] S7. Establish an index relationship between the forest pest images and the collection location, collection time, lighting conditions, species label, and morphological feature label, and store the forest pest images according to the three-level directory structure to form a forest pest database.

[0045] The forestry pest database refers to a collection of images of forestry pests and their associated information stored in a three-level directory structure and with multi-dimensional index relationships. For example, a structured data set containing images of all forestry pests A in forest farm area C of city B in province A, including images of each insect stage, the collection location, collection time, lighting conditions, species label, and morphological feature label for each image, is a forestry pest database.

[0046] The technical solution of this embodiment collects images of forest pests and their multi-dimensional information such as location, time, and illumination, performs species labeling and morphological feature point labeling, divides the regional image set according to the collection location, further subdivides the category image subsets according to the category label, and then clusters the insect life image subsets according to morphological feature labels such as head, thorax, and wing veins. It constructs a three-level directory structure of regional layer, category layer, and insect life layer and establishes a multi-dimensional index relationship, which solves the problems of insufficient sample coverage, lack of labeling standardization, and poor scene adaptability of existing forest pest databases, and improves the level of structured storage and refined management of forest pest data.

[0047] In one alternative approach, the step of annotating individual pests in the forestry pest image with morphological feature points to generate morphological feature labels includes:

[0048] The head contour point set of the pest individual is labeled as the head feature label, the pronotum contour point set of the pest individual is labeled as the thorax feature label, and the forewing vein intersection point set of the pest individual is labeled as the wing vein feature label.

[0049] The head outline point set refers to the set of coordinates of several pixels marked along the outer edge of the head of an individual forest pest in an image. For example, in an image of an adult forest pest A, the set of coordinates of 30 pixels marked sequentially along the outer edge of the head of pest A constitutes the head outline point set. The pronotum outline point set refers to the set of coordinates of several pixels marked along the outer edge of the pronotum of an individual forest pest in an image. For example, in an image of an adult forest pest A, the set of coordinates of 25 pixels marked sequentially along the outer edge of the pronotum of pest A constitutes the pronotum outline point set. The forewing vein intersection point set refers to the set of coordinates of several pixels marked at the intersections of the forewing veins of an individual forest pest in an image. For example, in an image of an adult forest pest A, the set of coordinates of 18 intersections formed by the convergence of the forewing veins marked by the annotator constitutes the forewing vein intersection point set.

[0050] Among the above-mentioned optional methods, the specific annotation content of the morphological feature labels can be further refined, and the annotation objects of the head outline point set, the pronotum outline point set, and the forewing vein intersection point set can be clarified, so that the morphological feature labels can form a precise correspondence with the insect anatomical structure, providing a standardized morphological basis for subsequent insect stage clustering.

[0051] In one alternative approach, the step of clustering the forestry pest images according to the morphological feature labels within each subset of said species images to divide the insect life image subset includes:

[0052] Extract the head contour point set, the pronotum contour point set, and the forewing vein intersection point set from each image of the forest pest. Calculate the centroid coordinates of the head contour point set, the pronotum contour point set, and the forewing vein intersection point set, respectively. Combine the three centroid coordinates into a morphological feature vector. Calculate the Euclidean distance between the morphological feature vectors of different forest pest images. Assign forest pest images whose Euclidean distance is less than a preset distance threshold to the same insect life image subset.

[0053] Centroid coordinates refer to the geometric center coordinates of the region or set of points enclosed by the contour point set or intersection point set. For example, the geometric center coordinates of the region enclosed by the head contour point set in the image of an adult forest pest A are pixel coordinates (345, 278), which are the centroid coordinates of the head contour point set. Morphological feature vectors refer to multidimensional vectors formed by combining the centroid coordinates of multiple morphological feature point sets of an individual pest. For example, the six-dimensional vector (345, 278, 352, 310, 428, 356) formed by combining the centroid coordinates of the head contour point set (345, 278), the pronotum contour point set (352, 310), and the forewing vein intersection point set (428, 356) in the image of an adult forest pest A is the morphological feature vector.

[0054] Here, Euclidean distance refers to a metric that measures the spatial distance between two morphological feature vectors. For example, the Euclidean distance calculated between the morphological feature vectors of one image of an adult forest pest A and another image of an adult forest pest A is 23.5. The preset distance threshold refers to a pre-set upper limit of the Euclidean distance used to determine whether two images of forest pests belong to the same insect stage subset. For example, if the preset distance threshold is set to 30, when the Euclidean distance between the morphological feature vectors of two images of adult forest pest A is 23.5 and less than 30, the two images are considered to belong to the same insect stage subset.

[0055] In the above-mentioned optional methods, the quantization and clustering of insect state image subsets are further realized based on the Euclidean distance of morphological feature vectors. Feature vectors are constructed by calculating the centroid coordinates of three types of contour point sets, and a preset distance threshold is used as the classification criterion, so that the insect state division process has a quantifiable geometric measurement basis.

[0056] In one alternative approach, the steps of acquiring the acquisition location, acquisition time, and lighting conditions of the forest pest images include:

[0057] The image acquisition device obtains the acquisition location through its built-in positioning module, the acquisition time through its built-in timing module, and the lighting conditions through its built-in light sensor.

[0058] Image acquisition equipment refers to hardware devices used to acquire images of forest pests; for example, a portable insect image acquisition terminal equipped with a 12-megapixel camera, built-in positioning components, timing components, and a light sensor is an image acquisition device. Positioning module refers to the component in the image acquisition device used to acquire geographical location information; for example, the GPS positioning chip built into a portable insect image acquisition terminal is a positioning component that automatically records the geographical location information of X degrees North latitude and Y degrees East longitude when capturing an image of forest pest A.

[0059] The timing module refers to the component in the image acquisition device used to obtain time information; for example, the real-time clock chip built into the portable insect image acquisition terminal is a timing component, which automatically records the time information of 10:23 AM on January 1, 2026 when capturing an image of forest pest A. The light sensor refers to the component in the image acquisition device used to sense the intensity of ambient light; for example, the ambient light sensor built into the portable insect image acquisition terminal is a light sensor, which automatically senses an ambient light intensity of 8500 lx when capturing an image of forest pest A.

[0060] Among the above-mentioned optional methods, further expand the automated acquisition methods of multi-dimensional information by using the positioning module, timing module and light sensor built into the image acquisition device to obtain the location, time and lighting conditions respectively, reducing the manual input process and improving the real-time performance and completeness of data acquisition.

[0061] In one alternative approach, the step of obtaining the illumination conditions further includes:

[0062] The original light intensity value collected by the light sensor is compared with a preset light level mapping table to generate a light level label, which is used to characterize the light environment category in which the forest pest image is located.

[0063] The original light intensity value refers to the unprocessed ambient light value directly collected by the light sensor. For example, the 8500 lx value output by the light sensor when capturing an image of forest pest A in a portable insect image acquisition terminal is the original light intensity value. The preset light level mapping table refers to a pre-established table that converts the original light intensity value into light level labels. For example, the preset light level mapping table specifies that 0 to 3000 lx corresponds to low light level, 3001 to 8000 lx corresponds to medium light level, and above 8001 lx corresponds to strong light level. Therefore, the 8500 lx original light intensity value of the forest pest A image is mapped to a strong light level.

[0064] The illumination level label refers to the level identifier used to characterize the illumination environment category of a forest pest image; for example, the "strong light level" generated after comparing the original illumination intensity value of 8500 lx of the forest pest image A with a preset illumination level mapping table is the illumination level label. Illumination environment category refers to the environment type classified according to illumination intensity; for example, the "strong light level" is the illumination environment category corresponding to the forest pest image A.

[0065] In the above-mentioned optional methods, the continuous illumination parameters are further converted into discrete level labels, and the original illumination intensity values ​​are compared and classified by a preset illumination level mapping table, so that the illumination environment has a unified semantic description standard, which facilitates subsequent data retrieval by illumination category.

[0066] In one alternative approach, after constructing a three-level directory structure comprising a regional layer, a species layer, and an insect stage layer, and associating the regional image set with the regional layer, the species image subset with the species layer, and the insect stage image subset with the insect stage layer, the method further includes:

[0067] An image thumbnail index is established for each subset of insect images under the insect state layer. The image thumbnail index is generated from the forest pest image with the highest standard of individual insect posture in the subset of insect images.

[0068] The image thumbnail index refers to a simplified image identifier used for quickly previewing the contents of a subset of insect life images. For example, in the subset of images of adult forest pest A, the image of adult forest pest A with the highest posture standard is selected to generate a thumbnail with a size of 200 pixels × 200 pixels. This thumbnail is the image thumbnail index.

[0069] In the above-mentioned optional methods, an image thumbnail indexing mechanism is further established under the insect stage layer. The images of forest pests with the highest pose standard are selected as representative samples of the subset, which optimizes the database visualization browsing experience and shortens the retrieval path for users to locate target insect stage images.

[0070] In one alternative approach, the step of determining the standardization of the individual pest's posture includes:

[0071] Calculate the angle between the main axis of the body of an individual forest pest in the image and a preset standard direction, and calculate the symmetry of the bilateral appendages of the individual pest. Calculate a posture score based on the angle and the bilateral appendage symmetry, and select the forest pest image with the highest posture score to generate the image thumbnail index.

[0072] The main axis direction refers to the direction in which the long axis of the insect's body extends in the image. For example, in an image of an adult forest pest A, the line connecting the front of the head to the end of the abdomen is horizontally oriented 15 degrees to the right and upward; this direction is the main axis direction. The preset standard direction refers to a pre-defined reference direction used to measure the standard of the insect's posture. For example, the preset standard direction is set to the right horizontally, used for comparison with the main axis direction of the adult forest pest A.

[0073] Among them, bilateral appendage symmetry refers to the degree of symmetry in shape and position between the left and right appendages of an individual pest. For example, in an image of an adult forest pest A, the difference in length between the left and right antennae is 2%, and the difference in joint angle between the left and right forelegs is 3%, resulting in a bilateral appendage symmetry of 92%. Posture score refers to a quantitative score used to evaluate the postural standard of an individual pest, calculated based on the angle between the main axis of the pest's body and a preset standard direction, as well as the bilateral appendage symmetry. For example, in an image of an adult forest pest A, the angle between the main axis of the pest's body and the preset standard direction is 15 degrees, and the bilateral appendage symmetry is 92%, resulting in a posture score of 88.

[0074] Among the above-mentioned optional methods, the evaluation index of the standardity of individual pest postures is further quantified. The posture score is calculated by combining the angle of the main axis of the body and the symmetry of the bilateral appendages, so that the selection of thumbnails has a quantifiable geometric judgment criterion, ensuring the posture standardization of representative samples.

[0075] In one alternative approach, the step of clustering the forestry pest images according to the morphological feature labels within each subset of said species images to divide the insect life image subset includes:

[0076] Calculate the insect development coefficient for each of the forest pest images, and divide the forest pest images into corresponding insect development image subsets based on the insect development coefficient;

[0077] The insect development coefficient is calculated using the following formula:

[0078] ;

[0079] in, This represents the developmental coefficient of the insect stage. This represents the area enclosed by the set of head contour points marked in the head feature label. Indicates the preset head reference area. This represents the area enclosed by the set of anterior chest back contour points marked in the chest feature label. This indicates the preset chest baseline area. This represents the area of ​​the convex hull formed by the set of intersection points of the forewing veins marked in the wing vein feature label. This indicates the preset wing vein reference area. This represents the average curvature of the set of forewing vein intersections. Indicates the preset curvature reference value. The first point in the set of forewing vein crossings indicates the... The length of the pulse segment between each intersection point and its adjacent intersection points. This represents the average length of all pulse segments. Indicates the total number of pulse segments. , , , , These represent the first area weighting coefficient, the second area weighting coefficient, the wing vein area inhibition coefficient, the curvature attenuation coefficient, and the vein segment variation adjustment coefficient, respectively.

[0080] It should be noted that the formula for the insect development coefficient is based on the biological characteristics of forest pests, which exhibit regular morphological evolution of the head, thorax, and wing vein structures at different developmental stages. The area enclosed by the head outline point set and the area enclosed by the pronotum outline point set are used as core indicators to characterize the developmental degree of the insect's trunk. The area of ​​the convex hull formed by the forewing vein intersection point set, the average curvature of the forewing vein intersection point set, and the degree of variation in vein segment length are used as key parameters to characterize wing vein maturity. The above multidimensional morphological features are integrated through a nonlinear function, and preset benchmark area, preset curvature benchmark value, and multiple weighting coefficients are introduced to achieve differentiated adjustment of sensitivity to different insect stages. Thus, a comprehensive evaluation index that can continuously quantify the developmental process of an individual pest from an early stage to a mature stage is constructed.

[0081] The above formula, by introducing a wing vein curvature attenuation term and a vein segment variation adjustment term, effectively overcomes the deficiency of traditional single morphological indicators in distinguishing closely related insect stages. It is particularly effective for forestry pests, where morphological differences between larvae and adults are significant, while morphological differences between larvae of different ages are subtle. This achieves refined insect stage classification and biological consistency. The formula provides a uniform and comparable quantified insect stage value for forestry pest images within each image subset, enabling the system to automatically group images of pests at the same developmental stage into the corresponding insect stage image subset based on the numerical range of this coefficient. This solves the clustering difficulties caused by the variability in insect stages of forestry pests and lays a numerical classification foundation for the subsequent structured organization of insect stage levels.

[0082] Among the above-mentioned optional methods, an insect development coefficient is further introduced as a clustering basis. Through nonlinear combination calculations of head area, thorax area, wing vein bulge area and its curvature, and vein segment variability, the physiological maturity of insect growth and development is comprehensively reflected, so that insect stage classification takes into account both morphological geometric characteristics and biological laws.

[0083] In one alternative approach, the step of clustering the forestry pest images according to the morphological feature labels within each subset of said species images to divide the insect life image subsets further includes:

[0084] Calculate the morphological coupling degree between each forest pest image and each candidate cluster center in the subset of insect life images, and take the insect life image subset to which the candidate cluster center with the largest morphological coupling degree belongs as the category to which the forest pest image belongs.

[0085] The morphological coupling degree is calculated using the following formula:

[0086] ;

[0087] in, This indicates the degree of morphological coupling. This represents the developmental coefficient of the current forest pest image. This represents the insect-like developmental coefficient of the candidate cluster center image. Indicates the developmental coefficient coupling index. Indicates the morphological difference attenuation coefficient. This represents the total number of categories of morphological feature points. This indicates the first [insect] in the current forestry pest image. The set of position vectors of morphological feature points. Indicates the first candidate cluster center in the image The set of position vectors of morphological feature points, wherein the first type of morphological feature points corresponds to the head contour point set in the head feature label, the second type of morphological feature points corresponds to the pronotum contour point set in the thorax feature label, and the third type of morphological feature points corresponds to the forewing vein intersection point set in the wing vein feature label.

[0088] It should be noted that the formula for morphological coupling degree is based on the idea of ​​dual similarity measurement of insect development coefficient and spatial distribution of morphological feature points. The coupling degree between the insect development coefficient of the current forestry pest image and the candidate cluster center image is calculated by a ratio function. The development coefficient coupling index is used to adjust the contribution weight of this ratio term to the overall coupling degree. At the same time, the spatial distribution difference between the two types of morphological feature point sets is measured by the ratio of the product of vector spatial distance and vector magnitude. The morphological difference attenuation coefficient is used to control the penalty of this difference term on the coupling degree. Finally, the development coefficient similarity and the spatial distribution similarity of morphological feature points are fused into a comprehensive index in the form of a product.

[0089] The above formula, by introducing a high-dimensional vector matching mechanism for the spatial distribution of morphological feature points, solves the problem of traditional clustering methods relying solely on global features while neglecting the correspondence of local morphological structures. This is particularly relevant for forestry pest images where significant morphological differences exist between images of the same species due to variations in shooting angle and posture. It achieves precise quantification of the matching degree of local anatomical structures, significantly improving the anti-interference capability of insect stage boundary delineation. The formula establishes a refined similarity evaluation standard between each forestry pest image to be classified and each candidate cluster center. By selecting the candidate cluster center with the highest morphological coupling degree as the belonging object, the boundary delineation of insect stage image subsets considers both the overall consistency of developmental stages and the matching degree of local morphological features of the head, thorax, and wing veins, thereby improving the accuracy and robustness of insect stage clustering.

[0090] Among the above-mentioned optional methods, the insect state classification mechanism is further optimized. The similarity of development coefficients and the spatial distribution difference of morphological feature points between the image to be classified and each candidate cluster center are measured by morphological coupling degree calculation. The maximum coupling degree is used as the basis for classification, thereby improving the accuracy of insect state classification and boundary discrimination ability.

[0091] Figure 2 A schematic diagram of an embodiment of a forestry pest species database construction system provided by the present invention is shown. Figure 2 As shown, the forestry pest species database construction system includes:

[0092] The acquisition module 201 is used to acquire images of forest pests, as well as the acquisition location, acquisition time, and lighting conditions of the forest pest images;

[0093] The annotation module 202 is used to annotate the forest pest images with pest species and generate species labels, and to annotate the individual pests in the forest pest images with morphological feature points and generate morphological feature labels.

[0094] The classification module 203 is used to classify the forestry pest images into corresponding regional image sets according to the collection location;

[0095] The segmentation module 204 is used to segment the forestry pest images into corresponding category image subsets according to the category labels in each of the regional image sets.

[0096] Clustering module 205 is used to cluster the forestry pest images in each of the image subsets according to the morphological feature labels, and divide the insect life image subsets. The morphological feature labels include at least head feature labels, thorax feature labels and wing vein feature labels.

[0097] The association module 206 is used to construct a three-level directory structure including a regional layer, a species layer, and an insect life layer, and associates the regional image set with the regional layer, the species image subset with the species layer, and the insect life image subset with the insect life layer.

[0098] The generation module 207 is used to establish an index relationship between the forest pest images and the collection location, collection time, lighting conditions, species labels, and morphological feature labels, and to store the forest pest images according to the three-level directory structure to form a forest pest database.

[0099] The technical solution of this embodiment collects images of forest pests and their multi-dimensional information such as location, time, and illumination, performs species labeling and morphological feature point labeling, divides the regional image set according to the collection location, further subdivides the category image subsets according to the category label, and then clusters the insect life image subsets according to morphological feature labels such as head, thorax, and wing veins. It constructs a three-level directory structure of regional layer, category layer, and insect life layer and establishes a multi-dimensional index relationship, which solves the problems of insufficient sample coverage, lack of labeling standardization, and poor scene adaptability of existing forest pest databases, and improves the level of structured storage and refined management of forest pest data.

[0100] Furthermore, the system provided in the above embodiments is only illustrated by the division of the above functional modules. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the system can be divided into different functional modules according to the actual situation to complete all or part of the functions described above. In addition, the system and method embodiments provided in the above embodiments belong to the same concept, and their specific implementation process can be found in the method embodiments, which will not be repeated here.

[0101] The above description is merely a preferred embodiment of the present invention and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of disclosure in this invention is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the above-disclosed concept. For example, technical solutions formed by substituting the above features with (but not limited to) technical features with similar functions disclosed in this invention.

[0102] It should be noted that the terms "first," "second," etc., used in the specification and claims of this application are used to distinguish similar objects and do not imply a specific order or sequence. Where appropriate, the order of use for similar objects can be interchanged so that the embodiments of this application described herein can be implemented in an order other than that shown or described.

[0103] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A method for constructing a database of forest pest species, characterized in that, The method includes: Acquire images of forest pests, including the acquisition location, acquisition time, and lighting conditions of the images; The forest pest images are labeled with pest species to generate species labels, and the individual pests in the forest pest images are labeled with morphological feature points to generate morphological feature labels. Based on the collection location, the forestry pest images are divided into corresponding regional image sets; In each of the regional image sets, the forestry pest images are divided into corresponding category image subsets according to the category labels; In each of the aforementioned image subsets, the forestry pest images are clustered according to the morphological feature labels to divide the insect life image subsets. The morphological feature labels include at least head feature labels, thorax feature labels, and wing vein feature labels. A three-level directory structure is constructed, comprising a regional layer, a species layer, and an insect life layer. The regional image set is associated with the regional layer, the species image subset is associated with the species layer, and the insect life image subset is associated with the insect life layer. Establish an index relationship between the forest pest images and the collection location, collection time, lighting conditions, species label, and morphological feature label, and store the forest pest images according to the three-level directory structure to form a forest pest database; In each of the aforementioned image subsets, the step of clustering the forestry pest images based on the morphological feature labels to divide the insect life image subsets includes: Calculate the insect development coefficient for each of the forest pest images, and divide the forest pest images into corresponding insect development image subsets based on the insect development coefficient; The insect development coefficient is calculated using the following formula: ； in, This represents the developmental coefficient of the insect stage. This represents the area enclosed by the set of head contour points marked in the head feature label. Indicates the preset head reference area. This represents the area enclosed by the set of anterior chest back contour points marked in the chest feature label. This indicates the preset chest baseline area. This represents the area of ​​the convex hull formed by the set of intersection points of the forewing veins marked in the wing vein feature label. This indicates the preset wing vein reference area. This represents the average curvature of the set of forewing vein intersections. Indicates the preset curvature reference value. The first point in the set of forewing vein crossings indicates the... The length of the pulse segment between each intersection point and its adjacent intersection points. This represents the average length of all pulse segments. Indicates the total number of pulse segments. , , , , These represent the first area weighting coefficient, the second area weighting coefficient, the wing vein area inhibition coefficient, the curvature attenuation coefficient, and the vein segment variation adjustment coefficient, respectively.

2. The method for constructing a forestry pest species database according to claim 1, characterized in that, The steps of annotating individual forest pests in the images with morphological feature points and generating morphological feature labels include: The head contour point set of the pest individual is labeled as the head feature label, the pronotum contour point set of the pest individual is labeled as the thorax feature label, and the forewing vein intersection point set of the pest individual is labeled as the wing vein feature label.

3. The method for constructing a forestry pest species database according to claim 2, characterized in that, In each of the aforementioned image subsets, the step of clustering the forestry pest images based on the morphological feature labels to divide the insect life image subsets includes: Extract the head contour point set, the pronotum contour point set, and the forewing vein intersection point set from each image of the forest pest. Calculate the centroid coordinates of the head contour point set, the pronotum contour point set, and the forewing vein intersection point set, respectively. Combine the three centroid coordinates into a morphological feature vector. Calculate the Euclidean distance between the morphological feature vectors of different forest pest images. Assign forest pest images whose Euclidean distance is less than a preset distance threshold to the same insect life image subset.

4. The method for constructing a forestry pest species database according to claim 1, characterized in that, The steps for obtaining the acquisition location, acquisition time, and lighting conditions of the forest pest images include: The image acquisition device obtains the acquisition location through its built-in positioning module, the acquisition time through its built-in timing module, and the lighting conditions through its built-in light sensor.

5. The method for constructing a forestry pest species database according to claim 4, characterized in that, The step of obtaining the illumination conditions further includes: The original light intensity value collected by the light sensor is compared with a preset light level mapping table to generate a light level label, which is used to characterize the light environment category in which the forest pest image is located.

6. The method for constructing a forestry pest species database according to claim 1, characterized in that, After constructing a three-level directory structure comprising a regional layer, a species layer, and an insect life stage layer, and associating the regional image set with the regional layer, the species image subset with the species layer, and the insect life stage image subset with the insect life stage layer, the method further includes: An image thumbnail index is established for each subset of insect images under the insect state layer. The image thumbnail index is generated from the forest pest image with the highest standard of individual insect posture in the subset of insect images.

7. The method for constructing a forestry pest species database according to claim 6, characterized in that, The steps for determining the standard of individual pest posture include: Calculate the angle between the main axis of the body of an individual forest pest in the image and a preset standard direction, and calculate the symmetry of the bilateral appendages of the individual pest. Calculate a posture score based on the angle and the bilateral appendage symmetry, and select the forest pest image with the highest posture score to generate the image thumbnail index.

8. The method for constructing a forestry pest species database according to claim 1, characterized in that, The step of clustering the forestry pest images according to the morphological feature labels and dividing them into insect life image subsets in each of the aforementioned image subsets further includes: Calculate the morphological coupling degree between each forest pest image and each candidate cluster center in the subset of insect life images, and take the insect life image subset to which the candidate cluster center with the largest morphological coupling degree belongs as the category to which the forest pest image belongs. The morphological coupling degree is calculated using the following formula: ； in, This indicates the degree of morphological coupling. This represents the developmental coefficient of the current forest pest image. This represents the insect-like developmental coefficient of the candidate cluster center image. Indicates the developmental coefficient coupling index. Indicates the morphological difference attenuation coefficient. This represents the total number of categories of morphological feature points. This indicates the first [insect] in the current forestry pest image. The set of position vectors of morphological feature points. Indicates the first candidate cluster center in the image The set of position vectors of morphological feature points, wherein the first type of morphological feature points corresponds to the head contour point set in the head feature label, the second type of morphological feature points corresponds to the pronotum contour point set in the thorax feature label, and the third type of morphological feature points corresponds to the forewing vein intersection point set in the wing vein feature label.

9. A system for constructing a database of forest pest species, characterized in that, The system, applied to the forestry pest species database construction method as described in any one of claims 1 to 8, comprises: The acquisition module is used to acquire images of forest pests, as well as the acquisition location, acquisition time, and lighting conditions of the forest pest images; The annotation module is used to annotate the forest pest images with pest species and generate species labels, and to annotate the individual pests in the forest pest images with morphological feature points and generate morphological feature labels. The classification module is used to classify the forestry pest images into corresponding regional image sets based on the collection location; The segmentation module is used to segment the forestry pest images into corresponding category image subsets in each of the regional image sets according to the category labels. The clustering module is used to cluster the forestry pest images in each of the image subsets according to the morphological feature labels, and divide the insect life image subsets. The morphological feature labels include at least head feature labels, thorax feature labels, and wing vein feature labels. The association module is used to construct a three-level directory structure that includes a regional layer, a species layer, and an insect life layer. It associates the set of regional images with the regional layer, the subset of species images with the species layer, and the subset of insect life images with the insect life layer. The generation module is used to establish an index relationship between the forest pest images and the collection location, collection time, lighting conditions, species labels, and morphological feature labels, and to store the forest pest images according to the three-level directory structure to form a forest pest database.

Images