Method for generating image data for insect identification

Abstract

To provide an insect identification image data generation method capable of enhancing the precision in identifying an insect included in image data.SOLUTION: An insect identification image data generation method according to the present invention is a method for generating insect identification image data by processing pre-processing image data in which an insect is included. The method comprises: a possibility determination step for determining whether portions of an insect included in the pre-processing image data are identifiable; and an identification step for identifying and extracting each portion of the insect whose portions are determined to be identifiable in the possibility determination step, and assigning information corresponding to each portion to the extracted range.SELECTED DRAWING: Figure 3

Description

[Technical Field]

[0001] The present invention relates to a method for generating image data for insect identification. [Background technology]

[0002] A conventional system for identifying captured insects is described in Patent Document 1. The system includes an image reading means for reading an image of an adhesive sheet to which an insect is attached, and an analysis device for analyzing the read image to identify the captured insect. The analysis device includes a preprocessing means for performing image processing on the read image to distinguish between an insect region, which is an area where the insect is captured, and a background, a feature extraction means for acquiring insect landmark data from the preprocessed image, a data storage means for storing standard landmark data that standardizes landmarks for insect morphological characteristics, and an identification means for identifying the insect by comparing the landmark data extracted by the feature extraction means with the standard landmark data stored in the data storage means.

[0003] The above-described captured insect identification system is said to be able to obtain identification results that are more objective and consistent than when insects are identified manually. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Japanese Patent Application Laid-Open No. 2014-142833 Summary of the Invention [Problem to be solved by the invention]

[0005] In a captured insect identification system such as that described in Patent Document 1, captured insects take on various postures, so if, for example, the insect is stuck to the adhesive sheet in a bent state, the insect's morphological characteristics cannot be properly extracted, resulting in incorrect recognition and low identification accuracy.

[0006] Therefore, an object of the present invention is to provide a method for generating image data for insect identification that can improve the accuracy of identifying insects contained in image data. [Means for solving the problem]

[0007] The method for generating image data for insect identification of the present invention is a method for generating image data for insect identification by processing pre-processed image data that includes insects in the image, and comprises a feasibility determination step of determining whether it is possible to identify parts of the insects contained in the pre-processed image data, and an identification step of identifying and extracting each part of the insects whose parts are determined to be identifiable in the feasibility determination step, and assigning information corresponding to each part to the extracted range.

[0008] Furthermore, it is also possible to configure the system so that information indicating that an insect is present is attached to any insect for which it is determined that the location cannot be identified in the determination step.

[0009] The image processing device may also include an insect extraction step for extracting insect data relating to insects from the pre-processing image data, and the possibility determination step may be configured to determine whether or not it is possible to identify the location of an insect corresponding to the insect data extracted in the insect extraction step.

[0010] The insect extraction process can also be configured to include an insect detection process for detecting insects contained in the unprocessed image data, and an insect identification process for adding frame information relating to a frame surrounding the insects detected in the insect detection process and extracting the insect data based on the added frame information.

[0011] The method may also be configured to include a point information assigning step of assigning point information corresponding to each part of the insect identified in the identifying step. [Effects of the Invention]

[0012] According to the present invention, it is possible to obtain a method for generating image data for insect identification that can improve the accuracy of identifying insects contained in image data. [Brief explanation of the drawings]

[0013] [Figure 1] 1 is a schematic diagram showing an image capturing device for capturing pre-processed image data in a method for generating an image for insect identification according to an embodiment of the present invention. [Figure 2] 10 is a flowchart showing the process of a method for generating an image for identifying the insect. [Figure 3] 10 is a flowchart showing the process of a region extraction step in the method for generating an image for identifying the insect. [Figure 4] 10 is a diagram showing an image corresponding to pre-processed image data to which frame information has been added in the method for generating an image for identification of the insect. FIG. [Figure 5] 10A and 10B show insects corresponding to insect data for which the part extraction process has been completed in the method for generating images for insect identification, where (a) shows an insect corresponding to insect data for which the part can be determined, and (b) shows an insect corresponding to insect data for which the part cannot be determined. [Figure 6] 10A and 10B show insects corresponding to insect data to which point information and order information have been added in the method for generating images for insect identification, where (a) shows an insect corresponding to insect data whose boundaries are recognizable, and (b) shows an insect corresponding to insect data whose boundaries are not recognizable. [Figure 7] 1 is a diagram showing an overview of an insect identification system according to an embodiment of the present invention. [Figure 8] FIG. 1 is a block diagram showing the same insect identification system. [Figure 9] FIG. 1 is a block diagram showing a pest control system that performs response control. [Figure 10] 10 is a flowchart showing the process of determining an increase in insects, determining the source of infestation, and determining a treatment method in a pest control system that performs countermeasure control. [Figure 11] 10 is a flowchart showing the process of determining the chemical treatment method in a pest control system that performs response control. [Figure 12] 10 is a flowchart showing the processing of an implementation feasibility determination step in a pest control system that performs response control. [Figure 13]1 is a data table used to determine various treatment decisions in a pest control system that performs response control. [Figure 14] FIG. 1 is a block diagram showing a pest control system that performs preventive control. [Figure 15] 10 is a flowchart showing the flow of processing in a pest control system that performs preventive control. [Figure 16] 10 is a data table used to determine information on the environment for insect proliferation in a pest control system that performs preventive control. DETAILED DESCRIPTION OF THE INVENTION

[0014] A method for generating image data for insect identification and an insect identification system 6 according to one embodiment of the present invention will be described with reference to FIGS.

[0015] First, the pre-processed image data processed by the insect identification image data generating method of this embodiment will be described with reference to FIG. 1. The pre-processed image data is generated by an imaging device 3 capable of imaging the capture surface 2a on which the insect 1 is captured, in a captured sample 2 having a capture surface 2a capable of capturing the insect 1. The captured sample 2 of this embodiment is a sample configured so that the insect 1 can adhere to the capture surface 2a, and multiple insects 1 are captured while adhering to the capture surface 2a; specifically, it is an insect paper on which the insects 1 are captured. Here, "insects" includes all small animals other than mammals, birds, and seafood, and in this embodiment refers to arthropods (especially insects and spiders). The capture surface 2a has a flat shape with a surface extending in one direction (in this embodiment, the direction penetrating the paper in FIG. 1). Furthermore, the capture surface 2a is plain, and in this embodiment, it is an entirely yellow surface without any lines or the like.

[0016] A plurality of insects 1 are adhered to the capture surface 2a of the captured sample 2. The captured sample 2 in this embodiment may be a sample created by sprinkling dead insects 1 on the capture surface 2a and causing them to adhere, or a sample created by attracting insects 1 with light or scent and causing them to adhere to the capture surface 2a, but is not limited to these.

[0017] The imaging device 3 is a device that images the captured sample 2 and generates pre-processing image data. Specifically, the imaging device 3 includes a box-shaped housing 31 that can block external light, a holding means (not shown) that is provided inside the housing 31 and can hold the captured sample 2 with the plane direction of the capture surface 2a extending in one direction (in this embodiment, a state in which the plane direction extends in the direction penetrating the paper surface of FIG. 1), an imaging means 32 that is provided inside the housing 31 and can image the capture surface 2a of the captured sample 2, an irradiation device 33 that is provided inside the housing 31 and can irradiate the capture surface 2a of the captured sample 2 with light, and an imaging processing unit (not shown) that processes the image captured by the imaging means 32 to generate pre-processing image data.

[0018] The imaging means 32 includes a camera 34 disposed at a distance from the capture surface 2a of the captured sample 2 held by the holding means so as to capture an image of the capture surface 2a, and a travel means (not shown) for moving the camera 34 along the surface direction of the capture surface 2a of the captured sample 2 held by the holding means. The camera 34 can be moved (traveled) along the surface direction of the capture surface 2a by the travel means, and can capture images of the capture surface 2a at multiple positions in the surface direction of the capture surface 2a while moving in the surface direction of the capture surface 2a. In addition, the camera 34 in this embodiment is disposed at a position facing the capture surface 2a. The camera 34 has a focal length such that only the portion of the capture surface 2a facing the camera 34 is in focus.

[0019] The illumination device 33 is configured to illuminate the capture surface 2a with light from both sides of the travel area in which the camera 34 travels. Specifically, the illumination device 33 has illumination units 35 on one side and the other side of the travel area, and each illumination unit 35 illuminates the capture surface 2a with light. The illumination units 35 are arranged to illuminate light from positions spaced approximately 45 degrees above the center of the width direction of the capture surface 2a (a direction perpendicular to the surface direction). The illumination device 33 of this embodiment is configured so that multiple illumination units 35 are arranged in a row along the travel area, and can illuminate light over the entire travel area in one direction. The illumination device 33 of this embodiment illuminates the capture surface 2a with white light.

[0020] The imaging processing unit generates pre-processed image data by stitching together multiple images captured by the imaging means 32. Specifically, the imaging processing unit processes multiple images captured by the camera 34 to stitch together in-focus portions of each image to generate pre-processed image data. In this embodiment, the imaging processing unit stitches together images of portions of the capture surface 2a facing the camera 34 from multiple images captured by the camera 34 to generate pre-processed image data.

[0021] The pre-processed image data generated as described above is data representing an image of an insect 1 resting on the capture surface 2a, for example, as shown in Fig. 4. The pre-processed image data of this embodiment is image data representing an image in which the capture surface 2a is used as a background and multiple insects 1 are captured on the background, and the pre-processed image data includes multiple pieces of insect data D relating to the insects 1. The insect data D is image data of the portion of the pre-processed image data that represents the insects 1.

[0022] Next, a method for generating image data for insect identification by processing pre-processed image data will be described with reference to Figures 2 to 6. The method for generating image data for insect identification according to this embodiment is a method for generating image data for insect identification as training data used in machine learning.

[0023] As shown in FIG. 2, the method for generating image data for insect identification according to this embodiment processes pre-processed image data containing an insect 1 as described above to generate image data for insect identification. Specifically, the method for generating image data for insect identification includes an insect extraction step S1 for extracting insect data D related to the insect 1 from the pre-processed image data, a part extraction step S2 for extracting parts of the insect 1 included in the extracted insect data D, a point information assignment step S3 for assigning point information 4 to the extracted insect data D, and an order information assignment step S4 for assigning order information 5 related to the order to the point information 4. In the method for generating image data for insect identification according to this embodiment, each of the above steps is performed for all insect data D before the next step is performed. Furthermore, in this embodiment, each step is performed manually.

[0024] The insect extraction step S1 is a step of executing an insect detection step S11 for detecting insects 1 included in an image corresponding to the unprocessed image data, and an insect identification step S12 for identifying the detected insects 1. That is, in the insect extraction step S1, the insect detection step S11 detects insects 1, and the insect identification step S12 extracts data to which information for identifying the detected insects 1 has been added as insect data D.

[0025] The insect detection step S11 is a step of processing the pre-processed image data to detect an insect 1 included in the image corresponding to the pre-processed image data. The insect detection step S11 of this embodiment is a step of searching for a part having characteristics of an insect 1 from the image corresponding to the pre-processed image data, and more specifically, when a part of a color different from the background has an outline of a predetermined shape, the part surrounded by the outline is determined to be one insect 1.

[0026] 4, the insect identification step S12 is a step of executing a frame information assignment step, which is a step of assigning frame information, which is information about a frame F surrounding the insect 1 discovered in the insect discovery step S11, an extraction step, which extracts insect data based on the frame information assigned in the frame information assignment step, and a type information assignment step, which assigns type information about the type of the discovered insect 1. That is, in the insect identification step S12 of this embodiment, information about the range and type of insect 1 included in the image corresponding to the unprocessed image data is identified, and insect data D about the identified insect 1 is extracted.

[0027] The frame information assigning step is a step of identifying the area occupied by the discovered insect 1 in the pre-processed image data and assigning frame information relating to a frame F that surrounds the area occupied by the insect 1. Specifically, the frame information assigning step is a step of identifying the outer edge of the insect 1 discovered in the insect detection step S11 and assigning information relating to a frame F that surrounds the outer edge. The frame F corresponding to the frame information assigned in this embodiment is square, particularly rectangular, and, for example, the coordinates of two points located at diagonal corners of the rectangular frame F are assigned as frame information.

[0028] The extraction step is a step of extracting insect data D based on the frame information. Specifically, in the extraction step, data in which the area where the insect 1 discovered in the insect detection step S11 is located is enclosed by a frame F is extracted as insect data D. In other words, the insect data D is image data extracted from the pre-processing image data by enclosing the area where the insect 1 is located by a frame F.

[0029] The type information assigning process is a process of identifying the type of insect 1 that has been discovered (specifically, the type in the classification required to identify the source of the insect 1, which will be described later, for example, the family in the biological classification hierarchy to which the insect 1 belongs), and assigning information about the type of insect 1 to the insect data D. Here, the assigned type information is information that can identify the type of insect 1, such as the type name of the insect 1 or an identifier related to the type of insect 1. In the type information assigning process, information about the type name of the insect 1 enclosed in the frame F corresponding to the frame information is assigned to the insect data D.

[0030] 2 and 3, the part extraction step S2 is a step of extracting parts of the insect 1 corresponding to each extracted insect data D. Furthermore, in the part extraction step S2 of this embodiment, it is determined whether or not the part of the insect 1 corresponding to each insect data D can be identified, and if the part is identifiable, the part of the insect 1 corresponding to each insect data D is identified and extracted. Specifically, the part extraction step S2 is a step of executing a selection step S21 of selecting the insect 1 corresponding to the insect data D to be processed, a feasibility determination step S22 of determining whether or not the part of the insect 1 corresponding to the insect data D selected in the selection step S21 can be identified, an identification step S23 of identifying and extracting the part of the insect 1 corresponding to the insect data D whose part is identifiable, and a non-designation processing step S24 of specifying not to identify the part of the insect 1 corresponding to the insect data D whose part cannot be identified. In the part extraction step S2 of this embodiment, it is determined whether or not there is unprocessed insect data D, and if there is unprocessed insect data D, a repeat step S25 of repeating each process in the part extraction step S2 is executed.

[0031] The body structure of the insect 1 will now be described with reference to Figures 4 to 6. The insect 1 comprises a main body 11 having a head 13 and a body 14 connected to the head 13 and including at least an abdomen 16, and appendages 12 extending outward from the main body 11. The appendages 12 include, for example, antennae, legs, wings, etc. The structure of the main body 11 varies depending on the insect 1. For example, the main body 11 of an insect comprises the head 13 and a thorax 15 and abdomen 16 as the body 14. The main body 11 of a spider comprises a cephalothorax as the head 13 and an abdomen 16 as the body 14. In this embodiment, the parts of the insect 1 will be described as referring to each part of the main body 11 of the insect 1.

[0032] 3, the selection step S21 is a step of selecting one insect 1 as a processing target from the insects 1 corresponding to the extracted plurality of insect data D. Also, in the selection step S21, an insect 1 corresponding to unprocessed insect data D that has not been processed in the part extraction step S2 is selected.

[0033] The possibility determination step S22 is a step of determining whether or not it is possible to identify the parts of the insect 1 corresponding to the selected insect data D. In this embodiment, the possibility determination step S22 determines whether or not each part is clearly visible in the image of the insect 1 corresponding to the selected insect data D. If each part is clearly visible, it is determined that the parts of the insect 1 corresponding to the selected insect data D are clearly visible. In this embodiment, the possibility determination step S22 determines that the parts are clearly visible if at least the head 13 and the body 14 are clearly visible. For example, as shown in FIG. 5(a), if the boundaries between each part (the head 13 and the body 14) in the main body 11 of the insect 1 corresponding to the insect data D are clearly visible, it is determined that the parts of the insect 1 are clearly visible. On the other hand, as shown in FIG. 5(b), if the main body 11 of the insect 1 corresponding to the insect data D is hidden by the appendages 12 (specifically, the wings), other insects 1, dust, etc., and the boundary between the head 13 and the body 14 cannot be clearly distinguished, it is determined that the parts cannot be identified. Furthermore, in the possibility determination step S22, it is determined that the part cannot be identified even when it is impossible to determine the part of the insect 1 corresponding to the insect data D, such as when the insect 1 is crushed or damaged on the capture surface 2a and part of the part is lost or separated from the main body 11. In this embodiment, even if part of the insect 1 corresponding to the insect data D is crushed or hidden, it is determined that the part of the insect 1 can be identified as long as the main body 11 can be recognized to an extent that allows the part of the insect 1 to be determined.

[0034] The identification step S23 is a step of identifying and extracting each part of the insect 1 that is determined to be identifiable in the possibility determination step S22. Specifically, the identification step S23 identifies and extracts the range in which each part is located from the insect 1 image data corresponding to the insect data D. For example, as shown in FIG. 5(a), if the entire main body 11 of the insect 1 corresponding to the insect data D can be recognized and the boundaries between the head 13, thorax 15, and abdomen 16 can be recognized in the identification step S23, the range in which the head 13, the range in which the thorax 15, and the range in which the abdomen 16 are located are identified and extracted from the main body 11 of the insect 1, and information corresponding to each part is assigned to each extracted range. Furthermore, in the identification step S23, if the boundary between the head 13 and thorax 15 of the insect 1 corresponding to the insect data D can be recognized but the boundary between the thorax 15 and abdomen 16 cannot be recognized, or if the insect 1 is an insect (e.g., a spider) whose torso 14 does not have at least one of the thorax 15 and the abdomen 16 (e.g., as shown in Figure 6(b)), at least two or more areas, the range where the head 13 is located and the range where the torso 14 is located, are identified and extracted, and information corresponding to each area is assigned to each extracted area.

[0035] The non-designation processing step S24 is a step of designating that the part of an insect 1 corresponding to the insect data D determined to be part-identifiable in the possibility determination step S22 will not be identified. For the insect 1 corresponding to the insect data D determined to be part-identifiable in the non-designation processing step S24, information indicating that the part cannot be identified is given, for example, as shown in FIG. 5(b).

[0036] The repetition step S25 is a step of determining whether there is an insect 1 corresponding to unprocessed insect data D that has not been processed in the part extraction step S2, and if there is an unprocessed insect 1, repeating each step in the part extraction step S2. Specifically, in the repetition step S25, if the multiple insect data D included in the raw image data include insect data D that has not been processed in the identification step S23 or the undesignated processing step S24, it is determined that there is an insect 1 corresponding to the unprocessed insect data D, and the process proceeds to the selection step S21. Furthermore, in the repetition step S25, if the processing in the identification step S23 or the undesignated processing step S24 has been completed for all insect data D included in the raw image data, the part extraction step S2 is terminated.

[0037] As shown in FIG. 2 , the point information assignment process S3 is a process of assigning point information 4 to at least three points on the insect data D. Furthermore, in the point information assignment process S3, three points on the surface of the insect 1 corresponding to the insect data D are assigned to the point information 4. Specifically, three points spaced apart along the entire length of the body 11 of the insect 1 (the direction connecting the head and abdomen of the insect 1) are assigned to the point information 4. That is, the point information assignment process S3 is a process of assigning point information 4 to at least three points within the range in which the insect 1 corresponding to the insect data D is located in the pre-processing image data. Specifically, the point information 4 is assigned to at least three points within the range in which each part of the insect 1 extracted in the part extraction process S2 is located. In the point information assignment process S3 of this embodiment, point information 4 is assigned so that the three points correspond to two or more parts of the multiple parts constituting the body 11 of the insect 1 corresponding to the insect data D. For example, point information 4 is assigned so that three points correspond to the head 13 and the torso 14. 6, in the point information assignment step S3 of this embodiment, for the insect 1 corresponding to each insect data D, point information 4 is assigned at least to a position corresponding to the head 13 of the insect 1 and a position corresponding to the body 14 of the insect 1. Specifically, the point information 4 is assigned by specifying three points: a head corresponding point 41 corresponding to the head 13 of the insect 1, an abdomen corresponding point 42 corresponding to the abdomen 16 of the insect 1, and an intermediate point 43 located between the head corresponding point 41 and the abdomen corresponding point 42. In this way, in the point information assignment step S3 of this embodiment, the point information 4 is assigned based on the part of the insect 1 corresponding to the insect data D. Therefore, the point information 4 is assigned to the insect data D whose part has been extracted in the part extraction step S2, but the point information 4 is not assigned to the insect data D whose part has been determined to be unidentifiable in the part extraction step S2.

[0038] The head corresponding point 41 is a point designated at the tip of the head 13 (the end on the head side in the overall length direction of the body main body 11) and is also a point designated at approximately the center in the width direction of the body main body 11 (a direction perpendicular to the overall length direction of the body main body 11 in the image of the insect 1 corresponding to the insect data D). That is, the head corresponding point 41 is a point designated at the tip of the body main body 11 of the insect 1 and is also a point designated at approximately the center in the width direction of the body main body 11. The abdomen corresponding point 42 is a point designated at the rear end of the abdomen 16 (the end on the ventral side in the overall length direction of the body main body 11) and is also a point designated at approximately the center in the width direction of the body main body 11. That is, the abdomen corresponding point 42 is a point designated at the rear end of the torso 14 (the rear end of the body main body 11 of the insect 1) and is also a point designated at approximately the center in the width direction of the body main body 11.

[0039] The midpoint 43 is a point specified in the midpoint of the main body 11 of the insect 1 in the lengthwise direction. The midpoint 43 is also specified in the approximate center of the insect 1 in the widthwise direction, and in this embodiment, it is a point located in the torso 14. Furthermore, the midpoint 43 is a point located at a characteristic point in the main body 11 of the insect 1. A characteristic point is a point that is characteristic of the insect 1's appearance, such as a boundary between parts, the base of a wing, a pattern on the insect 1, or a position that can be distinguished based on a characteristic point within a part, such as the center of the front and rear ends of each part of the insect 1. The center of the front and rear ends of each part of the insect 1 is, for example, the center of the front and rear ends of the abdomen or the center of the front and rear ends of the torso. For example, in the case of the insect 1 shown in FIG. 6(a), the center of the widthwise direction of the boundary between the thorax 15 and the abdomen 16 is used as the characteristic point to specify the position of the midpoint 43. 6(b), the position of the midpoint 43 is specified with the base of the wings of the body 14 as the feature point. In the point information assignment step S3 of this embodiment, if the boundaries between parts of the body 11 of the insect 1 corresponding to the insect data D are recognizable, the boundaries between the parts are treated as feature points, and if the boundaries between parts of the body 11 of the insect 1 corresponding to the insect data D are not recognizable but characteristic points within the parts are recognizable, the characteristic points within the parts are treated as feature points.

[0040] The order information assigning step S4 is a step of assigning order information 5 relating to the order of the three pieces of point information 4 assigned in the point information assigning step S3. The order information 5 is information relating to the order assigned to the point information 4 in order to identify the posture of the insect 1 corresponding to the insect data D. Specifically, the order information 5 is information relating to the order in which the head-corresponding point 41, the midpoint 43, and the abdomen-corresponding point 42 are arranged in a predetermined order. The order information 5 is also information relating to the order in which the points are arranged along the entire length of the body main body 11 of the insect 1. In this embodiment, as shown in FIG. 6, the order information 5 is information relating to the order from the front end (head side) to the rear end (ventral side). That is, in the order information assigning step S4, the order information 5 is assigned to the point information 4 based on the order of the parts of the body main body 11 of the insect 1. In this embodiment, the order information 5 relates to the order in which the head-corresponding point 41, the midpoint 43, and the abdomen-corresponding point 42 are arranged. In addition, in the sequential information assignment step S4 of this embodiment, sequential information 5 is assigned to point information 4, so sequential information 5 is not assigned to insect data D that has not been assigned point information 4 in the point information assignment step S3 (insect data D that has been determined to be unable to be part-identified in the part extraction step S2).

[0041] As described above, in the method for generating image data for insect identification according to this embodiment, the pre-processed image data is subjected to the processes in the above steps to generate image data for insect identification.

[0042] Next, the insect identification system 6 will be described with reference to FIGS.

[0043] As shown in FIG. 7, the insect identification system 6 is a system that identifies insects 1 contained in image data transmitted from a client terminal C (in this embodiment, the imaging device 3) via the Internet I. Specifically, the insect identification system 6 processes target image data obtained by capturing an image of a captured insect specimen by a client, and identifies and outputs the insects 1 contained in the target image data. In this embodiment, the target image data is captured by the imaging device 3 described above, and is transmitted to the insect identification system 6 via communication means built into the imaging device 3.

[0044] As shown in FIG. 8, the insect identification system 6 includes a communication unit 61 that communicates with the outside, an image data acquisition unit 62 that acquires target image data from the outside, an insect identification unit 63 that identifies the type of insect 1 included in the target image data, and an output unit 64 that outputs the identification result of the insect identification unit 63. The image data acquisition unit 62 acquires the target image data from a client terminal C (imaging device 3) via the communication unit 61 and the Internet I. The output unit 64 outputs the identification result to the client terminal C (specifically, a client computer P or a tablet terminal, etc.) via the communication unit 61 and the Internet I. Furthermore, such an insect identification system 6 is realized by a computer. Specifically, the insect identification system 6 of this embodiment is realized by a computer configured to function as the communication unit 61 that communicates with the outside, the image data acquisition unit 62 that acquires target image data from the outside, the insect identification unit 63 that identifies the type of insect 1 included in the target image data, and the output unit 64 that outputs the identification result of the insect identification unit 63.

[0045] The insect identification unit 63 is configured to identify the type of insect 1 contained in the target image by a trained model 65 constructed by machine learning using training data. The trained model 65 is constructed by machine learning using the insect identification image data generated by the above-mentioned method for generating image data for insect identification as training data.

[0046] The training data of the trained model 65 includes unprocessed image data as training image data containing an insect 1 in the image, pose information including point information 4 assigned by specifying at least three points to insect data D related to the insect 1 included in the training image data, and order information 5 related to the order assigned to the point information 4, and type information related to the type of insect 1 included in the training image data. The point information 4 is the same as the point information 4 assigned in the point information assignment step S3, the order information 5 is the same as the order information 5 assigned in the order information assignment step S4, and the type information is the same as the type information assigned in the type information assignment step. Furthermore, the training data of this embodiment is data assigned with frame information related to a frame F surrounding the insect 1 included in the training image data. The frame information is the same as the frame information assigned in the frame information assignment step.

[0047] The trained model 65 is constructed by so-called supervised learning, in which the insect identification image data generated by the above-described method for generating insect identification image data is used as training data. In other words, the trained model 65 is a model constructed by providing a large number of insect identification image data with type information added to a machine learning algorithm.

[0048] Such an insect identification unit 63 processes the target image data, discovers insects 1 contained in the target image data, and identifies the types of the discovered insects 1. The insect identification unit 63 of this embodiment assigns type information of the discovered insects 1 to the target image data, and outputs the number of insects 1 for each type as the identification result. In this embodiment, the types of insects 1 are classified at the family level.

[0049] According to the method for generating image data for insect identification configured as described above, the point information 4 and order information 5 are assigned to the extracted insect 1, so that the posture of the insect 1 can be determined based on the point information 4 and order information 5. For example, even if the insect 1 is bent, the type of insect 1 can be appropriately identified. Therefore, the accuracy of identifying the insect 1 contained in the image data can be improved.

[0050] Furthermore, since frame information is added to the extracted insects 1, it is easy to properly distinguish the insects 1 even when multiple insects 1 overlap each other.

[0051] Furthermore, since point information 4 is specified for two or more parts and the point information 4 is assigned based on the order of the parts, the posture of the insect 1 can be determined appropriately when the insect 1 is bent.

[0052] Furthermore, since the posture of the insect 1 can be determined based on at least three points including the head corresponding point 41, the abdomen corresponding point 42, and the midpoint 43, the posture of the insect 1 can be determined appropriately even if the insect 1 is bent.

[0053] Furthermore, since the point information 4 is assigned to the characteristic point as the midpoint 43, the point information 4 can be assigned to approximately the same position on the insects 1 for the same type of insects 1.

[0054] Furthermore, the imaging device 3 irradiates light from both sides of the travel area, preventing the insect 1 from being in shadow, and also prevents the insect 1 from being blurred in the pre-processing image data, because the imaging device 3 connects multiple images to generate pre-processing image data.

[0055] Furthermore, since the method includes a type information assigning step of assigning type information regarding the type of insect 1, even if the insect 1 included in the image data is bent, the image data including the bent insect 1 can be used as training data.

[0056] The method also includes a part extraction step S2 for extracting parts of the insect 1 corresponding to the insect data D from the insect data D extracted before the point information assignment step S3, and the part extraction step S2 extracts at least the head 13 and abdomen 16 of the insect 1. Therefore, in the point information assignment step S3, point information 4 can be appropriately assigned to parts of the main body 11 of the insect 1.

[0057] The part extraction step S2 also includes a possibility determination step S22 for determining whether parts can be extracted from the extracted insect data D, and part extraction is performed only on the insect data D that is determined to be part extractable in the possibility determination step S22 among the extracted insect data D. In the point information assignment step S3, point information 4 is assigned only to the insect data D from which parts have been extracted. Therefore, the point information 4 can be reliably assigned to the parts of the insect 1 corresponding to the insect data D.

[0058] Furthermore, in the point information assigning step S3, at least three points on the body main body 11 of the insect 1 are designated and assigned point information 4. Therefore, the posture of the insect 1 can be reliably determined regardless of the state of the appendages 12, such as the appendages of the insect.

[0059] In the sequential information assigning step S4, the sequential information 5 is assigned in the order of the head corresponding point 41, the midpoint 43, and the abdomen corresponding point 42, or in the order of the abdomen corresponding point 42, the midpoint 43, and the head corresponding point 41. Therefore, the sequential information 5 is assigned along the entire length of the main body 11, so that the posture of the insect 1 can be reliably determined.

[0060] Furthermore, according to the insect identification system 6 configured as described above, the trained model 65 is constructed by machine learning using training data that includes point information 4 relating to at least three points specified for the insect data D contained in the training image data and posture information including order information 5, so that the type of insect 1 can be appropriately identified even if, for example, the insect 1 is bent.

[0061] The inventors compared the insect identification system 6 using the trained model 65 described above with a conventional insect identification system using a trained model constructed as training data in which only information about the type of insect is attached to images of the insect, without including information about the insect's posture, such as the point information 4 and the sequential information 5. When the same captured insect samples were used to identify insects 1, the identification accuracy of the conventional insect identification system was approximately 60.0%, while the identification accuracy of the insect identification system 6 using the trained model 65 of the present invention was approximately 79.1%. Furthermore, when insects 1 were densely packed in the pre-processing image, the conventional insect identification system sometimes failed to identify the insects 1, resulting in a drop in the accuracy of detecting insects 1 (specifically, from 90.0% or more to 80.0% or less). However, the insect identification system 6 using the trained model 65 described above was able to maintain accuracy of detecting insects 1 (generally 90.0% or more) even when the insects 1 were densely packed.

[0062] The above describes an embodiment of the present invention using an example, but the present invention is not limited to the above embodiment, and various modifications can be made within the scope that does not deviate from the gist of the present invention.

[0063] For example, the case where image data as training data is created using the method for generating image data for insect identification has been described, but the present method for generating image data for insect identification can also be used as preprocessing in the task of identifying insects 1 contained in image data by humans or machines. Specifically, the method for generating image data for insect identification according to the present invention can be applied to sample image data obtained by capturing an image of a captured sample 2, and the sample image data as pre-processing image data can be used as identification data as insect identification image data to be used in the identification task.

[0064] Furthermore, although the case where a person adds information to pre-processed image data to create insect identification image data has been described, the present invention is not limited to such a case, and for example, the method can be configured so that information is added to pre-processed image data by a machine in all or some of the steps. When the method is configured so that information is added to pre-processed image data by a machine in all steps, the method can be configured as an insect identification image generation method in which a computer executes an insect extraction step S1, a part extraction step S2, a point information addition step S3, and a sequence information addition step S4 to generate an insect identification image.

[0065] Furthermore, although the case where each step is performed for all insects 1 before proceeding to the next step has been described, this is not limiting, and it is also possible to configure it so that, for example, all steps are performed for one insect 1 before all steps are performed for the next insect 1. It is also possible to perform each step independently for multiple insects 1.

[0066] Furthermore, although the case where the insect data D extracted in the insect extraction process S1 includes a type information assignment process for assigning type information regarding the type of insect 1 has been described, the present invention is not limited to such a case. For example, when the method for generating image data for insect identification is used as a preprocessing step in the work of identifying insect 1, the method can be configured so that type information of insect 1 is not assigned.

[0067] Furthermore, in the insect extraction process S1, the frame information assigning process has been described as assigning frame information relating to a rectangular frame F surrounding the insect 1 in the image, but this configuration is not limited to this, and it is also possible to assign frame information relating to a frame other than a rectangular shape, such as a circle or a shape that follows the outline of the insect 1, or it is possible not to assign a frame F, i.e., not to include the frame information assigning process.

[0068] In addition, in the part extraction step S2, it is determined whether or not parts can be extracted for the insect 1 corresponding to the insect data D, and parts of the insect 1 are extracted only for the insect data D for which it is determined that parts can be extracted. However, the present invention is not limited to this configuration, and it can also be configured to extract parts of the insect 1 for all insect data D.

[0069] Furthermore, although the case where the part extraction step S2 is included has been described, the part extraction step S2 may not be included. When the part extraction step S2 is not included, for example, the point information assignment step S3 may assign point information 4 to three points on the insect data D: one end, the other end, and a position between both ends of the main body 11, and the sequential information assignment step S4 may assign sequential information 5 to the point information 4 in order from one end to the other end. In this case, the point information 4 is essentially assigned so that at least three points correspond to two or more parts of the insect 1. Note that, in the above-described configuration, for example, one end of the main body 11 may be determined as the location of the antennae of the insect 1. Also, in the case of an insect 1 with a small head 13 and a large body 14, three or more points of point information 4 may be assigned to the body 14, and sequential information 5 may be assigned, for example, starting from the head 13. Furthermore, although the case where the point information 4 is assigned only to the main body 11 has been described, the configuration is not limited to this. Point information 4 may also be assigned to antennae, legs, or wings as appendages 12. For example, one or more pieces of point information 4 may be assigned to each of the antennae tips, the main body 11, and the legs, and sequential information 5 may be assigned starting from the antennae. In either case, by assigning three or more pieces of point information 4 and sequential information 5 at intervals sufficient to allow estimation of the insect 1's posture, the insect 1's posture can be determined, improving the accuracy of identifying the insect 1. Note that when four or more pieces of point information 4 are assigned, the posture can be determined with high accuracy.

[0070] It is also possible to configure the insect data D to be assigned line information connecting the point information 4 according to the sequential information 5 assigned in the sequential information assigning step S4. When line information is assigned, the assigned line information makes it easier to determine the posture of the insect 1. It is also possible to assign information about the area surrounded by the lines connecting the points assigned in the point information assigning step S3 to the insect data D. When area information is assigned, for example, if the area surrounded by the three pieces of point information 4 assigned in the point information assigning step S3 is large, it can be determined that the insect 1 is bent significantly.

[0071] Furthermore, the image data acquisition unit 62 in the insect identification system 6 has been described as acquiring image data via the Internet I, but it is not limited to this configuration and can also be configured to acquire image data directly from a device that captures an image of insect paper, etc.

[0072] Furthermore, although the case where the target image data is captured using the same method as the teacher image data has been described, the present invention is not limited to this configuration, and the target image data can also be captured using a method different from that used for capturing the teacher image data.

[0073] Furthermore, although the above description has been given of the case where the "family" in the biological classification hierarchy is used to indicate the type of insect 1, it is not limited to this case; "order," "genus," "species," etc. can also be used, and classifications not based on classification hierarchy (for example, classification based on appearance such as "crane flies") can also be adopted.

[0074] Next, the pest control system 7 will be described with reference to Figures 9 to 16. The pest control system 7 is a system that performs processing to suppress the adverse effects of pests when there is an increase in pests or when there is a high possibility that pests will increase, and in this embodiment, it is a system for controlling pests inside a building (e.g., a food factory) and / or pests that invade the building. The pest control system 7 of this embodiment uses the insect identification system 6 described above.

[0075] The pest control system 7 executes response control (feedback control) to reduce the number of insects in the building and / or insects invading the building in response to an increase in insects, and preventive control (feedforward control) to prevent an increase in insects when there is a possibility that the number of insects in the building and / or insects invading the building will increase. First, the response control will be described.

[0076] As shown in FIG. 9, the pest control system 7, which performs countermeasure control, includes a control unit 71 that receives and processes information related to the identification results from the insect identification system 6, and a storage unit 72 that stores various information. The control unit 71 outputs the processing results to the processing unit 8. The insect identification system 6, which outputs information related to the identification results to the pest control system 7 of this embodiment, outputs the number of each type of insect contained in target image data as an identification result to the control unit 71 for target image data generated (specifically, captured and image processed) by an imaging device 3 installed inside and outside the building and configured for fixed-point observation. The imaging device 3 captures images of insects trapped in multiple insect traps (e.g., insect paper) installed inside and outside the building to generate target image data. That is, the insect identification system 6 of this embodiment outputs an identification result related to the number of each type of insect trapped in the insect traps. Furthermore, in this embodiment, the insect identification system 6 is configured to output the identification results at predetermined time intervals.

[0077] The pest control system 7 includes a control unit 71 that executes predetermined processing, and a storage unit 72 that stores information about the identification results received from the insect identification system 6. The pest control system 7 of this embodiment is a system configured to identify the type of insect whose number has increased by more than a threshold based on the information about the identification results received from the insect identification system 6, and to output a processing method D4 corresponding to the identified type of insect. The pest control system 7 of this embodiment is also realized by cloud computing, but is not limited to this.

[0078] The storage means 72 stores at least the previously received identification result from the insect identification system 6. The storage means 72 is also capable of transmitting the previously received identification result that it has stored to the control unit 71. The storage means 72 also stores a database (for example, table T1 shown in FIG. 13) that the control unit 71 uses for various determinations, and transmits it to the control unit 71 in response to a request from the control unit 71. As shown in FIG. 13, the database stored in the storage means 72 of this embodiment is a database in which the type of insect (type information) is associated with intrusion route information D2, source information D3, and information on a processing method D4.

[0079] The control unit 71 is equipped with an increase determination unit 711 that compares the identification result received from the insect identification system 6 with the previously received identification result stored in the memory means 72 to determine whether the number of insects of each type has increased by more than a threshold value, an occurrence source determination unit 712 that determines, for the type of insect that the increase determination unit 711 determines has increased by more than the threshold value, entry route information D2 regarding the entry route of that type of insect and occurrence source information D3 regarding the occurrence source, and a processing method determination unit 713 that determines a processing method D4 for reducing insects in the building based on the determination of the occurrence source determination unit 712.

[0080] While receiving identification results from the insect identification system 6, the increase determination unit 711 also receives previously received identification results from the storage unit 72. The increase determination unit 711 compares the previously received identification results with the currently received identification results to determine whether the number of insects has increased by a threshold or more for each type of insect. That is, the increase determination unit 711 of this embodiment determines, for each type of insect, whether the number of insects newly captured by the insect trapping device between the time when the insect identification system 6 output the previous identification results and the time when the insect trapping device output the current identification results is equal to or greater than a threshold. The increase determination unit 711 can determine, for each type of insect, whether the number of insects newly captured by the insect trapping device is equal to or greater than a threshold during a predetermined time period (the interval between the output of the identification results by the insect identification system 6). Therefore, if a particular type of insect is abundant in the environment where the insect trapping device is installed, it can determine that the insects of that type are abundant. Note that the threshold in this embodiment is a preset value, but the present invention is not limited to this configuration. For example, the threshold can be set to a value twice the normal increase amount.

[0081] The infestation source determination unit 712 determines infestation source information D3 relating to the infestation source of each type of insect for which the increase determination unit 711 has determined that the number of insects has increased by more than a threshold. That is, when there are many insects of a particular type in the environment where the insect trapping device is installed, the infestation source determination unit 712 identifies the infestation source of that type of insect. The infestation source determination unit 712 of this embodiment determines infestation route information D2 relating to the infestation route of the type of insect determined to have increased and infestation source information D3 relating to the insect infestation source. The infestation route information D2 is information that distinguishes whether the type of insect is an external infestation type that occurs outside the building and infestation into the building, or an internal infestation type that occurs inside the building. The infestation source determination unit 712 of this embodiment identifies the infestation route information D2 and infestation source information D3 of the type of insect that has increased based on a pre-prepared database of infestation route information D2 and infestation source information D3 for each type of insect.

[0082] As shown in FIG. 13, the infestation source information D3 is information for specifying the source of that type of insect within a building, and for externally invading insects, it is information for specifying how that insect entered the building, and for internally occurring insects, it is information for specifying the environment inside the building in which the insects will be generated. Specifically, for externally invading insects, the infestation source information D3 includes insects that fly and invade the building, such as large flying flies belonging to large flies (for example, house flies, flesh flies, blow flies, etc.), and other flying insects that fly and invade the building other than large flies (for example, insects of the Diptera order other than large flies, thrips, moths, caddisflies, fire ants, wasps, planthoppers, etc.). This information indicates which of the four types of pests they belong to: crawling or creeping pests (which include insects such as ants, earwigs, crickets, springtails, centipedes, millipedes, house centipedes, woodlice, mites, spiders, ground beetles, and stink bugs); or other pests that do not fall into any of the above categories (such as rodents). Furthermore, the source information D3 includes, for internally occurring insects, drainage and water-related insects that originate from drainage and other water-related systems (for example, house flies, fruit flies, phorid flies, false house flies, sphaeroides, and Japanese black flies), fungivorous insects that originate from fungi (for example, winged booklice, booklice, stag beetles, and rove beetles), and indoor dust and food-related insects that originate from indoor dust and food (for example, booklice, silverfish, This information indicates whether the insect belongs to one of six types: insects that feed on insects (such as dermestids, acarid mites, beetles, flathead beetles, horn beetles, and moths); insect predators that emerge by feeding on insects (such as spiders of the families Pholcidae, Pholcidae, and Salticidae); heat source / gap insects that emerge from heat sources or gaps (such as cockroaches); or internally occurring insects that do not belong to any of the above types (such as bedbugs).

[0083] The source determination unit 712 of this embodiment references table T1 as shown in FIG. 13 to determine intrusion route information D2 and source information D3 for the type determined by the increase determination unit 711 to have increased. Specifically, the source determination unit 712 references type information D1 in table T1 to obtain intrusion route information D2 and source information D3 corresponding to the type determined to have increased. Specifically, if the type determined to have increased is a1, a2, or a3, the source determination unit 712 determines that the increasing insects are external invasive insects and that the source is a large flying fly. Furthermore, if the number of insects that are increasing is the same type related to the source information D3, the source determination unit 712 outputs only one type as the type of the increasing insects. For example, if the type determined to have increased is more than one of a4, a5, a6, a7, and a8, the source determination unit 712 determines that the increasing insects are external invasive insects and that the source is a flying or other type.

[0084] The treatment method determination unit 713 determines and outputs a treatment method D4 for reducing insects based on the infestation source information D3 determined by the infestation source determination unit 712. In this embodiment, the treatment method determination unit 713 outputs both a physical control D41 treatment method D4 and a chemical control D42 (chemical treatment) treatment method D4 as the treatment methods D4 for reducing insects. Furthermore, the treatment method determination unit 713 of this embodiment determines the treatment method D4 by referring to a database related to treatment methods D4, specifically, by referring to table T1 as shown in FIG. 13 . For example, if the infestation source determination unit 712 determines that the type of increasing infestation source is large flying flies, the physical control D41 in the treatment method D4 corresponding to the type of large flying flies outputs instructions such as checking for dead animals that may be infestation sources, installing insect nets, checking openings, and warnings such as shutter open time, and the chemical control D42 outputs spatial spraying of chemicals at the infestation site. Furthermore, when the infestation source determination unit 712 determines that the number of insects of the external intrusion type and the infestation source information D3 indicates an increase in other types of insects, the treatment method determination unit 713 outputs confirmation of paths that could become external intrusion holes as physical control D41 in the treatment method D4 corresponding to the other types. In this way, the treatment method determination unit 713 outputs the treatment method D4 for each type in the infestation source information D3 by dividing it into physical control D41 and chemical control D42.

[0085] Furthermore, for chemical control D42, the treatment method determination unit 713 determines the type and amount of chemical to be used, the chemical treatment method, and the treatment location, and determines whether or not to immediately perform chemical treatment using the determined method. The specific processing flow will be described later.

[0086] The processing unit 8 is configured to perform a predetermined process in accordance with the output of the treatment method determination unit 713. The processing unit 8 of this embodiment includes a notification means 81 configured to notify a manager of the building to be controlled of the treatment method D4 related to the physical control D41 and the chemical control D42 output by the treatment method determination unit 713, and an agent treatment means 82 that executes the agent treatment related to the chemical control D42 output by the treatment method determination unit 713.

[0087] The processing flow in the pest control system 7 configured as above will be described with reference to FIGS.

[0088] As shown in Figure 10, the pest control system 7 repeatedly executes a current result acquisition process S51, a previous result acquisition process S52, an increase determination process S53, an occurrence source determination process S54, a treatment method determination process S55, and a treatment output process S56.

[0089] The current result acquisition step S51 is a step of acquiring the identification results output by the insect identification system 6 and acquiring the number of insects of each type included in the identification results. The current result acquisition step S51 of this embodiment is executed by the increase determination unit 711, and the insect identification system 6 outputs the identification results at predetermined time intervals, thereby enabling the identification results to be acquired at predetermined time intervals. Furthermore, in the current result acquisition step S51, the acquired identification results are stored in the storage means 72.

[0090] The previous result acquisition step S52 is a step of acquiring the identification results previously output by the insect identification system 6 and acquiring the number of insects for each type in the previous identification results included in the identification results. The previous result acquisition step S52 in this embodiment is executed by the increase determination unit 711, which acquires the previous identification results from the storage means 72.

[0091] The increase determination step S53 is a step in which the number of insects for each type included in the identification result currently output by the insect identification system 6 is compared with the number of insects for each type included in the identification result previously output, and whether or not the number of insects for each type has increased by more than a threshold is determined for each type, and the types for which the number of insects has increased by more than a threshold are output. If it is determined in the increase determination step S53 that there is no type for which the number of insects has increased by more than a threshold (NO in the increase determination step S53), the process proceeds to the current result acquisition step S51, and if it is determined that there is one or more types for which the number of insects has increased by more than a threshold (YES in the increase determination step S53), the process proceeds to the infestation source determination step S54. Furthermore, the increase determination step S53 in this embodiment is executed by the increase determination unit 711.

[0092] The infestation source determination step S54 is a step of determining the intrusion route information D2 and the infestation source information D3 for the insects determined to have increased in number by more than the threshold in the increase determination step S53. Specifically, the infestation source determination step S54 is a step of referring to the database (table T1 shown in FIG. 13 in this embodiment) for the types of insects determined to have increased in number by more than the threshold in the increase determination step S53, and referring to the infestation route information D2 and the infestation source information D3 to identify the infestation source of the type that is increasing. For example, in the increase determination step S53, if one or more types of a1, a2, and a3 are increasing, it is determined that the type of externally invasive large flying flies is increasing; if one or more types of a4, a5, a6, a7, and a8 are increasing, it is determined that the type of externally invasive flying and other types is increasing; if one or more types of b1, b2, b3, b4, and b5 are increasing, it is determined that the type of externally invasive wandering / creeping flies is increasing; and if c1 or c2 are increasing, it is determined that the type of externally invasive other types is increasing. In the source determination step S54, if one or more of the types d1, d2, d3, d4, and d5 are increasing, it is determined that the internally generated drainage and water-related types are increasing; if one or more of the types e1, e2, e3, e4, and e5 are increasing, it is determined that the internally generated bacterium-eating types are increasing; if one or more of the types f1, f2, f3, f4, and f5 are increasing, it is determined that the internally generated indoor dust and food types are increasing; if one or more of the types g1, g2, and g3 are increasing, it is determined that the internally generated insect-eating types are increasing; if h1 is increasing, it is determined that the internally generated heat source and gap types are increasing; and if i1 is increasing, it is determined that the internally generated other types are increasing. In the source determination step S54, the source of all types determined to have increased by more than the threshold value in the increase determination step S53 is determined. The source determination step S54 is executed by the source determination unit 712.

[0093] The treatment method determination step S55 is a step of determining a treatment method D4 corresponding to the source of the increasing type of insect. Specifically, this is a step of determining a treatment to reduce the increasing number of insects based on the infestation route information D2 and infestation source information D3 determined in the infestation source determination step S54. In this embodiment, the treatment method determination step S55 outputs the treatment method D4 corresponding to the infestation source information D3 by referring to a database (e.g., table T1 shown in FIG. 13). Furthermore, in the treatment method determination step S55, physical control D41 and chemical control D42 are output separately for the treatment method D4. For example, in the treatment method determination step S55, if the type of large flying flies is increasing, physical control D41 includes instructions to check for dead animals that may be the source of infestation, install insect netting, check openings, and take precautions such as shutter open time, and chemical control D42 includes instructions to spray chemicals into the air at the infestation site and / or spray chemicals on the source of infestation. If the number of flying or other pests is increasing, physical control D41 includes instructions to install insect nets, check openings, and provide cautions such as shutter open times, as well as instructions to place attracting lights around the building and switch LED lights. Chemical control D42 includes instructions to spray chemicals in the air at the infestation site and / or spray chemicals at the source. Furthermore, if the number of wandering or creeping pests is increasing, physical control D41 includes installing sticky traps, and chemical control D42 includes spraying chemicals around the building and preventive chemical treatment of potential infestation sites. If the number of external infestations or other pests is increasing, physical control D41 includes checking for potential external entry points. Furthermore, in the treatment method determination step S55, if the infestation route information D2 indicates an increase in internal infestations, physical control D41 includes instructions to clean the location corresponding to the source information D3, and chemical control D42 includes chemical treatment to eradicate the infestation and preventive chemical treatment of the source.Specifically, in the treatment method determination step S55, as physical control D41, if the number of drainage and water-related systems is increasing, removal of sources such as drainage and water-related systems is output; if the number of bacterivorous systems is increasing, cleaning and removal of mold and other potential sources of infestation is output; if the number of indoor dust and food-related systems is increasing, cleaning and removal of dust and other potential sources of infestation is output; if the number of insect-eating systems is increasing, checking for the occurrence of insects that could be an infestation or attractant; and if the number of heat source and gap systems is increasing, cleaning and removal of food residues and other potential sources of infestation is output. Also, as chemical control D42, if the number of drainage and water-related systems, bacterivorous systems, indoor dust and food systems, or heat source and gap systems is increasing, chemical treatment to exterminate the infestation insects and prevent the occurrence of the infestation source are output; and if the number of insect-eating systems is increasing, chemical treatment to exterminate the infestation insects and chemical treatment to prevent the occurrence of attractant and infestation insects are output. Furthermore, if other types of internal infestation (specifically, bed bugs) are increasing, a bed bug detection notification is issued as physical control D41, and chemical treatment for exterminating infested insects and preventive chemical treatment are output as chemical control D42. In addition, the treatment method determination step S55 is executed by the treatment method determination unit 713.

[0094] As described above, in the treatment method determination step S55, if the number of externally invading insects is increasing, excluding other types, a method of preventing the invading of the increasing insects is output as physical control D41, and chemical treatment of at least one of the invading point and the source of infestation is output as chemical control D42. Also, if the number of internally infested insects is increasing, excluding other types, cleaning and removal of the source of infestation is output as physical control D41, and chemical treatment to exterminate the infested insects and preventive chemical treatment of the source of infestation are output as chemical control D42.

[0095] Furthermore, in the treatment method determination step S55, for chemical treatment as chemical control D42, the type and amount of chemical to be used, the chemical treatment method, and the location of treatment are determined. Specifically, as shown in Fig. 11, in the chemical treatment, a treatment method determination step S57 is executed to determine which type of chemical to use and how to perform the treatment, a treatment amount determination step S58 is executed to determine the amount of chemical to be used in the treatment, and an implementation feasibility determination step S59 is executed to determine whether to immediately perform the chemical treatment.

[0096] In the treatment method determination step S57, it is determined that the pesticide selected corresponding to the infestation source is to be sprayed on the infestation source and the location where the insects are detected. For example, the method of pesticide treatment corresponding to the infestation source information D3 is determined based on a database (not shown).

[0097] In the treatment amount determination step S58, the amount of the pesticide to be used for treatment selected in the treatment method determination step S57 is determined. The amount of pesticide to be used for treatment can be determined depending on the pesticide treatment method and the number of insects that have emerged. For example, the amount of pesticide to be used can be determined by deriving the treatment amount corresponding to the treatment method D4 and the number of insects based on a database (not shown).

[0098] The implementation feasibility determination process S59 is a process for determining whether or not to perform chemical treatment. Specifically, the implementation feasibility determination process S59 determines whether or not to perform chemical treatment depending on the situation at the location where chemical treatment is to be performed. In this embodiment, the implementation feasibility determination process S59, as shown in FIG. 12, includes a human presence determination process S591 for determining whether or not there is a person at the location where chemical treatment is to be performed, a time determination process S592 for determining whether a predetermined time has passed since the previous chemical treatment at the location where chemical treatment is to be performed, and a treatment available date and time determination process S593 for determining whether or not chemical spraying is permitted. The human presence determination process S591 acquires information, for example, from a human presence sensor or camera installed at the location where chemical treatment is to be performed, and determines whether or not there is a person at the location where chemical treatment is to be performed. Furthermore, the treatment available date and time determination process S593 is a process for determining whether or not a predetermined date and time is available for chemical treatment. For example, the available date and time for chemical treatment is set based on the time or day of the week. In this embodiment, if it is determined in the person determination step S591 that there is no person (YES in the person determination step S591), and if it is determined in the time determination step S592 that a predetermined time has passed since the previous drug treatment (YES in the time determination step S592), and if it is determined in the processable date and time determination step S593 that the date and time is available for treatment (YES in the processable date and time determination step S593), it is determined that drug treatment can be performed, and the processing unit 8 is instructed to perform drug treatment (drug treatment execution instruction step S594). Also, if it is determined in the person determination step S591 that there is a person (NO in the person determination step S591), or if it is determined in the time determination step S592 that a predetermined time has not passed since the previous drug treatment (NO in the time determination step S592), and if it is determined in the processable date and time determination step S593 that the date and time is not available for treatment (NO in the processable date and time determination step S593), it is determined that drug treatment will not be performed immediately, and skip processing S595 is executed. In the skip processing S595, the notification means 81 of the processing unit 8 notifies the administrator or the like of the need for drug processing, and prompts the administrator to manually perform the drug processing or to specify the date and time for drug processing.

[0099] In the treatment output step S56, various instructions are given to the processing unit 8 based on the content output in the treatment method determination step S55. Specifically, if an output related to physical control D41 is given in the treatment method determination step S55, the output content is notified to the building manager or the like by the notification means 81, and the manager or the like is prompted to carry out physical control D41. Furthermore, if an output related to chemical control D42 is given, the chemical treatment means 82 is controlled to carry out chemical treatment, or skip processing S595 is executed. Furthermore, when a notification is given by the notification means 81, it can also be notified that the notification is for response control.

[0100] As described above, according to the pest control system 7 of this embodiment, it is determined whether there are any species of insects whose number has increased by more than a threshold based on the identification results output by the insect identification system 6, and if there are any species of insects whose number has increased by more than the threshold, a processing method D4 corresponding to the species whose number has increased is output, thereby suppressing the adverse effects of insects.

[0101] Furthermore, for the increased types, the source information D3 is determined and a processing method D4 corresponding to the source information D3 is output, so that, for example, when multiple types of insects are emerging from the same source, it is possible to prevent duplicate processing methods D4 from being output.

[0102] Although the pest control system 7 that performs response control has been described using one example, the system is not limited to this configuration.

[0103] For example, although the case where the identification results output by the insect identification system 6 are used as the identification results has been described, the present invention is not limited to such a configuration, and it is also possible to configure it to use, for example, identification results output by a conventional insect identification system 6 or identification results determined by a person. Furthermore, it is not limited to the case where the identification results for insects captured by an insect trap are used, and it is also possible to employ an insect identification system 6 that detects insects moving within a building, identifies the type of insect, and outputs identification results for each type (for example, an insect identification system 6 that outputs the type and number of insects as identification results based on an odor sensor).

[0104] Furthermore, the case where the pest control system 7 determines and outputs the treatment method D4 has been described, but the configuration is not limited to this, and the system can also be configured to output, for example, intrusion route information D2 and source information D3 and notify a building manager, etc. With this configuration, the building manager, etc. can perform the work required for pest control based on the notified intrusion route information D2 and source information D3.

[0105] Furthermore, although the case where the generation source determination unit 712 determines the generation source information D3 and the processing method determination unit 713 outputs the processing method D4 corresponding to the generation source information D3 has been described, the configuration is not limited to this. For example, the processing method determination unit 713 can be configured to output the processing method D4 corresponding to the type of the increase in the number output by the increase determination unit 711, rather than based on the generation source information D3.

[0106] Furthermore, although the case where the treatment method determination unit 713 outputs both physical control D41 and chemical control D42 as control methods has been described, it may also be configured to output only one of physical control D41 and chemical control D42. When such a configuration is adopted, the treatment method determination unit 713 may also be configured to select and output an appropriate treatment method D4 from physical control D41 and chemical control D42 based on information such as the number of insects.

[0107] Furthermore, the processing method D4 of the physical control D41 has been described in the case where the notification means 81 notifies the building manager or the like, but the configuration is not limited to this, and the physical control D41 can also be configured to be executed automatically. For example, the processing unit 8 can be configured to close the opening when a confirmation of the opening is output.

[0108] In addition, in the increase determination step S53, a case where the determination is made using the previous identification result has been described, but this configuration is not limiting, and it is also possible to determine, for example, for each type of insect, whether the current number has increased by more than a threshold value compared to the average number of insects in the past few identification results.

[0109] Furthermore, although the processing method D4 has been described as being determined based on source information D3, it is not limited to this configuration, and can also be configured to be determined based on type information D1, or based on intrusion route information D2, for example.

[0110] Next, the pest control system 7 that performs preventive control will be described with reference to FIGS.

[0111] As shown in Figure 14, the pest control system 7 that performs preventive control includes a control unit 71 that determines whether there is a risk of insect increase based on the identification results received from the insect identification system 6 and environmental information related to the environment inside and outside the building received from a factor sensor unit 9 that monitors the environment inside and outside the building, and outputs a preventive method for preventing the increase of insects if there is a risk of insect increase, and a storage unit 72 that stores databases (e.g., tables T1 and T2 shown in Figures 13 and 16) necessary for various judgments by the control unit 71 and environmental information received from the factor sensor. The configurations of the insect identification system 6 and the processing unit 8 are the same as those in the pest control system 7 that performs countermeasure control.

[0112] The factor sensor unit 9 is a sensor that monitors the environment inside and outside the building, specifically, a sensor that monitors at least humidity and temperature. The factor sensor unit 9 of this embodiment is provided to monitor the environment of locations within the building where internally-spawning insects may occur (specifically, locations indicated by the occurrence source information D3 for internally-spawning insects shown in FIG. 13) and the environment outside the building (specifically, the surroundings of the building). In addition, the factor sensor unit 9 outputs environmental information regarding the environment obtained by monitoring to the pest control system 7.

[0113] The storage means 72 stores the environmental information received from the factor sensor unit 9 for a predetermined period (e.g., one year). The storage means 72 also stores databases (e.g., tables T1 and T2 shown in FIGS. 13 and 16) used by the control unit 71 for various determinations, and transmits them to the control unit 71 in response to a request from the control unit 71. Specifically, the storage means 72 stores a database in which insect species and factors causing their increase (prevention conditions D7) are associated with each other, as shown in FIG. 16, and a database in which insect species (species information D1) are associated with intrusion route information D2, infestation source information D3, and information on treatment methods D4, as shown in FIG. 13. In the database shown in FIG. 16, the development zero point D5 and the effective accumulated temperature D6 required for infestation are associated with each insect species.

[0114] The control unit 71 includes an insect species extraction unit 714 that extracts insect species that exist in numbers greater than or equal to a threshold from the identification results received from the insect identification system 6; a factor extraction unit 715 that extracts factors (prevention condition D7) that will increase the existing species extracted by the insect species extraction unit 714; a condition determination unit 716 that determines whether the factors for increase (prevention condition D7) extracted by the factor extraction unit 715 are satisfied; and a prevention method determination unit 717 that determines a method for preventing an increase in species that the condition determination unit 716 determines to satisfy the factors for increase (prevention condition D7), and outputs the method to the processing unit 8.

[0115] The insect species extraction unit 714 extracts species whose numbers are equal to or greater than a threshold value by referring to the number of each type of insect output as an identification result from the insect identification system 6. The insect species extraction unit 714 of this embodiment is configured to extract species whose numbers are equal to or greater than one from the identification result.

[0116] The factor extraction unit 715 extracts factors (preventive conditions D7) that cause an increase in the type of insect extracted by the insect species extraction unit 714. Specifically, the factor extraction unit 715 refers to a database (specifically, table T2 shown in FIG. 16) stored in the storage means 72 to extract factors (preventive conditions D7) that cause an increase in the type of insect extracted by the insect species extraction unit 714. The factor extraction unit 715 extracts conditions recorded as preventive conditions D7 in table T2 shown in FIG. 16 as factors, and specifically extracts one or more conditions from among the effective accumulated temperature D6 (a value obtained by accumulating temperatures exceeding the development zero point D5 over a period of days; in this embodiment, daily degrees), average temperature (average daily temperature in this embodiment), and relative humidity as factors. For example, the factor extraction unit 715 extracts, for type a1, an increasing factor that the effective accumulated temperature is higher than 250 degrees (prevention condition D7), for type a4, an increasing factor that the effective accumulated temperature is higher than 1300 degrees or an average daily temperature is higher than 25 degrees (prevention condition D7), for type d2, an increasing factor that the effective accumulated temperature is higher than 300 degrees and an average daily temperature is lower than 20 degrees (prevention condition D7), and for type f1, an increasing factor that the effective accumulated temperature is higher than 50 degrees and a relative humidity is lower than 90% (prevention condition D7). Furthermore, the factor extraction unit 715 extracts information for determining whether each of the above factors is satisfied (specifically, the growth zero point D5) from the database.

[0117] The condition determination unit 716 determines whether factors causing an increase in insects (prevention conditions D7) are satisfied for each type of insect, based on the environmental information output by the factor sensor unit 9 and the factors extracted by the factor extraction unit 715. Specifically, the condition determination unit 716 derives the average temperature, relative humidity, effective accumulated temperature, etc., based on the environmental information output by the factor sensor unit 9 and stored in the storage means 72, and information for determining whether each factor extracted by the factor extraction unit 715 is satisfied (e.g., growth zero point D5), and determines whether the factors extracted by the factor extraction unit 715 are satisfied based on the derived results. Furthermore, the condition determination unit 716 determines whether factors are satisfied for externally invading insects based on environmental information outside the building, and determines whether factors are satisfied for internally occurring insects based on environmental information inside the building (more specifically, environmental information of places within the building where internally occurring insects may be occurring).

[0118] The prevention method determination unit 717 determines a method for preventing an increase in the number of insects for a type determined by the condition determination unit 716 to satisfy the factors. The prevention method determination unit 717 determines a prevention method for preventing an increase in the number of insects for a type determined by the condition determination unit 716 to satisfy the factors, with reference to table T1 as shown in FIG. 13 . The prevention method determination unit 717 of this embodiment determines the intrusion route information D2 and the infestation source information D3 corresponding to the type determined to satisfy the factors, and specifically, with reference to the type information D1 in table T1, acquires the intrusion route information D2 and the infestation source information D3 corresponding to the type determined to have increased. Furthermore, the prevention method determination unit 717 determines a prevention method for preventing an increase in the number of insects based on the acquired infestation source information D3, and outputs the determined method to the processing unit 8.

[0119] The prevention method determination unit 717 of this embodiment outputs, as a prevention method, a treatment method D4 similar to the treatment method D4 output by the above-mentioned treatment method determination unit 713. Note that this configuration is not limited to this, and it is also possible to configure it so that different methods are output as the treatment method D4 and the prevention method. For example, as a prevention method, it is possible to output only physical control D41 without outputting chemical control D42, or to output adjustment of temperature and humidity (environment), or it is possible to output, as chemical control D42, chemical treatment using a slow-acting chemical that has less repellency than the chemical used in the treatment method D4.

[0120] The flow of processing in the pest control system 7 as described above will be described with reference to FIG.

[0121] As shown in Figure 15, the pest control system 7 repeatedly executes an identification result acquisition process S61, a type information extraction process S62, an increase factor extraction process S63, a factor information acquisition process S64, a factor determination process S65, a source information acquisition process S66, a prevention method acquisition process S67, and a processing output process S56.

[0122] The identification result acquisition step S61 is a step of acquiring the identification result output by the insect identification system 6. The identification result acquisition step S61 of this embodiment is executed at predetermined time intervals. The identification result acquisition step S61 of this embodiment is executed by the insect identification system 6 outputting the identification result at predetermined time intervals. The identification result acquisition step S61 is executed by the insect species extraction unit 714.

[0123] The type information extraction step S62 is a step of extracting types of insects whose number is equal to or exceeds a threshold value from the identification results acquired in the identification result acquisition step S61. In this embodiment, this is a step of extracting types of insects whose number is equal to or exceeds a threshold value from the identification results. The type information extraction step S62 is executed by the insect type extraction unit 714.

[0124] The increase factor extraction step S63 is a step of extracting from a database an increase factor (prevention condition D7) for the type extracted in the type information extraction step S62. Specifically, the increase factor extraction step S63 refers to the database stored in the storage means 72, and extracts the increase factor (prevention condition D7) of the type of insect extracted in the type information extraction step S62 and information (specifically, growth zero point D5) for determining whether the increase factor (prevention condition D7) is satisfied. The increase factor extraction step S63 is executed by the factor extraction unit 715.

[0125] The factor information acquisition step S64 is a step of acquiring environmental information related to the factors (preventive condition D7) of insect increase extracted in the factor increase extraction step S63. Specifically, the factor information acquisition step S64 is a step of acquiring environmental information stored in the storage means 72, and deriving factor information such as average temperature, relative humidity, and effective accumulated temperature based on the acquired environmental information and information for determining whether the factors of increase (preventive condition D7) extracted in the factor increase extraction step S63 are satisfied. The factor information acquisition step S64 is executed by the factor extraction unit 715.

[0126] The factor determination step S65 is a step of determining whether the factor information derived in the factor information acquisition step S64 satisfies the factor of insect increase (preventive condition D7) extracted in the increase factor extraction step S63. If it is determined in the factor determination step S65 that the factor information does not satisfy the preventive condition D7 for all types extracted in the type information extraction step S62 (NO in the factor determination step S65), the process proceeds to the identification result acquisition step S61. If it is determined in the factor determination step S65 that the factor information satisfies the preventive condition D7 for at least one type extracted in the type information extraction step S62 (YES in the factor determination step S65), the process proceeds to the infestation source information acquisition step S66. The factor determination step S65 is executed by the condition determination unit 716.

[0127] The infestation source information acquisition step S66 is a step of identifying infestation source information D3 and intrusion route information D2 for a type of insect that satisfies the cause of insect increase (prevention condition D7). Specifically, the infestation source information acquisition step S66 is a step of referring to a database (table T1 shown in FIG. 13 in this embodiment) for a type of insect that has been determined to satisfy the condition of insect increase (prevention condition D7), and referring to the infestation route information D2 and the infestation source information D3 to identify the source of the increasing type of insect. The specific processing in the infestation source information acquisition step S66 is the same as the infestation source determination step S54 in the pest control system 7 that executes response control. The infestation source information acquisition step S66 is executed by the prevention method determination unit 717.

[0128] The prevention method acquisition step S67 is a step of determining a treatment method D4 corresponding to the infestation source of the type of insect determined to satisfy the factors for increase. Specifically, it is a step of determining a treatment to suppress the increase of the increasing number of insects based on the infestation route information D2 and the infestation source information D3 determined in the infestation source information acquisition step S66. In the treatment method determination step S55 of this embodiment, the treatment method D4 corresponding to the infestation source information D3 is output by referring to a database (for example, table T1 shown in FIG. 13). The prevention method acquisition step S67 of this embodiment is similar to the treatment method determination step S55 in the pest control system 7 that executes countermeasure control. In other words, the prevention method acquisition step S67 includes a step of determining the specific method of chemical treatment in chemical pest control D42 and whether or not to implement chemical treatment.

[0129] The processing output step S68 issues various instructions to the processing unit 8 based on the content output in the prevention method acquisition step S67. The processing output step S68 of this embodiment is similar to the processing output step S56 in the pest control system 7 that executes response control. Specifically, when output related to physical pest control D41 is issued in the prevention method acquisition step S67, the output content is notified to the building manager or the like by the notification means 81, and the manager or the like is prompted to execute physical pest control D41. Furthermore, when output related to chemical pest control D42 is issued, the chemical treatment means 82 is controlled to execute chemical treatment, or skip processing S595 is executed.

[0130] According to the pest control system 7 configured as described above, for a species of insect that exists in numbers above a threshold, it is determined whether or not the factors that would cause an increase in that species of insect are met, and if the factors that would cause an increase are met, preventive treatment can be carried out, thereby suppressing the increase in insects.

[0131] Furthermore, for species that satisfy the factors for increase, the source information D3 is determined and a processing method D4 corresponding to the source information D3 is output. Therefore, for example, when factors for the increase of multiple types of insects from the same source are observed, it is possible to prevent the output of duplicate processing methods D4.

[0132] In the above-mentioned pest control system 7, we have described the case where environmental information is acquired from the factor sensor unit 9, but this configuration is not limited to this. For example, environmental information regarding the environment outside the building can also be acquired from weather information, etc., without using a sensor.

[0133] Furthermore, although the factors that cause an increase in insects have been explained using the effective accumulated temperature D6, average temperature, and relative humidity as examples, the present invention is not limited to such a configuration and can also use information regarding, for example, the season, weather, and hours of sunlight.

[0134] Furthermore, although the pest control system 7 that performs response control and the pest control system 7 that performs preventive control have been described separately, the configuration is not limited to this, and one system can also be configured to perform both response control and preventive control. In such a configuration, for example, when the insect identification system 6 outputs an identification result, both response control and preventive control can be performed, and both the processing required for response and the processing required for prevention can be output to the processing unit 8.

[0135] The above embodiment will be summarized below.

[0136] The method for generating image data for insect identification in the above embodiment is a method for generating image data for insect identification by processing pre-processed image data containing insects in the image to generate image data for insect identification, and includes an insect extraction step of extracting insect data related to insects from the pre-processed image data, a point information assignment step of assigning point information by specifying at least three points for the extracted insect data, and an order information assignment step of assigning order information to the point information of the at least three points specified by the point information assignment step.

[0137] With this configuration, point information and order information are assigned to the extracted insects, so that the posture of the insects can be determined based on the point information and order information, and even if the insects are bent, the type of insect can be appropriately identified, thereby improving the accuracy of identifying insects contained in image data.

[0138] The insect extraction step may also include a frame information assignment step of assigning frame information relating to a frame surrounding an insect included in the image, and the insect data may be extracted based on the frame information assigned in the frame information assignment step.

[0139] With this configuration, the extracted insects are assigned frame information, making it easier to properly identify insects even when multiple insects overlap, thereby improving the accuracy of identifying insects included in image data.

[0140] The point information assigning step is a step of assigning point information so that at least three points correspond to two or more of the multiple parts of the insect, and the order information assigning step can be configured to assign the order information to the point information based on the order in which the two or more parts are arranged.

[0141] With this configuration, point information is assigned to two or more parts and assigned based on the order of the parts, so that the posture of the insect can be determined appropriately even if the insect is bent, thereby improving the accuracy of identifying insects included in image data.

[0142] The point information assigning step can also be configured to assign point information to the insect data by specifying at least three points: a head corresponding point corresponding to the head of the insect corresponding to the insect data, an abdomen corresponding point corresponding to the abdomen of the insect, and an intermediate point located between the head corresponding point and the abdomen corresponding point.

[0143] With this configuration, the posture of the insect can be determined based on at least three points including the head corresponding point, the abdomen corresponding point, and the midpoint, so that the posture of the insect can be determined appropriately even if the insect is bent, thereby improving the accuracy of identifying insects contained in image data.

[0144] The point information assigning step may also be configured to assign point information to a feature point located between the head corresponding point and the abdomen corresponding point as the intermediate point.

[0145] With this configuration, point information is assigned to the feature points as intermediate points, so that point information can be assigned to approximately the same positions on insects of the same type, thereby improving the accuracy of identifying insects included in image data.

[0146] In addition, the pre-processing image data is generated by an imaging device that includes a camera that is positioned at a distance from the capture surface so as to capture an image of the capture surface in a captured sample in which an insect has been captured on the capture surface, and that moves along the surface direction of the capture surface, and an illumination device that illuminates the capture surface from both sides of the traveling area in which the camera moves, and the imaging device can be configured to move the camera to capture multiple images and connect the captured multiple images to generate pre-processing image data.

[0147] With this configuration, the imaging device irradiates light from both sides of the driving area, preventing insects from being cast in shadow, and the pre-processed image data is generated by stitching together multiple images, preventing the insects in the pre-processed image data from becoming blurred, thereby improving the accuracy of identifying insects contained in the image data.

[0148] Furthermore, a method for generating image data for insect identification that generates image data for insect identification as training data used in machine learning can also be configured to include a type information assignment step that assigns type information regarding the type of insect to the extracted insect data.

[0149] According to this configuration, since it includes a type information assigning step of assigning type information regarding the type of insect, even if the insect included in the image data is bent, the image data including the bent insect can be used as training data.

[0150] In addition, the insect identification system of the above embodiment includes an image data acquisition unit that acquires target image data in which the insect to be identified is included in the image, and an insect identification unit that identifies the type of insect included in the target image data, wherein the insect identification unit is configured to identify the type of insect included in the target image data acquired by the image data acquisition unit using a trained model constructed by machine learning using teacher data, and the trained model is constructed by machine learning using teacher data that includes: teacher image data in which the insect is included in the image; pose information including point information assigned by specifying at least three points to the insect data related to the insect included in the teacher image data and order information regarding the order assigned to the point information; and type information related to the type of insect included in the teacher image data.

[0151] According to this configuration, the trained model is constructed by machine learning using training data that includes point information on at least three points specified for the insect data included in the training image data and posture information including order information, so that the type of insect can be appropriately identified even if the insect is bent, for example. [Explanation of symbols]

[0152] 1... insect, 11... main body, 12... accessory part, 13... head, 14... torso, 15... thorax, 16... abdomen, 2... captured sample, 2a... capture surface, 3... imaging device, 31... housing, 32... imaging means, 33... illumination device, 34... camera, 35... lighting unit, 4... point information, 41... head corresponding point, 42... abdomen corresponding point, 43... intermediate point, 5... sequence information, 6... insect identification system, 61... communication unit, 62... image data acquisition unit, 63... insect identification unit, 64... Output unit, 65...trained model, 7...pest control system, 71...control unit, 711...increase determination unit, 712...source determination unit, 713...treatment method determination unit, 714...insect species extraction unit, 715...factor extraction unit, 716...condition determination unit, 717...prevention method determination unit, 72...storage means, 8...processing unit, 81...notification means, 82...chemical treatment means, C...client terminal, D...insect data, F...frame, I...Internet, P...computer

Claims

1. A method for generating image data for insect identification, which processes pre-processed image data containing insects to generate image data for insect identification, a determination step of determining whether or not it is possible to identify an insect part included in the unprocessed image data; A method for generating image data for insect identification, comprising: an identification process for identifying and extracting each part of an insect whose part is determined to be identifiable in the possibility determination process, and assigning information corresponding to each part to the extracted range.

2. The method for generating image data for insect identification according to claim 1 , wherein information indicating that an insect is an insect is attached to the insect for which it is determined that the part cannot be identified in the determination step.

3. an insect extraction step of extracting insect data relating to insects from the unprocessed image data; 3. The method for generating image data for insect identification according to claim 1, wherein the possibility determining step determines whether or not it is possible to identify a part of an insect corresponding to the insect data extracted in the insect extracting step.

4. The insect extraction step comprises: an insect detection step of detecting insects contained in the unprocessed image data; The method for generating image data for insect identification according to claim 3, further comprising an insect identification step of adding frame information relating to a frame surrounding the insect discovered in the insect discovery step and extracting the insect data based on the added frame information.

5. The method for generating image data for insect identification according to claim 1 , further comprising a point information assigning step of assigning point information corresponding to each part of the insect identified in the identifying step.

Images