Insect traps and systems

Abstract

This application provides an insect trap and system, relating to the technical field of insect trapping devices. The insect trap is used to collect target infrared data containing a dual-channel discrete time series within itself and send the target infrared data to a cloud server. The insect trap is also used to collect images of pests that have fallen into it upon receiving an image acquisition control command from the cloud server. The insect trap includes an insect-catching body, which includes a control module, a trapping module, and a monitoring module. This application enables the insect trap to perform timely and effective image acquisition of pests only upon confirmation of their entry, thereby effectively reducing the number of invalid acquisitions and frequent starts of the image acquisition device within the insect trap, thus reducing energy consumption and preventing heat buildup.

Description

Technical Field

[0001] This application relates to the field of insect-catching devices, and more particularly to insect traps and systems. Background Technology

[0002] Pests in stored grain are a significant factor jeopardizing safe grain storage. Infestations not only lead to losses in grain quantity and quality but can also cause food safety issues. Currently, sampling and screening methods are commonly used to assess pest infestations, but these methods are labor-intensive, cumbersome, and produce results with a time lag. In particular, as pest density increases, both the number of samples per thousand and the quantity of samples taken per batch need to increase to obtain an accurate assessment, which is difficult to achieve in practical management. Therefore, insect traps such as electronic probe traps within grain piles are becoming a growing trend in the automatic monitoring of pests in grain piles.

[0003] Currently, insect traps such as electronic probe traps in grain piles are usually equipped with image acquisition devices. These devices collect images of the inside of the trap in real time or at regular intervals and send the collected images to other processing devices. These devices then store the images and identify the types of pests in the area where the trap is located, allowing for targeted pest control in the region.

[0004] However, for existing insect traps, if the image acquisition interval is too long, it can easily lead to the inability to deal with regional pests in a timely manner. If the image acquisition interval is too short, it will cause the image acquisition device to start frequently, resulting in a large amount of heat accumulation. This will not only increase the energy consumption of the image acquisition device, but also accelerate the reproduction and growth of pests, and pose a significant fire safety hazard. At the same time, it will also generate many invalid images of non-pests, which will further reduce the efficiency and effectiveness of regional pest control.

[0005] Therefore, there is an urgent need to design a method that can effectively avoid invalid image acquisitions and frequent image acquisition starts in the insect trap. Summary of the Invention

[0006] In view of this, embodiments of this application provide an insect trap and system to eliminate or improve one or more defects existing in the prior art.

[0007] The first aspect of this application provides an insect trap for collecting target infrared data containing a dual-channel discrete time series within itself and sending the target infrared data to a cloud server.

[0008] The insect trap is also used to capture images of pests that fall into the trap when or after receiving an image acquisition control command sent by the cloud server.

[0009] The beneficial effects of this technical solution are: it can control the insect trap to perform timely and effective image acquisition of the insect that has fallen inside it only when the insect is confirmed to have fallen in, thereby effectively reducing the number of invalid acquisitions and frequent starts of the image acquisition device in the insect trap, thus reducing the energy consumption of the insect trap and making it less likely for heat to accumulate.

[0010] In some embodiments of this application, the insect trap includes an insect trapping body, which includes a control module, a trapping module, and a monitoring module;

[0011] The trapping module includes a first tube and an insect-collecting funnel installed inside the first tube. The insect-collecting funnel is coaxially arranged with the first tube. In a first direction, one end of the insect-collecting funnel is the first end and the other end is the second end. The diameter of the insect-collecting funnel gradually narrows from the first end to the second end. An insect passage hole is provided at the second end.

[0012] The monitoring module includes a second tube, and an infrared data acquisition component and an insect-swatting component, both installed inside the second tube. The second tube is coaxially arranged with and connected to the first tube. Both the infrared data acquisition component and the insect-swatting component are communicatively connected to the control module. The infrared data acquisition component is used to acquire target infrared data containing a dual-channel discrete time series inside the insect trap according to the instructions of the control module. The insect-swatting component is used to acquire images of pests falling into the insect trap according to the instructions of the control module.

[0013] The beneficial effects of this technical solution are as follows: When in use, it is placed in a grain pile with the first tube positioned above the second tube. Insects in the grain pile enter the first tube and then the insect-collecting funnel. Since the first infrared receiving tube can receive the light emitted by the first infrared emitting tube, the range of light covers the insect-passing hole in the first direction. After the insect falls out of the insect-passing hole on the insect-collecting funnel, it will enter the range of light emitted by the first infrared receiving tube. The control module receives the signal collected by the infrared data acquisition component and controls the insect-swatting component to take a picture of the insect. In this way, the insect trap can only activate the insect-swatting component when an object falls into the insect-passing hole, or choose whether to activate the insect-swatting component, without having to keep the insect-swatting component constantly on. This reduces energy consumption and prevents heat from accumulating. Compared with existing insect traps, it weakens the beneficial effects on the growth and reproduction of pests and also reduces the probability of safety accidents.

[0014] In some embodiments of this application, the infrared data acquisition component includes a first infrared emitting tube and a first infrared receiving tube. The insect-passing hole is disposed close to the infrared data acquisition component. The projection of the insect-passing hole in a first plane is located within the range of light emitted by the first infrared emitting tube that the first infrared receiving tube can receive. The first plane is perpendicular to the first direction, and the axis of the first infrared emitting tube and the axis of the first infrared receiving tube are both located in the plane. The first direction is the axial direction of the first tube.

[0015] The infrared data acquisition component further includes a second infrared receiver and a second infrared emitter coaxially arranged. The axes of the second infrared receiver and the second infrared emitter are both located within the first plane. The first infrared emitter and the first infrared receiver are coaxially arranged, with the axis of the first infrared receiver being the first axis and the axis of the second infrared receiver being the second axis. The intersection of the first axis and the second axis is located at the center of the projection of the insect hole onto the first plane, and the projection of the insect hole onto the first plane is located within the range of light emitted by the second infrared emitter that the second infrared receiver can receive.

[0016] The beneficial effects of this technical solution are as follows: increasing the number of infrared receivers and infrared emitters improves infrared monitoring sensitivity, enabling the control module to receive signals promptly and accurately when an object falls through the insect hole; the intersection of the first axis and the second axis is located at the center of the projection of the insect hole onto the first plane, ensuring that the range of light emitted by the first infrared receiver and the range of light emitted by the second infrared receiver accurately cover the projection of the insect hole onto the first plane. This prevents the projection of the insect hole from being biased to one side, thus avoiding a portion of the projection falling outside the light range and ultimately preventing a decrease in infrared monitoring sensitivity when an object falls through the insect hole.

[0017] In some embodiments of this application, the first axis is perpendicular to the second axis.

[0018] The beneficial effect of this technical solution is that the range of light emitted by the first infrared receiver tube and the range of light emitted by the second infrared receiver tube can overlap significantly, making it less likely for objects falling through the insect hole to fall outside the light range, thus further improving the sensitivity of infrared monitoring.

[0019] In some embodiments of this application, the distance between the first infrared emitting tube and the projection of the insect passage hole in the first plane is greater than the distance between the first infrared receiving tube and the projection of the insect passage hole in the first plane, and the distance between the second infrared emitting tube and the projection of the insect passage hole in the first plane is greater than the distance between the second infrared receiving tube and the projection of the insect passage hole in the first plane.

[0020] The beneficial effects of this technical solution are as follows: Since the light emitted by the infrared emitting tube is distributed in a cone shape, the farther away from the infrared emitting tube, the larger the area covered by the light. Therefore, the distance between the projection of the insect hole in the first plane and the infrared emitting tube is relatively far, and the projection of the light emitted by the infrared emitting tube can more easily cover the projection. The greater the distance between the infrared receiving tube and the infrared emitting tube, the smaller the area that the light emitted by the infrared emitting tube can directly illuminate on the infrared receiving tube. Therefore, the infrared receiving tube is placed closer to the projection of the insect hole in the first plane, so that the projection is more likely to fall within the light range that the infrared receiving tube can receive.

[0021] In some embodiments of this application, the insect-catching assembly includes a camera and a shooting tube. A shooting chamber is formed inside the shooting tube. A shooting chamber inlet is formed at one end of the shooting tube in the first direction, and a shooting chamber outlet is formed at the other end of the shooting tube. The insect passage hole communicates with the shooting chamber inlet. In the first direction, the infrared data acquisition assembly is located between the insect collecting funnel and the shooting chamber inlet. The camera is positioned close to the shooting chamber inlet and is fixed to the inner wall of the shooting chamber. The camera is communicatively connected to the control module to capture images into the shooting chamber.

[0022] The beneficial effects of this technical solution are as follows: when an insect falls out of the insect hole and triggers the infrared data acquisition component, the control module receives the signal from the infrared data acquisition component. During the process of the object falling inside the shooting chamber, the camera is controlled to shoot into the shooting chamber. The process of the object staying inside the shooting chamber provides the camera with sufficient shooting time.

[0023] In some embodiments of this application, the diameter of the shooting chamber gradually decreases from the shooting chamber inlet to the shooting chamber outlet.

[0024] The beneficial effect of this technical solution is that, since the diameter of the shooting chamber gradually decreases from the entrance to the exit, the object will come into contact with the side wall of the shooting chamber after falling into it and before falling out of it. Under the action of friction, the falling speed is slowed down, giving the camera sufficient shooting time.

[0025] In some embodiments of this application, the monitoring module further includes a cleaning component installed inside the second tube. The cleaning component includes a housing, an insect-carrying platform, and a drive unit. The housing is connected to the imaging tube, the drive unit is installed on the housing, and the insect-carrying platform is installed on the drive unit. The drive unit is used to drive the insect-carrying platform to reciprocate between a first position and a second position. In the first position, the insect-carrying platform engages with the imaging chamber outlet, and in the second position, a gap is formed between the insect-carrying platform and the imaging chamber outlet. The drive unit is communicatively connected to the control module.

[0026] The beneficial effects of this technical solution are as follows: After the camera finishes photographing the object that has fallen into the shooting chamber, the control module can activate the drive mechanism, moving the insect-carrying platform from the first position to the second position, allowing the object to fall out of the shooting chamber. Before photographing the objects inside the shooting chamber again, the control module activates the drive mechanism again, moving the insect-carrying platform from the second position to the first position, thus achieving automatic cleaning. The entire structure is relatively simple and also reduces labor costs. The shell and the shooting chamber are connected by screws and nuts.

[0027] In some embodiments of this application, the driving component includes a driving component body and a first output shaft mounted on the driving component body. The driving component is a rotary driving component. The first output shaft is coaxially arranged with the second tube. The cleaning assembly further includes a threaded rod mounted on the first output shaft. A threaded hole is formed on the housing. The threaded rod engages with the threaded hole.

[0028] The beneficial effects of this technical solution are as follows: when the first output shaft of the drive unit rotates, it drives the threaded rod to rotate. Since the threaded rod is engaged with the threaded hole on the housing, the rotation of the threaded rod will drive the threaded rod, the drive unit, and the insect-carrying platform to move in the first direction. When the rotation direction of the first output shaft changes, it drives the rotation direction of the threaded rod to change, and the corresponding drive unit and the insect-carrying platform will also change their movement direction in the first direction.

[0029] In some embodiments of this application, the cleaning component further includes a trigger and a first limit switch and a second limit switch, both mounted on the housing. The first limit switch and the second limit switch are both used to communicate with the control module. The trigger is mounted on the drive body and is used to move with the insect-carrying platform so that when the insect-carrying platform is in a first position, the trigger contacts the first limit switch, and when the insect-carrying platform is in a second position, the trigger contacts the second limit switch.

[0030] The beneficial effects of this technical solution are as follows: when the insect-carrying platform moves to the first position, the limiting component triggers the first limit switch, causing the movement of the insect-carrying platform toward the shooting chamber to stop automatically; and when the insect-carrying platform moves to the second position, the limiting component triggers the second limit switch, causing the movement of the insect-carrying platform away from the shooting chamber to stop automatically.

[0031] In some embodiments of this application, the drive unit further includes a second output shaft, the first output shaft and the second output shaft are coaxially arranged, and the first output shaft and the second output shaft are located at opposite ends of the drive unit body in the first direction, and the insect-carrying platform is fixed to the second output shaft.

[0032] The beneficial effects of this technical solution are as follows: the first output shaft and the second output shaft are driven simultaneously by the drive unit body. After the shooting operation of the object in the shooting chamber is completed, the control module starts the drive unit, which drives the first output shaft and the second output shaft to rotate. On the one hand, it moves the insect-carrying platform from the first position to the second position, and on the other hand, it drives the insect-carrying platform to rotate. Under the action of centrifugal force, the object falling on the insect-carrying platform is thrown off the insect-carrying platform. After the object on the insect-carrying platform is thrown off, the control module causes the drive unit to drive the first output shaft and the second output shaft to change the direction of rotation, and causes the insect-carrying platform to move from the second position to the first position.

[0033] In some embodiments of this application, the insect-carrying platform includes a top surface facing the imaging chamber, on which a first convex ridge and a second convex ridge are formed. The length directions of the first convex ridge and the second convex ridge are both parallel to the top surface, and the first convex ridge and the second convex ridge intersect perpendicularly. A convex cone is also formed at the intersection of the first convex ridge and the second convex ridge.

[0034] In some embodiments of this application, the insect-catching body further includes a collection tube coaxially arranged with the first tube, the collection tube being detachably connected to the end of the second tube away from the first tube, and a collection cavity communicating with the first tube is formed in the collection tube.

[0035] The beneficial effects of this technical solution are as follows: When in use, it is placed in the grain pile so that the first tube is above the second tube, and the second tube is above the collection tube. When the object that falls onto the insect-carrying platform is thrown off the platform by centrifugal force, the object will fall down into the collection tube. After the insect catcher has been used for a period of time, the collection tube can be separated from the second tube, and the object in the collection chamber can be poured out.

[0036] In some embodiments of this application, the insect-catching body further includes a thread tube, which is coaxially arranged with the first tube. One end of the thread tube extends to the end of the first tube away from the second tube, and the other end of the thread tube extends to the insect-collecting funnel. A thread-passing hole is formed on the insect-collecting funnel, and the thread tube communicates with the thread-passing hole.

[0037] The beneficial effects of this technical solution are: by setting up a conduit, it is easy to run the wire inside the insect trap, making it less likely for the wire to interfere with the various components inside the insect trap, and making it less likely for the wire to obstruct the movement of pests or other objects inside the insect trap.

[0038] In some embodiments of this application, the cloud server is used to execute a pest landing identification method, which includes:

[0039] Receive target infrared data containing a dual-channel discrete time series collected from the current insect trap;

[0040] Based on a preset time series classification model, the waveform recognition type corresponding to the discrete time series of the dual channels in the target infrared data is obtained. If the waveform recognition type is a pest detection waveform, it is confirmed that a pest has fallen into the insect trap and the insect trap is controlled to collect images of the pest that has fallen into it.

[0041] The method of obtaining the waveform recognition type corresponding to the discrete time series of the dual channels in the target infrared data based on the preset time series classification model includes:

[0042] Global features are extracted from the discrete time series of the two channels in the target infrared data to obtain the global features of the target infrared data;

[0043] The discrete time series of the two channels and the global features are input into a preset time series classification model so that the time series classification model extracts the local features corresponding to each of the discrete time series of the two channels, and fuses each of the local features with the global features to obtain the waveform type identification result data of the target infrared data. The waveform type identification result data includes the probability of different waveform identification types, and the waveform identification types include pest detection waveforms and at least one non-pest detection waveform.

[0044] The step of performing global feature extraction on the discrete time series of the dual channels in the target infrared data to obtain the global features of the target infrared data includes:

[0045] The discrete time series of the dual channels in the target infrared data are respectively processed by shifting the waveform downward with a minimum value of 0 to obtain two preprocessed time series corresponding to the target infrared data;

[0046] Based on a preset effective threshold, effective sampling points are selected in the two preprocessed time series respectively to form the reaction zone corresponding to each of the two preprocessed time series;

[0047] The global features of the target infrared data are determined based on the reaction regions corresponding to the two preprocessed time series.

[0048] The second aspect of this application provides a pest infestation identification system, comprising: a cloud server interconnected with the first aspect of this application and the insect trap described above;

[0049] The cloud server is used to execute the pest ingress identification method;

[0050] The pest identification method includes:

[0051] Receive target infrared data containing a dual-channel discrete time series collected from the current insect trap;

[0052] Based on a preset time series classification model, the waveform recognition type corresponding to the discrete time series of the dual channels in the target infrared data is obtained. If the waveform recognition type is a pest detection waveform, it is confirmed that a pest has fallen into the insect trap and the insect trap is controlled to collect images of the pest that has fallen into it.

[0053] The method of obtaining the waveform recognition type corresponding to the discrete time series of the dual channels in the target infrared data based on the preset time series classification model includes:

[0054] Global features are extracted from the discrete time series of the two channels in the target infrared data to obtain the global features of the target infrared data;

[0055] The discrete time series of the two channels and the global features are input into a preset time series classification model so that the time series classification model extracts the local features corresponding to each of the discrete time series of the two channels, and fuses each of the local features with the global features to obtain the waveform type identification result data of the target infrared data. The waveform type identification result data includes the probability of different waveform identification types, and the waveform identification types include pest detection waveforms and at least one non-pest detection waveform.

[0056] The step of performing global feature extraction on the discrete time series of the dual channels in the target infrared data to obtain the global features of the target infrared data includes:

[0057] The discrete time series of the dual channels in the target infrared data are respectively processed by shifting the waveform downward with a minimum value of 0 to obtain two preprocessed time series corresponding to the target infrared data;

[0058] Based on a preset effective threshold, effective sampling points are selected in the two preprocessed time series respectively to form the reaction zone corresponding to each of the two preprocessed time series;

[0059] The global features of the target infrared data are determined based on the reaction regions corresponding to the two preprocessed time series.

[0060] The beneficial effects of this technical solution are: it can control the insect trap to perform timely and effective image acquisition of the insect that has fallen inside it only when the insect is confirmed to have fallen in, thereby effectively reducing the number of invalid acquisitions and frequent starts of the image acquisition device in the insect trap, thus reducing the energy consumption of the insect trap and making it less likely for heat to accumulate.

[0061] The insect trap provided in this application is used to collect target infrared data containing a dual-channel discrete time series inside its interior and send the target infrared data to a cloud server; the insect trap is also used to collect images of pests falling into the insect trap when or after receiving an image acquisition control command sent by the cloud server; the insect trap includes an insect trapping body, which includes a control module, a trapping module, and a monitoring module; the trapping module includes a first tube and an insect-collecting funnel installed in the first tube, the insect-collecting funnel being coaxially arranged with the first tube, one end of the insect-collecting funnel being the first end and the other end being the second end in a first direction, the diameter of the insect-collecting funnel gradually narrowing from the first end to the second end, and an insect passage hole being provided at the second end; the monitoring module includes a second tube and infrared data acquisition components installed in the second tube. The system includes an insect-swatting component, with the second tube coaxially arranged and connected to the first tube. Both the infrared data acquisition component and the insect-swatting component are communicatively connected to the control module. The infrared data acquisition component acquires target infrared data containing a dual-channel discrete time series from inside the insect trap according to instructions from the control module. The insect-swatting component acquires images of pests falling into the insect trap according to instructions from the control module. This allows the system to control the insect trap to acquire images of pests only when it is confirmed that they have fallen in, thus effectively reducing the number of invalid acquisitions and frequent starts of the image acquisition device in the insect trap. This reduces the energy consumption of the insect trap, prevents heat buildup, and weakens the beneficial effects on the growth and reproduction of pests compared to existing insect traps. Furthermore, it reduces the probability of safety accidents.

[0062] Additional advantages, objectives, and features of this application will be set forth in part in the description which follows, and will in part become apparent to those skilled in the art upon review of the following description, or may be learned by practice of the application. The objectives and other advantages of this application can be realized and obtained by means of the structures specifically pointed out in the specification and drawings.

[0063] Those skilled in the art will understand that the purposes and advantages that can be achieved with this application are not limited to those specifically described above, and that the above and other purposes that this application can achieve will be more clearly understood from the following detailed description. Attached Figure Description

[0064] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, do not constitute a limitation thereof. The components in the drawings are not drawn to scale but are merely for illustrating the principles of this application. For ease of illustration and description of certain parts of this application, corresponding portions in the drawings may be enlarged, i.e., may appear larger relative to other components in an exemplary device actually manufactured according to this application. In the drawings:

[0065] Figure 1 This is a schematic diagram of the first process of the pest identification method in the embodiments of this application.

[0066] Figure 2 This is a schematic diagram of the second process of the pest falling into the identification method in the embodiments of this application.

[0067] Figure 3 This is a schematic diagram of the second process of the pest falling into the identification method in the embodiments of this application.

[0068] Figure 4 This is a schematic diagram illustrating the network architecture of the feature extraction and classification network in the embodiments of this application.

[0069] Figure 5 This is a schematic diagram of the first process of the pest type identification method in the embodiments of this application.

[0070] Figure 6 This is a schematic diagram of the second process of the pest type identification method in the embodiments of this application.

[0071] Figure 7 This is a schematic diagram of the third process of the pest type identification method in the embodiments of this application.

[0072] Figure 8 This is a schematic diagram illustrating an example of how a recursive graph algorithm is used in this application to transform raw data into a two-dimensional image.

[0073] Figure 9This is an example schematic diagram of the curve of H(Z) changing with ∈ (0 < ∈ < 0.24) in an embodiment of this application.

[0074] Figure 10 This is a schematic diagram of the fourth process of the pest type identification method in the embodiments of this application.

[0075] Figure 11(a) is a schematic diagram illustrating the first structural example of the Inception module in the embodiments of this application.

[0076] Figure 11(b) is a schematic diagram illustrating a second structural example of the Inception module in an embodiment of this application.

[0077] Figure 11(c) is a schematic diagram illustrating a third structure of the Inception module in an embodiment of this application.

[0078] Figure 12 This is a schematic diagram of the pest identification device in the embodiments of this application.

[0079] Figure 13 This is a schematic diagram of the pest type identification device in the embodiments of this application.

[0080] Figure 14 This is a three-dimensional structural schematic diagram of one embodiment of the insect trap provided in this application.

[0081] Figure 15 This is a front view structural diagram of one embodiment of the insect-catching body provided in this application.

[0082] Figure 16 for Figure 15 A schematic diagram of the cross-sectional structure at point AA.

[0083] Figure 17 for Figure 16 A magnified view of a portion of point B in the middle.

[0084] Figure 18 This is a partial three-dimensional structural schematic diagram of one embodiment of the insect trap provided in this application.

[0085] Figure 19 This is a partial left-side view of one embodiment of the insect trap provided in this application.

[0086] Figure 20 This is a partial top view of one embodiment of the insect trap provided in this application.

[0087] Figures 21 to 23 This is a partial top view of one embodiment of the infrared data acquisition component provided in this application.

[0088] Figure 24 This is a partial three-dimensional structural schematic diagram of one embodiment of the insect trap provided in this application.

[0089] Figure 25 This is a partial left-side view of one embodiment of the insect trap provided in this application.

[0090] Figure 26 for Figure 25 A schematic diagram of the cross-sectional structure at point CC.

[0091] Figure 27 This is a top view of the insect-carrying platform provided in an embodiment of this application.

[0092] Figure 28 This is a schematic diagram of the connection relationship of the pest falling into the identification system provided in the embodiments of this application.

[0093] Figure 29 Functional block diagram of the main control circuit board provided for the application example of this application.

[0094] Figure 30 Example diagram of pest voltage sequence provided for application of this application.

[0095] Figure 31 Example diagram of voltage sequence for grain debris provided as an application example of this application.

[0096] Figure 32 A schematic diagram illustrating the workflow of the pest infestation identification algorithm provided for the application example of this application.

[0097] Figure 33 Example diagram of pest identification waveform provided for application of this application.

[0098] Figure 34(a) is an example of a booklice-triggered waveform in the non-pest identification waveform provided in the application example of this application.

[0099] Figure 34(b) is an example of a non-pest identification waveform provided in the application example of this application, which is a waveform repeatedly triggered by pests staying in the detection area.

[0100] Figure 34(c) is an example of the first type of false triggering waveform triggered by two objects falling in succession in the non-pest identification waveform provided in the application example of this application.

[0101] Figure 34(d) is an example of a second type of false-triggered waveform generated by equipment failure or large shaking in the non-pest identification waveform provided in the application example of this application.

[0102] Figure 35A waveform diagram of the retained data in the manual cleaning example of the pest species identification dataset provided as an application example of this application.

[0103] Figure 36 A waveform diagram of the cleaned data in the example of manually cleaning the pest species identification dataset provided as an application example of this application.

[0104] Figure label:

[0105] 100-Insect catching body;

[0106] 110 - Trapping Module;

[0107] 111 - First tube;

[0108] 111a - Pipe cap section;

[0109] 111b - Trapping section;

[0110] 111ba - Insect trap;

[0111] 112-Insect-collecting funnel;

[0112] 112a - Through hole;

[0113] 112b - Through the wormhole;

[0114] 112b'-projection;

[0115] 113-Through the insect tube;

[0116] 114 - Cable conversion board;

[0117] 115 - Light-guided trap;

[0118] 120 - Monitoring Module;

[0119] 121 - Second tube;

[0120] 122-Shooting tube;

[0121] 122a - Camera compartment;

[0122] 122b - Entrance to the shooting compartment;

[0123] 122c - Camera compartment exit;

[0124] 123 - Clean up components;

[0125] 123a - Insect-carrying platform;

[0126] 123aa - First convex ridge;

[0127] 123ab - Convex cone;

[0128] 123ac - Second convex ridge;

[0129] 123b - Housing;

[0130] 123ba - Wiring hole;

[0131] 123c - Threaded rod;

[0132] 123d - Drive components;

[0133] 124-Infrared data acquisition component;

[0134] 124a - First infrared receiver tube;

[0135] 124b - First Infrared Emitter;

[0136] 124c - Infrared data acquisition circuit board;

[0137] 124d - Second infrared receiver tube;

[0138] 124e - Second Infrared Emitter;

[0139] 125-camera;

[0140] 130 - Collection tube;

[0141] 140 - Cable management conduit;

[0142] 150 - Control Module;

[0143] 200-cable;

[0144] 300 - Wireless communication device;

[0145] 400-Gran Head. Detailed Implementation

[0146] To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail below with reference to the embodiments and accompanying drawings. Here, the illustrative embodiments and their descriptions are used to explain this application, but are not intended to limit it.

[0147] It should also be noted that, in order to avoid obscuring this application with unnecessary details, only the structures and / or processing steps closely related to the scheme according to this application are shown in the accompanying drawings, while other details that are not closely related to this application are omitted.

[0148] It should be emphasized that the term "including / comprises" as used herein refers to the presence of a feature, element, step, or component, but does not exclude the presence or addition of one or more other features, elements, steps, or components.

[0149] It should also be noted that, unless otherwise specified, the term "connection" in this article can refer not only to a direct connection, but also to an indirect connection involving an intermediary.

[0150] In the following description, embodiments of the present application will be illustrated with reference to the accompanying drawings. In the drawings, the same reference numerals represent the same or similar parts, or the same or similar steps.

[0151] To design a method that effectively avoids invalid image acquisitions and frequent image acquisition restarts in an insect trap, embodiments of this application provide an insect fall-in identification method, an insect type identification method, an insect fall-in identification device for executing the insect fall-in identification method, an insect type identification device for executing the insect type identification method, a physical device and a computer-readable storage medium, an insect trap, and an insect fall-in identification system. These systems can control the insect trap to perform timely and effective image acquisition of the insect only when it is confirmed that a pest has fallen inside. This effectively reduces the number of invalid acquisitions and frequent restarts of the image acquisition device in the insect trap, thereby reducing the energy consumption of the insect trap and preventing heat buildup.

[0152] The following examples will provide a detailed description.

[0153] Based on this, embodiments of this application provide a pest entry identification method that can be implemented by a pest entry identification device, see [link to relevant documentation]. Figure 1 The pest identification method specifically includes the following:

[0154] Step 1000: Receive the target infrared data containing a dual-channel discrete time series collected from the current self-trapping insect trap.

[0155] In one or more embodiments of this application, the target infrared data containing a dual-channel discrete time series collected from inside the insect trap can be collected by an infrared data acquisition component installed inside the insect trap, and then the acquired dual-channel discrete time series is transmitted to the pest landing identification device via a communication module through a control module inside the insect trap. The pest landing identification device can be specifically implemented in a cloud server.

[0156] It is understood that the target infrared data can be simply referred to as the target raw data. The waveform of the target raw data consists of two sets of related discrete time series. That is, the target infrared data contains a dual-channel discrete time series. The discrete time series specifically refers to a voltage sequence that is continuous over time.

[0157] In one example of this application, the length of each time series can be set to 128, and the selection of this time series length is related to the ADC sampling frequency. Experiments show that when the sampling rate is set to 10kHz, a time series length of 128 provides a capture time of 12.8ms, which is sufficient to capture the complete falling process of the pest (typically less than 10ms). Selecting a time series of 128 points not only records the details of the pest's fall but also facilitates data storage. The corresponding total length of the dual-channel discrete time series is 256.

[0158] Step 2000: Based on the preset time series classification model, obtain the waveform recognition type corresponding to the discrete time series of the dual channels in the target infrared data. If the waveform recognition type is a pest detection waveform, confirm that a pest has fallen into the insect trap and control the insect trap to collect images of the pest that has fallen into it.

[0159] In step 2000, the time series classification model can specifically refer to a feature extraction and classification network trained using a learnable time series classification algorithm. Specifically, the learnable time series classification algorithm refers to the learnable Shaples algorithm. The core of the Shaples algorithm is to find highly informative subsequences in the time series dataset, called shaplets. The biggest difference between the learnable Shaples algorithm and the traditional Shaples algorithm is that it can automatically iteratively find the shaplet that achieves the best time series classification effect.

[0160] The execution of step 1000 is based on the insect-swatting component inside the insect trap being in a dormant or off state. Only when the waveform recognition type corresponding to the dual-channel discrete time series is identified as an insect detection waveform through step 2000, the insect-falling recognition device will send a corresponding notification message to the insect trap. This will cause the controller of the insect-falling recognition device to issue a start command to the insect-swatting component inside the insect trap after recognizing that the content displayed in the notification message is a dual-channel discrete time series waveform recognition type of insect detection waveform. The insect-swatting component will then collect images of the insects that have fallen into the insect trap and send the collected real-scene image data to a device for storing and displaying the real-scene image data. This device can be the insect-falling recognition device used to execute the insect-falling recognition method, or it can send the data to other devices with data processing and display functions.

[0161] As can be seen from the above description, the pest entry identification method provided in this application embodiment can effectively identify and distinguish whether the object falling into the insect trap is a pest, and can effectively improve the accuracy and reliability of pest entry identification. It can also control the insect trap to perform timely and effective image acquisition of the pest that has fallen into it only when it is confirmed that a pest has fallen in, thereby effectively reducing the number of invalid acquisitions and frequent starts of the image acquisition device in the insect trap, thereby reducing the energy consumption of the insect trap and making it less likely for heat to accumulate. Compared with existing insect traps, it weakens the beneficial effects on the growth and reproduction of pests, and also reduces the probability of safety accidents.

[0162] To further improve the accuracy and effectiveness of waveform type identification results for target infrared data, a pest landing identification method is provided in this application embodiment, see [link to relevant documentation]. Figure 2 Step 2000 in the pest identification method specifically includes the following:

[0163] Step 2100: Perform global feature extraction on the discrete time series of the dual channels in the target infrared data to obtain the global features of the target infrared data.

[0164] Step 2200: Input the discrete time series of the two channels and the global features into a preset time series classification model, so that the time series classification model extracts the local features corresponding to the discrete time series of the two channels respectively, and fuses each of the local features with the global features to obtain the waveform type identification result data of the target infrared data. The waveform type identification result data includes the probability of different waveform identification types, and the waveform identification types include pest detection waveforms and at least one non-pest detection waveform.

[0165] Step 2300: If the waveform type recognition result data shows that the waveform recognition type of the target infrared data is a pest detection waveform, then it is confirmed that a pest has fallen into the insect trap and the insect trap is controlled to collect images of the pest that has fallen into it.

[0166] As can be seen from the above description, the pest landing identification method provided in this application can effectively improve the accuracy and effectiveness of the waveform type identification result data of the target infrared data by performing global feature extraction of dual-channel discrete time series and using a time series classification model to achieve local feature extraction and fusing each of the local features with the global features.

[0167] To further improve the accuracy and effectiveness of global feature extraction, in a pest landing identification method provided in this application embodiment, see [link to relevant documentation]. Figure 3 Step 2100 in the pest identification method specifically includes the following:

[0168] Step 2110: Perform waveform downward shifting with a minimum value of 0 on the discrete time series of the dual channels in the target infrared data to obtain two preprocessed time series corresponding to the target infrared data.

[0169] Step 2120: Based on a preset effective threshold, select effective sampling points in the two preprocessed time series respectively to form the reaction zone corresponding to each of the two preprocessed time series.

[0170] In step 2120, the effective threshold can be equal to a first percentage of the maximum sampled value in the preprocessed time series, which can be set between 5% and 50%, preferably 20%.

[0171] Specifically, for the preprocessed time series, since its 128 sampling points are continuous time samples, only a portion of these sampling points correspond to situations where objects have fallen. Furthermore, when an object falls through the detection area, the sampling value exhibits a trend of first rising, then falling, and finally leveling off. To concentrate the features extracted by the algorithm on situations where objects have fallen, in one example of this application, sampling points with values ​​greater than 20% of the maximum value in the time series are generally considered as more relevant and effective sampling points; the set of these effective sampling points is defined as the reaction zone.

[0172] Step 2130: Determine the global features of the target infrared data based on the reaction regions corresponding to the two preprocessed time series.

[0173] In the case where only one pest falls through the detection area, there is only one reaction zone. However, for various waveforms in the actual warehouse, there may be more than one reaction zone. To obtain the number of reaction zones in different waveforms, a set class Ω is introduced, consisting of all subsets of reaction zones M that satisfy the conditions (hereinafter referred to as sub-reaction zones). For the case of only one reaction zone, Ω = {M}. Based on this, in one or more embodiments of this application, the global features of the target infrared data may include: the maximum sampled value in the preprocessed time series, the sub-reaction zones corresponding to the two preprocessed time series, and the Fast Fourier Transform (FFT) result of the preprocessed time series.

[0174] As can be seen from the above description, the pest landing identification method provided in this application can effectively improve the effectiveness of global feature extraction by preprocessing the discrete time series of dual channels and generating reaction zones, thereby further improving the accuracy and effectiveness of the waveform type identification results of infrared data.

[0175] To further improve the effectiveness and accuracy of waveform recognition using time series classification models, a pest landing identification method is provided in an embodiment of this application, see [link to relevant documentation]. Figure 2 The method for identifying pests includes the following steps prior to step 1000 or step 2000:

[0176] Step 0100: Obtain each historical infrared data containing different dual-channel discrete time series and their corresponding waveform identification type labels, wherein the waveform identification type labels include labels corresponding to pest detection waveforms and at least one non-pest detection waveform.

[0177] In step 0100, in order to further improve the accuracy and reliability of the waveform identification type results corresponding to the discrete time series of the dual channels in the target infrared data, the non-pest detection waveforms include: booklice-triggered waveforms, pest-repeated-triggered waveforms, and false-triggered waveforms.

[0178] Step 0200: Perform global feature extraction on the discrete time series of the dual channels in each of the historical infrared data to obtain the global features corresponding to each of the historical infrared data.

[0179] Step 0300: Based on each of the historical infrared data, the corresponding global features and waveform recognition type labels, a preset feature extraction and classification network is trained using a learnable time series classification algorithm, so as to train the feature extraction and classification network into a time series classification model for recognizing the waveform recognition type of infrared data.

[0180] As can be seen from the above description, the pest falling identification method provided in this application improves the effectiveness and accuracy of waveform identification by training the feature extraction and classification network into a time series classification model.

[0181] To improve the reliability and effectiveness of training the feature extraction and classification network, a pest landing identification method is provided in an embodiment of this application, see [link to relevant documentation]. Figure 3 Step 0200 of the pest identification method specifically includes the following:

[0182] Step 0210: Perform waveform downshifting with a minimum value of 0 on the discrete time series of the dual channels in each of the historical infrared data to obtain two preprocessed time series corresponding to each of the historical infrared data.

[0183] In step 0210, historical infrared data can be written as raw data waveforms, consisting of two sets of related discrete time series, each with a length of 128. In this embodiment, the time series features of a single set are first calculated, and then the two sets of features are combined to form a global feature of a single data point. Any set of raw time series is defined as f(k), k∈K; where K is the set of sampling points k, K={k|k∈Z,0≤k≤127}; the time series is preprocessed by shifting it downwards to its minimum value of 0, resulting in the preprocessed time series F(k)=f(k)-min[f(k)].

[0184] Step 0220: Based on a preset effective threshold, select effective sampling points in the two preprocessed time series corresponding to each of the historical infrared data to form the reaction zone of each of the two preprocessed time series corresponding to each of the historical infrared data.

[0185] In step 0220, for the time series F(k), since its 128 sampling points are continuous time samples, only a portion of these sampling points correspond to the situation where an object falls. Furthermore, when an object falls through the detection area, the sampling value exhibits a trend of first rising, then falling, and finally leveling off. To concentrate the features extracted by the algorithm on situations where an object falls, this application, without loss of generality, considers sampling points with sampling values ​​greater than 20% of the maximum value in the time series as more important effective sampling points t. The set of these effective sampling points t is defined as the reaction region M = {t|t∈K, F(t)>0.2×max[F(k)]}.

[0186] Step 0230: Obtain at least one sub-reaction zone for each of the preprocessed time series, and select the sub-reaction zone with the largest number of valid sampling points for each preprocessed time series as the maximum length reaction zone of the preprocessed time series.

[0187] Specifically, when only one pest falls through the detection area, there is only one reaction zone. However, for various waveforms in the actual warehouse, there may be more than one reaction zone. To obtain the number of reaction zones in different waveforms, a set class Ω is introduced, consisting of all subsets of reaction zone M that satisfy the conditions (hereinafter referred to as sub-reaction zones). For the case of only one reaction zone, Ω = {M}; otherwise, Ω is calculated by the following formula (1):

[0188]

[0189] The length of the reaction region is defined as the number of sampling points contained within it. For each sub-reaction region in Ω, the one with the largest length is denoted as G, then:

[0190]

[0191] i represents the index, Ai and Ai+1 represent distinct subsets of M, and card(A i ) represents the number of elements in Ai.

[0192] Step 0240: Determine the global features of each historical infrared data according to the maximum length reaction zone of the two preprocessed time series corresponding to each historical infrared data.

[0193] After the above calculations, the global features of the time series are shown in Table 1:

[0194] Table 1

[0195]

[0196] In Table 1, max[F(k)] represents the maximum value in F(k); card(Ω) represents the number of elements in Ω; F(i) represents the amplitude of the sampling point with index i in the sequence; min(G) represents the minimum value of the elements in G; max(G) represents the maximum value of the elements in G; FFT[F(k)] represents the FFT transformation result of the sequence F(k). Since the energy of the four types of waveforms is mainly concentrated in the low frequency, only the amplitudes corresponding to the eight frequency points near the zero point in the FFT transformation result are used as features.

[0197] As can be seen from the above description, the pest landing identification method provided in this application can effectively improve the reliability and effectiveness of the training feature extraction and classification network process by preprocessing the training data and generating the reaction zone.

[0198] To further improve the reliability and effectiveness of the training process for the feature extraction and classification network, in the pest landing identification method provided in this application embodiment, the feature extraction and classification network specifically includes the following components:

[0199] The input layer, local feature extraction layer, and feature fusion layer are connected in sequence.

[0200] The input layer is used to receive the dual-channel discrete time series, corresponding global features, and waveform recognition type labels from the historical infrared data;

[0201] The local feature extraction layer is used to extract local features from the discrete time series of the two channels in the historical infrared data based on a learnable temporal classification algorithm, so as to obtain the local features corresponding to the discrete time series of the two channels respectively. The local feature extraction layer contains multiple feature extraction blocks.

[0202] The feature fusion layer is used to perform activation function calculation, feature fusion, and probability transformation on the global features and local features corresponding to the historical infrared data to obtain the probability of each different waveform recognition type corresponding to the historical infrared data. Then, based on the cross-entropy loss function, the loss value between the waveform recognition type of the historical infrared data and the probability of each different waveform recognition type corresponding to the historical infrared data is calculated.

[0203] Specifically, the local feature extraction layer can be abbreviated as the soft min layer. See also Figure 4 The network architecture of the feature extraction and classification network shown is used to extract local features of data waveforms, and the local and global features of the data waveforms are fused in the feature extraction and classification network to finally achieve waveform classification.

[0204] like Figure 4 As shown, the input to the feature extraction and classification network is raw time series waveform data with a length Q of 256.

[0205] The Soft min layer uses a learnable Shaples algorithm to extract local features from the input data. In the traditional Shaples algorithm, the distance d from each shaplet to the original time series is calculated. Sm，n As a feature, this distance is calculated as follows:

[0206]

[0207] In equation (3), TSm represents the m-th sequence in the dataset, Sn represents the n-th shaplet found, and L is the length of the shaplet, although this distance is not differentiable for Sn. j represents the index; l represents the index; TS m，j+l-1 S represents the (j+l-1)th element in the m-th sequence of the dataset; n，l This represents the l-th element of the n-th shaplet, where m represents the index and n represents the index.

[0208] The learnable Shaples algorithm uses a differentiable soft-min distance estimate d. Sm，n In equation (4), α = -100, which is the experimental value that can accurately estimate d. Sm，n The parameter value can be calculated by the following formula (4):

[0209]

[0210] in:

[0211]

[0212] Where, soft min dist represents the minimum differentiable distance from the shaplet to the original time series, D m，n，j This represents the Euclidean distance between the j-th subsequence extracted from the m-th sequence in the dataset and the n-th shaplet. This indicates that an intermediate quantity is being calculated, and j′ represents the index. This indicates that intermediate quantities are being calculated.

[0213] Each block in the soft min layer outputs the soft min distance between itself and the input time series, serving as a local feature of the input sequence. The number of blocks and the length of each block in the soft min layer are configurable hyperparameters. Since the original time series length is H = 256, this paper sets the block lengths in the soft min layer to 0.25H, 0.5H, and 0.75H, resulting in 0.2H = 51 blocks of each length. Therefore, the feature vector output by the soft min layer has dimensions (1, 51x3). After passing through a fully connected layer, the feature vector output by the soft min layer becomes (1, 24) in dimension, and... Figure 4 The global features of the input sequence calculated in the previous step are concatenated, resulting in a fused feature vector with dimensions (1, 52). Before concatenation, all features are activated using the sigmoid activation function. The fused features are then passed through a fully connected layer and another sigmoid activation function to transform them into probabilities for four waveform categories. These probabilities are then compared with the true class labels of the input data using the cross-entropy loss function to calculate the loss value.

[0214] As can be seen from the above description, the pest infestation identification method provided in this application, by designing a specific architecture for a feature extraction and classification network based on a learnable temporal classification algorithm, can effectively improve the effectiveness and accuracy of the feature extraction and classification network in identifying local features, thereby improving the reliability and effectiveness of the training process of the feature extraction and classification network.

[0215] To further improve the effectiveness and accuracy of feature extraction and classification networks in identifying local features, a pest landing identification method is provided in this application embodiment, see [link to relevant documentation]. Figure 3 The pest identification method further includes the following specific content between steps 0200 and 0300:

[0216] Step 0250: Obtain a subsequence with the same length as each feature extraction block in the local feature extraction layer based on the sliding window method for each of the historical infrared data.

[0217] Step 0260: Obtain the cluster center of each subsequence based on the clustering algorithm, and use the cluster center as the initial parameter of the local feature extraction layer to complete the initialization of the local feature extraction layer.

[0218] As can be seen from the above description, the pest landing identification method provided in this application embodiment can further improve the effectiveness and accuracy of feature extraction and classification network identification of local features by initializing the local feature extraction layer.

[0219] To further improve the reliability and effectiveness of time series classification models, in the pest landing identification method provided in this application embodiment, the time series classification model specifically includes the following:

[0220] The input layer is used to receive the dual-channel discrete time series and corresponding global features from the target infrared data;

[0221] The local feature extraction layer is used to extract local features from the discrete time series of the two channels in the target infrared data based on a learnable temporal classification algorithm, so as to obtain the local features corresponding to the discrete time series of the two channels respectively.

[0222] The feature fusion layer is used to perform activation function calculation, feature fusion, and probability transformation on the global and local features corresponding to the target infrared data to obtain the probabilities of different waveform recognition types.

[0223] It is understood that the network architecture of the time series classification model can be the same as that of the feature extraction and classification network, wherein the feature fusion layer of the time series classification model does not need to calculate the loss value of the probability of waveform recognition type and label.

[0224] As can be seen from the above description, the pest infestation identification method provided in this application, by designing a specific architecture of a time series classification model based on a learnable time series classification algorithm, can effectively improve the effectiveness and accuracy of the time series classification model in identifying local features.

[0225] This application also provides an embodiment of a pest type identification method that can be implemented by a pest type identification device, see [link to embodiment]. Figure 5 The pest type identification method specifically includes the following:

[0226] Step 3000: Obtain a notification message indicating that the waveform identification type of the target infrared data is a pest detection waveform;

[0227] In step 3000, the pest type identification device can obtain a notification message from other devices indicating that the waveform identification type corresponding to the target infrared data is a pest detection waveform, or it can be set on the same cloud server as the pest landing identification device. If it is the latter, the pest type identification device can directly determine whether to trigger step 4000 based on the identification result of the pest landing identification method in the aforementioned embodiment.

[0228] Step 4000: Obtain pest type identification result data in the current insect trap based on the preset pest type identification model.

[0229] For example, if the pest falling into the identification method provided in the foregoing embodiments of this application determines that the waveform identification type is a pest detection waveform to confirm that a pest has fallen into the current insect trap and controls the insect trap to collect images of the pest that has fallen into it, then step 4000 obtains the pest type identification result data in the current insect trap based on a preset pest type identification model.

[0230] As can be seen from the above description, the pest type identification method provided in this application obtains the pest type identification result data in the current insect trap based on a preset pest type identification model. On the basis of identifying pests falling into the insect trap, it can further realize the automation and efficiency of pest type identification, thereby effectively improving the timeliness and pertinence of the user's pest control in the area where the insect trap is located based on the pest type identification result.

[0231] To improve the accuracy and efficiency of pest type identification, the pest type identification method provided in this application refers to... Figure 6 Step 4000 in the pest type identification method may include the following:

[0232] Step 4100: Receive real-world image data of pests that have fallen into the trap.

[0233] Step 4200: Input the real-scene image data into a preset pest type recognition model so that the preset pest type recognition model outputs pest type recognition result data of the real-scene image data, wherein the pest type recognition model includes a first convolutional neural network.

[0234] As can be seen from the above description, the pest type identification method provided in this application embodiment can effectively improve the accuracy and efficiency of pest type identification by automatically identifying pest types from real-scene image data.

[0235] To improve the accuracy and efficiency of pest type identification, the pest type identification method provided in this application refers to... Figure 7 Step 4000 in the pest type identification method may also include the following:

[0236] Step 4300: Based on a preset recursive graph algorithm, convert the discrete time series of the dual channels in the target infrared data into a two-dimensional recursive image.

[0237] Step 4400: Input the two-dimensional recursive image into a preset pest type recognition model so that the preset pest type recognition model outputs pest type recognition result data of the two-dimensional recursive image data, wherein the pest type recognition model includes a second convolutional neural network.

[0238] Specifically, in response to the problem that the features contained in the original one-dimensional time series data generated by the infrared electronic probe trap for catching pests are insufficient to distinguish pest species with similar body size, this application embodiment will first start by deeply mining the features of the original waveform data of pest detection, convert the target infrared data from dual-channel one-dimensional time series data into a two-dimensional image through a recursive graph algorithm, and then use a deep neural network to extract features and classify the two-dimensional image after the original data is converted, thereby realizing the identification of the species of pests caught.

[0239] As can be seen from the above description, the pest type identification method provided in this application can effectively improve the convenience and efficiency of pest type identification by identifying pest types from two-dimensional recursive images after waveform conversion.

[0240] Specifically, the pest type identification algorithm is implemented as follows:

[0241] ① Use a recursive graph algorithm to transform the original data into a two-dimensional image.

[0242] First, the phase space of the original time series is generated using an embedded delay method. Assume the original time series X is:

[0243] X={x i |i=1,2,...N} (6)

[0244] Among them, x i represents the i-th element in the original sequence; N represents the length of the original sequence.

[0245] A subsequence v of X can be obtained by choosing the embedding dimension d and the time delay τ. i :

[0246] v i (d, τ) = {x i+kτ ∈X|k=0,1,2...d-1} (7)

[0247] Among them, v i (d, τ) represents a subsequence of x obtained by applying the embedding dimension d and the time delay τ; x i+kτ Let represent the (i+kτ)th element in the above sequence. Because we need to ensure that v... i Each element belongs to X, that is, the subscript index of x in equation (7) cannot exceed N, where x represents an element in the sequence vi(d, τ), then we have:

[0248] i≤N-(d-1)τ (8)

[0249] Under the condition of satisfying equation (8), all subsequences vi of X constitute a vector space of dimension (d, N-(d-1)τ), which is called the phase space of the original sequence X. The recursion graph indicates whether there is recursion between any two vectors in the phase space, that is, the recursion graph R is an N-(d-1)τ square matrix:

[0250]

[0251] Any element Ri in the recursive graph R j Represent vectors vi and v in phase space j The degree of recursion between them is calculated using equation (10):

[0252] R ij =θ(t) (10)

[0253] Where t represents vectors vi and v j Euclidean distance between them:

[0254] t = ||v i -v j || (11)

[0255] Where θ represents the function name; v j Let represent the j-th vector in the vector space.

[0256] There are many possible expressions for the function θ, among which the function θopt(t) was proposed in the optimization of recursive graph algorithms. This function makes Rij change with vectors vi and v j The specific numerical value of the Euclidean distance between them changes continuously.

[0257] First, θopt(t) is decreasing with respect to t and the rate of decrease gradually decreases, where ∈ is the threshold. Therefore, θopt(t) satisfies the differential equation (12):

[0258]

[0259] Furthermore, θopt(t) also satisfies two boundary conditions. Assuming that the set of Euclidean distances between any two vectors in the phase space is D, then the boundary conditions satisfied by θopt(t) are:

[0260]

[0261] θ opt (max(D))=0 (14)

[0262] Where max(D) represents the maximum value of the set of Euclidean distances between any two vectors in phase space, which is the maximum value of the elements in D.

[0263] Solving equations (12) to (14) simultaneously yields:

[0264]

[0265] aln(ct) represents the calculation of intermediate quantities.

[0266] in:

[0267]

[0268]

[0269] Where ln(c∈) represents the intermediate quantity to be calculated.

[0270] The above is the algorithm flow for converting a single time series data into a two-dimensional recursive graph. However, the insect detection waveform data of the trap consists of two time series with a length of 128 in each channel. If they are spliced ​​together as a time series with a length of 256 and then converted into a recursive graph, the state information of each channel in the original data, as well as the relevant information between the two channels, will be mixed together. Therefore, this application designs a method to convert the original time series of the trap into a three-channel optimized recursive graph. That is, firstly, the phase spaces Vch1 and Vch2 are generated for the two channels of original data according to equations (6) to (8), and then the three-channel recursive graph is generated according to equations (18) to (20). Channel 1 and channel 2 represent the recursiveness of vectors in Vch1 and Vch2, respectively, and channel 3 represents the recursiveness between any vector in Vch1 and any vector in Vch2.

[0271]

[0272]

[0273]

[0274] Among them, R ch1 Represents the recursion graph of channel 1; R ch2 Represents the recursion graph of channel 2; R ch3 Represents the recursion graph for channel 3; R ij This represents an element in the recursive graph R.

[0275] In summary, the process of transforming raw data into a two-dimensional image using a recursive graph algorithm is illustrated below. Figure 8 As shown.

[0276] ② Parameter selection for recursive graph algorithm

[0277] In the method described above for converting raw time-series data of pest detection into a two-dimensional recursive graph, there are three variable parameters: the embedding dimension d, the embedding delay τ, and the threshold ∈ when calculating the vector recursion. Among them, the embedding dimension d and the embedding delay τ only affect the dimension of the generated phase space, and the threshold ∈ only affects the specific recursive graph generated from the phase space.

[0278] The practical meaning of phase space is to represent all possible states of a system. To measure the quality of the phase space reconstructed from a single time series data point through embedding delay, it's necessary to compare the similarity between the reconstructed phase space and the system's true phase space. However, for the process of a pest falling through a detection area, there are actually countless possible states. Because it's impossible to determine the exact posture of the pest as it passes through the detection area, the true phase space of the system is difficult to calculate, making it impossible to quantify the quality of the phase space reconstruction. Furthermore, the threshold ∈, rather than the embedding dimension d and embedding delay τ, directly determines the shape of the two-dimensional recursive graph transformed from the original data. Therefore, the value of the threshold ∈ will be given priority when selecting parameters.

[0279] When ∈ takes the value 0, all elements in the recursion graph will be 0; when ∈ takes the value 1, all elements in the recursion graph will be 1. When the values ​​of elements in the recursion graph are too concentrated, their characteristics will disappear; therefore, it is desirable for the values ​​of elements in the recursion graph to be as dispersed as possible. Assume that set U is the set of all values ​​in Rij, Z is a discrete random variable defined on U, and PZ(z) is the probability density function of Z. Information entropy is used to evaluate whether the distribution of Z is discrete or concentrated:

[0280]

[0281] Where H(Z) represents the information entropy of Z, and P Z Let z represent the probability density function of Z, and z represent the values ​​of the random variable Z. k This represents the k-th value of the random variable Z.

[0282] Clearly, H(Z) is related to ∈. Optimizing the parameter ∈ means maximizing H(Z). In fact, to avoid all elements in the recursion graph being 0 or 1, the range of values ​​for the parameter ∈ should satisfy:

[0283] min(D)<∈ <max(D) (22)

[0284] Here, min(D) represents the minimum value of set D.

[0285] Within the range of values ​​of ∈, the optimal parameter ∈ is found using uniform sampling. Ten points are uniformly selected within the range of equation (22) as the values ​​of ∈, and the curve of H(Z) as a function of ∈ is plotted as follows. Figure 9 As shown.

[0286] When generating recursive graphs for all data in the dataset, this paper found that when the phase space of each channel's time series is transformed into a recursive graph, the changing trend of H(Z) with ∈ is as follows: Figure 9 As shown, therefore, this application determines the value of ∈ according to the following process, and the final determined value of ∈ is 0.0473:

[0287] 1) Calculate the average of min(D) and max(D) of all data in the dataset as the min(D) and max(D) of the entire dataset.

[0288] 2) Use equation (23) to determine ∈.

[0289]

[0290] To improve the accuracy and efficiency of pest type identification, the pest type identification method provided in this application refers to... Figure 10 The pest type identification method may further include the following content after step 4000:

[0291] Step 5000: If the pest type identification model includes a first convolutional neural network and a second convolutional neural network, wherein the first convolutional neural network is used to identify the pest type identification result data corresponding to the real-scene image data of the pests that have fallen into the trap collected by the insect trap, and the second convolutional neural network is used to identify the pest type identification result data corresponding to the two-dimensional recursive image obtained by converting the dual-channel discrete time series in the target infrared data, then the second convolutional neural network is iteratively optimized based on the pest type identification result data corresponding to the real-scene image data output by the first convolutional neural network.

[0292] As can be seen from the above description, the pest type identification method provided in this application embodiment can effectively improve the accuracy and effectiveness of pest type identification by using the pest type identification result data corresponding to the real scene image data to iteratively optimize the second convolutional neural network.

[0293] To further improve the accuracy of pest type identification by transforming the waveform into a two-dimensional recursive image, the second convolutional neural network in the pest type identification method provided in this application is an InceptionV3 network.

[0294] In other words, after transforming a single time series data point into a two-dimensional recursive graph, the InceptionV3 network can be used for feature extraction and classification. The InceptionV3 network structure list is shown in Table 2, where the Inception module structure is as follows: Figures 11(a) to 11(c)As shown. In Figures 11(a) to 11(c) In this context, Base represents the input; Filter Concat represents the concatenation at the kernel channel level.

[0295] Table 2

[0296]

[0297]

[0298] Wherein, Conv represents convolution; Conv padded represents padded convolution; Pool represents pooling; inception represents decomposed convolutional layer; Linear represents fully connected layer; and Softmax represents Softmax classification layer.

[0299] To further improve the timeliness and targeting of pest control in the area where the insect trap is located based on the pest type identification results, the pest type identification method provided in this application embodiment specifically includes the following pest types: longhorned flat flour beetle, Turkish flat flour beetle, sawtooth flour beetle, red flour beetle, mixed flour beetle, grain borer, rice weevil, and maize weevil.

[0300] As can be seen from the above description, the pest type identification method provided in this application embodiment can effectively improve the applicability and comprehensiveness of pest type identification, and thus effectively improve the timeliness and pertinence of users' pest control in the area where the insect trap is located based on the pest type identification results.

[0301] From a software perspective, this application also provides a pest landing identification device for performing all or part of the pest landing identification method, see [link to relevant documentation]. Figure 12 The pest identification device specifically includes the following components:

[0302] Infrared data receiving module 10 is used to receive target infrared data containing a dual-channel discrete time series collected in the current self-trapping insect trap.

[0303] The pest falling in identification and control image acquisition module 20 is used to obtain the waveform recognition type corresponding to the discrete time series of the dual channels in the target infrared data based on a preset time series classification model. If the waveform recognition type is a pest detection waveform, it is confirmed that a pest has fallen into the insect trap and the insect trap is controlled to acquire images of the pest that has fallen into it.

[0304] The embodiments of the pest falling into the identification device provided in this application can be used to execute the processing flow of the pest falling into the identification method embodiments described above. Its functions will not be repeated here, but can be referred to the detailed description of the pest falling into the identification method embodiments described above.

[0305] The pest entry detection component of the aforementioned pest entry detection device can be executed on a server or on a client device. The choice depends on the processing power of the client device and the limitations of the user's usage scenario. This application does not impose any limitations in this regard. If all operations are completed on the client device, the client device may further include a processor for the specific processing of pest entry detection.

[0306] The aforementioned client device may have a communication module (i.e., a communication unit) that can communicate with a remote server to achieve data transmission. The server may include a server on the task scheduling center side; in other implementation scenarios, it may also include a server on an intermediate platform, such as a server on a third-party server platform that has a communication link with the task scheduling center server. The server may include a single computer device, a server cluster consisting of multiple servers, or a distributed server structure.

[0307] The server and the client device can communicate using any suitable network protocol, including those not yet developed as of the date of this application. Such network protocols may include, for example, TCP / IP, UDP / IP, HTTP, HTTPS, etc. Furthermore, such network protocols may also include RPC (Remote Procedure Call Protocol) and REST (Representational State Transfer Protocol) protocols used on top of the aforementioned protocols.

[0308] As can be seen from the above description, the pest entry identification device provided in this application embodiment can effectively identify and distinguish whether the object falling into the insect trap is a pest, and can effectively improve the accuracy and reliability of pest entry identification. It can also control the insect trap to perform timely and effective image acquisition of the pest that has fallen into it only when it is confirmed that a pest has fallen in. This can effectively reduce the number of invalid acquisitions and frequent starts of the image acquisition device in the insect trap, thereby reducing the energy consumption of the insect trap and making it less likely for heat to accumulate. Compared with existing insect traps, it weakens the beneficial effects on the growth and reproduction of pests, and also reduces the probability of safety accidents.

[0309] From a software perspective, this application also provides a pest type identification device for performing all or part of the pest type identification method, see [link to relevant documentation]. Figure 13 The pest type identification device specifically includes the following components:

[0310] The trigger detection module 30 is used to obtain a notification message that the waveform identification type of the target infrared data is a pest detection waveform;

[0311] In the trigger detection module 30, the pest type identification device can obtain a notification message from other devices indicating that the waveform identification type of the target infrared data is a pest detection waveform, or it can be set in the same cloud server as the pest landing identification device. If it is the latter, the pest type identification device can directly determine whether to trigger the type identification module 40 based on the identification result of the pest landing identification method in the aforementioned embodiment.

[0312] The type recognition module 40 is used to obtain the pest type recognition result data in the current insect trap based on the preset pest type recognition model.

[0313] The embodiments of the pest type identification device provided in this application can be used to execute the processing flow of the pest type identification method embodiments described above. Its functions will not be repeated here, but can be referred to the detailed description of the pest type identification method embodiments described above.

[0314] The pest type identification part of the pest type identification device can be executed on the server or on the client device.

[0315] As can be seen from the above description, the pest type identification device provided in this application embodiment obtains the pest type identification result data in the current insect trap based on the preset pest type identification model. On the basis of identifying pests falling into the insect trap, it can further realize the automation and efficiency of pest type identification, thereby effectively improving the timeliness and pertinence of the user's pest control in the area where the insect trap is located based on the pest type identification result.

[0316] This application also provides an electronic device, which can be a cloud server. The cloud server may include a processor, a memory, a receiver, and a transmitter. The processor is used to execute the pest landing identification method or pest type identification method mentioned in the above embodiments. The processor and memory can be connected via a bus or other means, taking a bus connection as an example. The receiver can be connected to the processor and memory via wired or wireless means.

[0317] The processor can be a central processing unit (CPU). The processor can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, or combinations of the above types of chips.

[0318] Memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs, non-transitory computer-executable programs, and modules, such as the program instructions / modules corresponding to the pest landing identification method or pest type identification method in the embodiments of this application. The processor executes various functional applications and data processing by running the non-transitory software programs, instructions, and modules stored in the memory, thereby implementing the pest landing identification method or pest type identification method in the above method embodiments.

[0319] The memory may include a program storage area and a data storage area. The program storage area may store the operating system and applications required for at least one function; the data storage area may store data created by the processor, etc. Furthermore, the memory may include high-speed random access memory and non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, the memory may optionally include memory remotely located relative to the processor, which can be connected to the processor via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0320] The one or more modules are stored in the memory, and when executed by the processor, the pest falling identification method or pest type identification method in the embodiment is executed.

[0321] In some embodiments of this application, the user equipment may include a processor, a memory, and a transceiver unit. The transceiver unit may include a receiver and a transmitter. The processor, memory, receiver, and transmitter may be connected via a bus system. The memory is used to store computer instructions, and the processor is used to execute the computer instructions stored in the memory to control the transceiver unit to send and receive signals.

[0322] As one implementation method, the functions of the receiver and transmitter in this application can be implemented by transceiver circuits or dedicated transceiver chips, and the processor can be implemented by dedicated processing chips, processing circuits or general-purpose chips.

[0323] As another implementation approach, the server provided in this application embodiment can be implemented using a general-purpose computer. That is, the program code implementing the processor, receiver, and transmitter functions is stored in memory, and the general-purpose processor implements the processor, receiver, and transmitter functions by executing the code in memory.

[0324] This application also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the aforementioned pest infestation identification method or pest type identification method. The computer-readable storage medium can be a tangible storage medium, such as random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, register, floppy disk, hard disk, removable storage disk, CD-ROM, or any other form of storage medium known in the art.

[0325] This application also provides an embodiment of an insect trap, which is used to collect target infrared data containing a dual-channel discrete time series inside it and send the target infrared data to a cloud server so that the cloud server can execute the pest falling identification method described in the first aspect or implement the pest type identification method described in the second aspect.

[0326] The insect trap is also used to capture images of pests that fall into the trap when or after receiving an image acquisition control command sent by the cloud server.

[0327] As can be seen from the above description, the insect trap provided in this application embodiment can control the insect trap to perform timely and effective image acquisition of the insect that has fallen inside it only when it is confirmed that the insect has fallen in. This can effectively reduce the number of invalid acquisitions and frequent starts of the image acquisition device in the insect trap, thereby reducing the energy consumption of the insect trap and making it less likely for heat to accumulate.

[0328] Existing electronic detectors with image capabilities are designed to be normally open, which leads to excessive power dissipation and heat buildup, potentially contributing to pest growth and reproduction. The insect trap provided in this application embodiment (see...) Figures 14 to 27 The insect trap specifically includes the following components:

[0329] The insect-trapping body 100 includes a control module 150, a trapping module 110, and a monitoring module 120.

[0330] The trapping module 110 includes a first tube 111 and an insect-collecting funnel 112 installed inside the first tube 111. The insect-collecting funnel 112 is coaxially arranged with the first tube 111. In a first direction, one end of the insect-collecting funnel 112 is the first end and the other end is the second end. The diameter of the insect-collecting funnel 112 gradually narrows from the first end to the second end. An insect-passing hole 112b is provided at the second end.

[0331] The monitoring module 120 includes a second tube 121, and an infrared data acquisition component and an insect-swatting component, both installed in the second tube 121. The second tube 121 is coaxially arranged with and connected to the first tube 111. The infrared data acquisition component 124 and the insect-swatting component are both communicatively connected to the control module 150. The infrared data acquisition component 124 includes a first infrared emitting tube 124b and a first infrared receiving tube 124a. The insect-passing hole 112b is located close to the infrared data acquisition component 124. The projection 112b' of the insect-passing hole 112b in a first plane is located within the range of light emitted by the first infrared emitting tube 124b that the first infrared receiving tube 124a can receive. The first plane is perpendicular to the first direction, and the axis of the first infrared emitting tube 124b and the axis of the first infrared receiving tube 124a are both located in this plane. The first direction is the axial direction of the first tube 111.

[0332] In this embodiment, the insect passage hole 112b is positioned close to the infrared data acquisition component 124; that is, the second end of the insect collecting funnel 112 is positioned closer to the infrared data acquisition component 124 than the first end of the funnel. The insect-capturing component refers to a component for capturing insects. The edge of the insect collecting funnel 112 is preferably snapped between the first tube 111 and the second tube 121. The diameter of the insect passage hole 112b is 2mm to 10mm (preferably 4mm). Figure 21 As shown, the range between the solid lines on both sides of the projection 112b' of the insect hole 112b in the first plane is the range of light emitted by the first infrared emitting tube 124b, and the range between the dashed lines on both sides of the projection 112b' of the insect hole 112b in the first plane is the range of light emitted by the first infrared receiving tube 124a that can be received by the first infrared emitting tube 124b.

[0333] Preferably, the insect trap provided in this application embodiment further includes a grate head 400, a cable 200 (which can be a thicker bus or a thinner wire), and a wireless communication device 300, all made of tensile and wear-resistant material. The first tube 111 includes a cap section 111a and a trapping section 111b connected to each other. The trapping section 111b is connected to the second tube 121. The cap section 111a is located at the end of the trapping section 111b away from the second tube 121. The cap section 111a is preferably a hollow conical structure, and the trapping section 111b is preferably a circular hollow plastic tube. Several insect-attracting holes 111ba with a diameter not exceeding 2.5 mm are distributed on the trapping section 111b; the grate head 400... The cable 200 is fixed at the end of the cap section 111a. One end of the cable 200 passes through the gland 400, enters the insect trap body 100 through the cap section 111a, and connects with the electrical components inside the insect trap body 100. The wireless communication device 300 is installed at the other end of the cable 200. The trapping module 110 also includes a cable conversion board 114 and a light trapping element 115 installed on the cable conversion board 114. The light trapping element 115 is preferably a surface-mount LED with adjustable wavelength, and the LED is set to the wavelength with the best trapping effect. Both the cable conversion board 114 and the light trapping element 115 are installed in the cap section 111a, and the light trapping element 115 is located close to the trapping section 111b. The LED is fixed to the bottom surface of the cable conversion board 114 by soldering. The light trapping element 115 can also be a surface-mount LED with a fixed wavelength or a through-hole LED.

[0334] Preferably, the trapping module 110 further includes an insect passage tube 113, the top end of which is connected to the insect passage hole 112b, and the bottom end of which extends to the infrared data acquisition component 124. Objects falling out of the insect passage hole 112b enter the insect passage tube 113, pass through the insect passage tube 113, and fall into the infrared data acquisition component 124.

[0335] The insect trap provided in this embodiment is placed in a grain pile with the first tube 111 positioned above the second tube 121. Insects in the grain pile enter the first tube 111 and then the insect collection funnel 112. Since the first infrared receiving tube 124a can receive the light emitted by the first infrared emitting tube 124b, the range of light emitted by the first infrared emitting tube 124b covers the insect passage hole 112b in the first direction. After the insect falls out of the insect passage hole 112b on the insect collection funnel 112, it enters the range of light emitted by the first infrared receiving tube 124a that can receive the light emitted by the first infrared emitting tube 124b. The control module 150 receives the signal collected by the infrared data acquisition component 124 and controls the insect-swatting component to take a picture of the insect. In this way, the insect trap can activate the insect-swatting component only when an object falls into the insect passage hole 112b, or choose whether to activate the insect-swatting component, without having to keep the insect-swatting component constantly on. This reduces energy consumption and prevents heat from accumulating. Compared with existing insect traps, it weakens the beneficial effects on the growth and reproduction of pests and also reduces the probability of safety accidents.

[0336] like Figures 21 to 23As shown, optionally, the infrared data acquisition component 124 further includes a second infrared receiver 124d and a second infrared emitter 124e coaxially arranged. The axes of the second infrared receiver 124d and the second infrared emitter 124e are both located within the first plane. The first infrared emitter 124b and the first infrared receiver 124a are coaxially arranged, with the axis of the first infrared receiver 124a being the first axis and the axis of the second infrared receiver 124d being the second axis. The intersection of the first axis and the second axis is located at the center of the projection 112b' of the insect hole 112b in the first plane, and the projection 112b' of the insect hole 112b in the first plane is located within the range where the second infrared receiver 124d can receive the light emitted by the second infrared emitter 124e. It can be understood that both the first axis and the second axis are infinitely extendable straight lines. Increasing the number of infrared receivers and emitters improves infrared detection sensitivity, enabling the control module 150 to receive signals promptly and accurately when an object falls through the insect hole 112b. The intersection of the first axis and the second axis is located at the center of the projection 112b' of the insect hole 112b in the first plane. This ensures that the range of light emitted by the first infrared receiver 124a and the range of light emitted by the second infrared receiver 124e can accurately cover the projection 112b' of the insect hole 112b in the first plane. This prevents the projection 112b' of the insect hole 112b from being biased to one side, thus avoiding a portion of the projection falling outside the light range and preventing a decrease in infrared detection sensitivity when an object falls through the insect hole 112b. Figure 22 In the diagram, the range between the two dashed lines on both sides of the first infrared emitting tube 124b is the range of light emitted by the first infrared emitting tube 124b, and the range between the two dashed lines on both sides of the second infrared emitting tube 124e is the range of light emitted by the second infrared emitting tube 124e. Figure 23 This is a partial top view of one embodiment of the infrared data acquisition component 124 provided in this application. Figure 23 The diameters of the first infrared receiver 124a, the first infrared emitter 124b, the second infrared receiver 124d, and the second infrared emitter 124e are all 5 mm. Figure 23 All values ​​marked in the figure are in millimeters, except for 45.0°.

[0337] Optionally, the first axis is perpendicular to the second axis. In this way, the range in which the first infrared receiver 124a can receive the light emitted by the first infrared emitter 124b and the range in which the second infrared receiver 124d can receive the light emitted by the second infrared emitter 124e overlaps significantly, making it less likely for objects falling through the insect hole 112b to fall outside the light range, thus further improving the infrared monitoring sensitivity.

[0338] Optionally, the distance between the first infrared emitting tube 124b and the projection 112b' of the insect passage 112b in the first plane is greater than the distance between the first infrared receiving tube 124a and the projection 112b' of the insect passage 112b in the first plane, and the distance between the second infrared emitting tube 124e and the projection 112b' of the insect passage 112b in the first plane is greater than the distance between the second infrared receiving tube 124d and the projection 112b' of the insect passage 112b in the first plane. Because the light emitted by the infrared emitter is cone-shaped, the farther away from the infrared emitter, the larger the area covered by the light. Therefore, the distance between the projection 112b' of the wormhole 112b in the first plane and the infrared emitter is relatively far, so the projection of the light emitted by the infrared emitter can more easily cover the projection. The greater the distance between the infrared receiver and the infrared emitter, the smaller the area that the light emitted by the infrared emitter can directly illuminate on the infrared receiver. Therefore, the infrared receiver is placed closer to the projection 112b' of the wormhole 112b in the first plane, so that the projection can more easily fall within the range of light that the infrared receiver can receive.

[0339] Optionally, the insect-swatting assembly includes a camera 125 and a shooting tube 122. A shooting chamber 122a is formed inside the shooting tube 122. A shooting chamber inlet 122b is formed at one end of the shooting tube 122 in the first direction, and a shooting chamber outlet 122c is formed at the other end of the shooting tube 122. The insect passage hole 112b communicates with the shooting chamber inlet 122b. In the first direction, the infrared data acquisition component 124 is located between the insect collecting funnel 112 and the shooting chamber inlet 122b. The camera 125 is disposed close to the shooting chamber inlet 122b and is fixed to the inner wall of the shooting chamber 122a. The camera 125 is communicatively connected to the control module 150 to capture images into the shooting chamber 122a. Thus, when an insect falls out of the insect passage 112b, triggering the infrared data acquisition component 124, the control module 150 receives the signal from the infrared data acquisition component 124. As the object falls within the shooting chamber 122a, the control module 150 controls the camera 125 to capture images into the shooting chamber 122a. The object's stay within the shooting chamber 122a provides sufficient time for the camera 125 to capture images. Preferably, the camera 125 is embedded in a groove at the top of the shooting chamber 122a. The insect-catching component also preferably includes a supplementary light, which and the camera 125 are turned on and off simultaneously.

[0340] Optionally, the diameter of the shooting chamber 122a gradually decreases from the shooting chamber inlet 122b to the shooting chamber outlet 122c. Because the diameter of the shooting chamber 122a gradually decreases from the shooting chamber inlet 122b to the shooting chamber outlet 122c, an object falling into the shooting chamber 122a and before falling out of the shooting chamber outlet 122c will contact the side wall of the shooting chamber 122a. The frictional force slows down the falling speed, giving the camera 125 sufficient time to capture the image.

[0341] In this embodiment, preferably, the control module 150 is a main control circuit board, which preferably includes six main parts: a power supply module, a microprocessor, a communication module, a multi-dimensional data acquisition module, a camera 125 driving module, and an infrared signal processing module. The power supply module is used to output appropriate voltage and current to power the system; the microprocessor is used to run processing programs, drive each acquisition module, and execute instructions issued by the host computer; the communication module is used to send information from the microprocessor to the bus and obtain instructions from the bus to the microprocessor; the multi-dimensional data acquisition module is used to collect multi-dimensional data such as carbon dioxide concentration, oxygen concentration, temperature, and humidity in the grain pile; the camera 125 driving module is used to output appropriate driving power and control signals to the camera 125 module and read image information from the camera 125 module; the infrared signal processing module is used to collect infrared waveform data of pests to identify whether pests have fallen, thereby determining whether to activate the camera 125 to obtain images.

[0342] In this embodiment of the application, preferably, the infrared data acquisition component 124 further includes an infrared data acquisition circuit board 124c. The first infrared emitting tube 124b and the first infrared receiving tube 124a are both mounted on the infrared data acquisition circuit board 124c. The infrared data acquisition circuit board 124c is connected to the main control circuit board through pin headers and socket headers. The pin headers and socket headers not only play a role in fixing the position, but also realize the transmission of power and signals.

[0343] Existing traps lack pest removal mechanisms, requiring manual removal of pests. This increases the complexity of the monitoring system operation and labor costs, especially as the number of traps deployed in grain warehouses increases, making pest monitoring and removal more difficult and time-consuming. Optionally, in this embodiment, the monitoring module 120 further includes a cleaning component installed within the second tube 121. The cleaning component 123 includes a housing 123b, an insect-carrying platform 123a, and a drive component 123d. The housing 123b is connected to the imaging tube 122. The drive component 123d is installed on the housing 123b, and the insect-carrying platform 123a is installed on the drive component 123d. The drive component 123d drives the insect-carrying platform 123a to reciprocate between a first position and a second position. In the first position, the insect-carrying platform 123a engages with the imaging chamber outlet 122c; in the second position, a gap is formed between the insect-carrying platform 123a and the imaging chamber outlet 122c. The drive component 123d is communicatively connected to the control module 150. Thus, after the camera 125 finishes photographing the object that has fallen into the shooting chamber 122a, the control module 150 can control the drive component 123d to start, moving the insect-carrying platform 123a from the first position to the second position, allowing the object to fall out of the shooting chamber outlet 122c. Before proceeding with the next round of photographing the objects inside the shooting chamber 122a, the control module 150 restarts the drive component 123d, moving the insect-carrying platform 123a from the second position to the first position, thereby achieving automatic cleaning. The entire structure is relatively simple and also reduces labor costs. The housing 123b is connected to the shooting chamber 122a by screws and nuts.

[0344] Optionally, the drive component 123d includes a drive component 123d body and a first output shaft mounted on the drive component body. The drive component 123d is a rotation drive component. The first output shaft is coaxially arranged with the second tube 121. The cleaning assembly 123 further includes a threaded rod 123c mounted on the first output shaft. A threaded hole is formed on the housing 123b, and the threaded rod 123c engages with the threaded hole. Thus, when the first output shaft of the drive component 123d rotates, it drives the threaded rod 123c to rotate. Since the threaded rod 123c engages with the threaded hole on the housing 123b, the rotation of the threaded rod 123c will drive the threaded rod 123c, the drive component 123d, and the insect-carrying platform 123a to move in the first direction. The change in the rotation direction of the first output shaft causes the rotation direction of the threaded rod 123c to change, and the corresponding drive component 123d and insect-carrying platform 123a also change their movement direction in the first direction. Of course, the driving component 123d can also be a linear driving component, which connects the insect-carrying platform 123a to the transmission rod of the linear driving component. Under the drive of the transmission rod of the linear driving component, the insect-carrying platform 123a moves between a first position and a second position. Preferably, the rotary driving component is a DC motor.

[0345] Optionally, the cleaning component 123 further includes a trigger, and a first limit switch and a second limit switch, both mounted on the housing 123b. Both the first and second limit switches are communicatively connected to the control module 150. The trigger is mounted on the drive component 123d and moves with the insect-carrying platform 123a, such that when the insect-carrying platform 123a is in a first position, the trigger contacts the first limit switch, and when the insect-carrying platform 123a is in a second position, the trigger contacts the second limit switch. Thus, when the insect-carrying platform 123a moves to the first position, the limit switch triggers the first limit switch, automatically stopping the movement of the insect-carrying platform 123a towards the imaging chamber 122a; and when the insect-carrying platform 123a moves to the second position, the limit switch triggers the second limit switch, automatically stopping the movement of the insect-carrying platform 123a away from the imaging chamber 122a.

[0346] Some existing probe traps with cleaning functions have structural design flaws, making them ineffective at cleaning live insects with strong climbing abilities. For example, using a stepper motor to control the flipping of the insect-falling plate cannot clean some pests with strong climbing abilities, thus affecting the accuracy of monitoring. In addition, some probe traps use electric strip brushes to clean pests. When cleaning live insects, the pests may grab the brush, leading to cleaning failure. In this embodiment, optionally, the drive component 123d further includes a second output shaft. The first output shaft and the second output shaft are coaxially arranged, and in the first direction, the first output shaft and the second output shaft are located at opposite ends of the drive component 123d body. The insect-carrying platform 123a is fixed to the second output shaft. In other words, the first and second output shafts are simultaneously driven by the drive unit 123d. After the imaging operation of the object in the imaging chamber 122a is completed, the control module 150 activates the drive unit 123d. The drive unit 123d drives the first and second output shafts to rotate, causing the insect-carrying platform 123a to move from the first position to the second position. Simultaneously, it rotates the insect-carrying platform 123a, causing the object falling onto it to be thrown off under centrifugal force. After the object is thrown off, the control module 150 causes the drive unit 123d to change the rotation direction of the first and second output shafts, and moves the insect-carrying platform 123a from the second position to the first position. Preferably, the first output shaft is connected to the first reduction gear set, the threaded rod 123c is mounted on the output shaft of the first reduction gear set, the second output shaft is connected to the second reduction gear set, and the insect-carrying platform 123a is mounted on the output shaft of the second reduction gear set.

[0347] Optionally, the insect-carrying platform 123a includes a top surface facing the imaging chamber 122a, on which a first protruding ridge 123aa and a second protruding ridge 123ac are formed. The length directions of the first protruding ridge 123aa and the second protruding ridge 123ac are both parallel to the top surface, and the first protruding ridge 123aa and the second protruding ridge 123ac intersect perpendicularly. A protruding cone 123ab is also formed at the intersection of the first protruding ridge 123aa and the second protruding ridge 123ac.

[0348] Preferably, the control module 150 also includes a motor drive board. Each limit switch can control the movement amplitude of the drive component 123d and the insect-carrying platform 123a. When the movement reaches a preset position, the corresponding limit switch will be triggered, providing a control signal to the motor drive board. A wiring hole 123ba is formed on the housing 123b. The power line of the motor drive board passes through the wiring hole 123ba on the housing 123b and connects to the infrared data acquisition circuit board 124c. The housing 123b has two parts, which are connected to each other by a snap-fit ​​mechanism.

[0349] Optionally, the insect-trapping body 100 further includes a collection tube 130 coaxially arranged with the first tube 111. The collection tube 130 is detachably connected to the end of the second tube 121 away from the first tube 111, and a collection cavity communicating with the first tube 111 is formed in the collection tube 130. In use, the insect trap provided in this embodiment is placed in a grain pile, with the first tube 111 positioned above the second tube 121, and the second tube 121 positioned above the collection tube 130. When an object falling onto the insect-carrying platform 123a is thrown off by centrifugal force, the object falls into the collection tube 130. After the insect trap has been used for a period of time, the collection tube 130 can be separated from the second tube 121 to empty the object from the collection cavity. In this embodiment, preferably, the cap section 111a, the trapping section 111b, the second tube 121, and the collection tube 130 are sequentially threaded together.

[0350] Optionally, the insect-trapping body 100 further includes a wiring conduit 140, which is coaxially arranged with the first tube 111. One end of the wiring conduit 140 extends to the end of the first tube 111 away from the second tube 121, and the other end extends to the insect-collecting funnel 112. A wire-passing hole 112a is formed on the insect-collecting funnel 112, and the wiring conduit 140 communicates with the wire-passing hole 112a. By providing the wiring conduit 140, it is convenient to run wires within the insect-trapping body 100, making it less likely for the wires to interfere with the components within the insect-trapping body 100, and making it less likely for the wires to obstruct the movement of pests or other objects within the insect-trapping body 100. When the first tube 111 includes a cap section 111a and a trapping section 111b, the cable 200 passes through the gland 400 and enters the cap section 111a. The cable 200 is connected to the cable conversion plate 114, which is also connected to a wire. At the cable conversion plate 114, the thicker cable 200 is converted into a thinner wire. The wire enters the cable routing tube 140 and extends along the cable routing tube 140 to the main control circuit board and connects to the main control circuit board. In this embodiment, the cable routing tube 140 is preferably a cable routing metal tube, which is a hollow long tube with a diameter of 2mm to 10mm (preferably 4mm). The inside can be used for wires to pass through, and the diameter of the wire through hole 112a is slightly larger than the inner diameter of the wire through hole 123ba. The cable routing metal tube is snapped and fixed to the cable conversion plate 114 through a through hole reserved in the center of the cable conversion plate 114. The cable routing tube 140 can also be made of materials such as plastic. The main control circuit board is fixed to the side of the insect collecting funnel 112 with screws and nuts, and the infrared data acquisition circuit board 124c is fixed between the top of the shooting chamber 122a and the bottom of the insect collecting funnel 112 with screws and nuts.

[0351] To better illustrate the insect trap provided in this application, this application also provides an application example of the insect trap, which is as follows:

[0352] This application example presents an electronic probe trap for pests in grain piles.

[0353] From the outside, the electronic probe trap consists of two parts: the electronic probe body and the wireless communication device. The electronic probe trap body and the wireless communication device are connected by tensile and wear-resistant cables, which provide power to the body and transmit signals.

[0354] The electronic probe body and the cable are connected using a gland connector.

[0355] The main body of the electronic probe consists of a cap section, a trapping section, an electronic monitoring section, and an insect collection section. All connections between the cap section and the trapping section, the trapping section and the electronic monitoring section, and the electronic monitoring section and the insect collection section are threaded.

[0356] The tube cap section is a hollow conical structure with an opening at the top for mounting the gland; the tube cap section contains a cable conversion board, a light-attracting module, and a metal tube for cable routing.

[0357] The cable conduit is a hollow tube with a diameter of 2mm to 10mm (preferably 4mm), which allows wires to pass through.

[0358] The cable conversion board is used to convert the thicker bus on the outside of the probe into thinner wires, which are then connected to the main control circuit board inside the electronic monitoring section through a metal conduit.

[0359] The light-attracting module uses tunable wavelength surface-mount LEDs, which can be used to set the optimal wavelength for the trapping effect.

[0360] The light-attracting module is fixed to the bottom of the cable conversion board by welding.

[0361] The metal conduit for cable routing is secured to the cable conversion board by snapping it into place through a pre-drilled hole in the center.

[0362] The trapping section is a circular hollow plastic tube with several insect-attracting holes no larger than 2.5 mm in diameter.

[0363] The electronic monitoring section comprises the following parts: insect collection funnel, wiring hole, main control circuit board, camera connector, infrared data acquisition circuit board, camera module, shooting chamber, insect carrying platform, and cleaning mechanism.

[0364] The edge of the insect-collecting funnel is engaged between the outer shell of the trapping section and the electronic monitoring section.

[0365] The insect collecting funnel has a wire hole and an insect passage hole in the middle.

[0366] The inner diameter of the cable routing hole is slightly larger than the diameter of the cable routing metal conduit.

[0367] The metal conduit for wiring is inserted into the wiring hole of the insect collecting funnel for fixation, and the wires pass through the wiring hole to connect to the main control circuit board.

[0368] The diameter of the wormhole is 2mm to 10mm (preferably 4mm).

[0369] The main control circuit board is fixed to the side of the insect collection funnel with screws and nuts.

[0370] The infrared data acquisition circuit board is connected to the main control circuit board via pin headers and sockets. The pin headers and sockets not only fix the position, but also enable the transmission of power and signals.

[0371] Use screws and nuts to secure the infrared data acquisition circuit board between the top of the imaging chamber and the bottom of the insect collection funnel.

[0372] A white surface-mount LED is soldered to the bottom of the infrared data acquisition circuit board to provide supplementary lighting for the shooting chamber when the camera is shooting.

[0373] The camera module is embedded in a groove at the top of the shooting chamber.

[0374] The insect-carrying platform and the cleaning mechanism are connected by a rotating shaft.

[0375] The cleaning mechanism is connected to the shooting chamber by screws and nuts.

[0376] The main control circuit board comprises six main parts: a power supply module, a microprocessor, a communication module, a multi-dimensional data acquisition module, a camera driver module, and an infrared signal processing module. The power supply module outputs appropriate voltage and current to power the system; the microprocessor runs the processing program, drives the various acquisition modules, and executes instructions from the host computer; the communication module sends information from the microprocessor to the bus and receives instructions from the bus back to the microprocessor; the multi-dimensional data acquisition module collects multi-dimensional data from the grain pile, such as carbon dioxide concentration, oxygen concentration, temperature, and humidity; the camera driver module outputs appropriate drive power and control signals to the camera module and reads image information from the camera module; and the infrared signal processing module collects infrared waveform data of pests to identify whether pests have fallen, thereby determining whether to activate the camera to acquire images.

[0377] The infrared data acquisition circuit board uses two pairs of infrared light emitting tubes and infrared receiving tubes with a diameter of 5mm.

[0378] The infrared emitting tubes and infrared receiving tubes are placed facing each other on the same straight line, with the two pairs of tubes placed perpendicularly to each other.

[0379] The insect passage hole in the insect collecting funnel passes between the infrared emitting tube and the receiving tube, and is completely covered by the light emitted by the infrared emitting tube.

[0380] The cleaning mechanism comprises a rotating shaft, a cleaning mechanism housing, wiring holes, a first reduction gear set, a drive circuit board, limit switches, a DC motor, a second reduction gear set, and a threaded rod. The rotating shaft engages with the bottom of the insect-carrying platform to drive its rotation. The DC motor's rotating shaft, through reduction gear sets one and two, drives both the rotating shaft and the threaded rod. The rotation of the threaded rod allows for relative movement between the internal structure and the cleaning mechanism housing. Limit switches control the range of motion of the internal structure; when a preset position is reached, the limit switches are triggered, providing a control signal to the motor drive board. The power cable of the motor drive board passes through the wiring holes and connects to the infrared data acquisition circuit board. The cleaning mechanism housing consists of two parts connected by a snap-fit ​​mechanism.

[0381] The diameters and lengths of the various parts of the electron probe body in this application example may also be realized in this application example.

[0382] The diameter of the metal conduit used for wiring can also be used to achieve this application example.

[0383] The wiring conduit is made of metal, but other materials (such as plastic) can also be used to achieve this application example.

[0384] The bus uses thicker wires, so a cable adapter board is required. However, this application example can also be achieved using thinner wires, and it is also possible to achieve this application example without using a cable adapter board if the bus uses thinner wires.

[0385] The light-attracting module uses a surface-mount LED with adjustable wavelength, but this application example can also be achieved using a surface-mount LED with a fixed wavelength or a through-hole LED.

[0386] This application example can also be achieved using adhesives for cable conduits and cable conversion boards.

[0387] The trapping section is a transparent hollow plastic tube with an inner diameter of 44mm and an outer diameter of 50mm. This application example can also be achieved by scaling its inner or outer diameter, changing its transparency, or replacing its material.

[0388] The diameter of the insect-attracting hole is 2mm, and the angle is 45 degrees upward. This application example can also be achieved by scaling up or down the diameter of the insect-attracting hole or changing its angle.

[0389] The edge of the insect-collecting funnel is fixed between the outer shell of the trapping section and the electronic monitoring section using a snap-fit ​​method, but this application example may also be achieved by using an adhesive.

[0390] The inner diameter of the wiring hole is 4.1mm, and this application example can also be achieved by scaling its inner diameter.

[0391] The diameter of the wormhole is 4mm, and this application example can also be achieved by scaling its diameter.

[0392] The main control circuit board and the insect collection funnel are connected using screws and nuts, but this application example may also be achieved by using adhesives, fastening, or snap-fitting.

[0393] The infrared data acquisition circuit board and the main control circuit board are connected using pin headers and female headers, but wire connections can also be used to achieve the same application.

[0394] The infrared data acquisition circuit board is connected to the imaging chamber and the insect collection funnel using screws and nuts, but this application example can also be achieved by using adhesive, fastening, or snap-fitting.

[0395] The supplementary LED used is a white surface-mount LED, but other colored surface-mount LEDs or through-hole LEDs may also achieve this application example.

[0396] The cleaning mechanism and the shooting chamber are connected by screws and nuts, but this application example may also be achieved by adhesive, fastening, or snap-fitting.

[0397] The infrared emitting and receiving tubes used have a diameter of 5 mm, but this application example can also be achieved by using infrared emitting and receiving tubes with other diameters (e.g., 3 mm).

[0398] The relative positions of the infrared transmitter and receiver on the infrared data acquisition circuit board and the insect hole are shown, but this application example can also be achieved by translating or rotating the infrared transmitter and receiver.

[0399] The surface of the insect-carrying platform has two mutually perpendicular ridges and a cone in the middle, but this application example may also be achieved without ridges or cones, or with other shapes of ridges (such as wavy shapes).

[0400] This application example can also be achieved without using a reduction gear.

[0401] Limit switches are used as the control signal for the motion amplitude, but pressure sensors or infrared sensors can also be used to achieve this application example.

[0402] A DC motor is used, but this application example could also be achieved using a stepper motor.

[0403] Based on the above embodiments of insect traps and cloud servers, this application also provides an embodiment of a pest infestation identification system, see [link to embodiment]. Figure 28The pest identification system specifically includes the following:

[0404] The cloud server and the insect trap are interconnected; the cloud server is used to execute the aforementioned pest falling identification method or the aforementioned pest type identification method.

[0405] As can be seen from the above description, the pest entry identification system provided in this application embodiment can control the insect trap to perform timely and effective image acquisition of the pest that has fallen inside it only when it is confirmed that a pest has fallen in. This can effectively reduce the number of invalid acquisitions and frequent starts of the image acquisition device in the insect trap, thereby reducing the energy consumption of the insect trap and making it less likely for heat to accumulate.

[0406] To further illustrate the above embodiments, this application also provides an electronic probe trap with automatic cleaning function and a real-time pest monitoring system implemented using the above-mentioned insect trap (i.e., electronic probe trap) and various methods, which relates to the field of grain storage pest monitoring, namely how to trap and accurately detect pests in grain piles during the storage of raw grain.

[0407] The current electronic probe traps and systems for monitoring pests in grain piles have the following problems:

[0408] 1. Some of the probe traps lack pest removal mechanisms, requiring manual removal of pests. This increases the complexity of the monitoring system and labor costs, especially as the number of traps deployed in grain warehouses increases, making pest monitoring and removal more difficult and time-consuming.

[0409] 2. Some counting-type insect traps with cleaning functions require the installation of negative pressure vacuum suction tubes inside the grain pile. When using an external suction pump to clean the insects from inside the traps via these tubes, this deployment method limits the system's flexibility, and the suction tubes are prone to clogging. If clogging occurs, manual cleaning is required, increasing labor costs. Furthermore, this counting method itself has a certain degree of error, and the suction operation is not performed periodically. After the insects trapped inside the tubes die, their bodies lose water, and the negative pressure suction process can damage them, making it difficult for grain warehouse personnel to inspect and identify the insect species, potentially affecting pest control.

[0410] 3. Some probe traps with cleaning functions have structural design flaws that prevent them from effectively cleaning live insects with strong climbing abilities. For example, the method of using a stepper motor to control the flipping of the insect-falling plate cannot clean some pests with strong climbing abilities, thus affecting the accuracy of monitoring. In addition, some probe traps use motorized strip brushes to clean pests. When cleaning live insects, the pests may grab the brush, leading to cleaning failure.

[0411] 4. The electronic probe traps currently available cannot effectively distinguish between grain debris or impurities and pests, which can lead to incorrect pest counting. For some electronic probe traps that use cameras, this can cause the cameras to be activated accidentally. Frequent activation of the cameras can generate a lot of heat, which can help pests reproduce and grow in the grain pile, and also poses a significant fire safety hazard.

[0412] 5. Limited data collection dimensions: Common electronic probe traps are designed only to count the number of pests trapped. A few probe traps collect temperature and humidity data inside the grain pile, but these are insufficient to fully reflect the state of the ecosystem inside the grain pile and cannot effectively predict the occurrence of pests.

[0413] 6. Remote upgrades of the probe trap driver are not possible. This leads to difficulties during program version iterations. When improvements to the probe trap's functionality or performance are needed, the driver cannot be updated remotely; on-site upgrades by dedicated personnel or technical teams are required. This involves investment of human and material resources, increasing maintenance and upgrade costs.

[0414] To address the aforementioned issues, the main control circuit board in the electronic probe trap provided in this application example functions as follows: Figure 29 As shown, when pests or grain debris fall through the insect hole, the infrared data acquisition circuit board continuously monitors the voltage of the infrared receiver to obtain a continuous voltage sequence. Figure 30 This is an example of a voltage sequence for pests. Figure 31 This is an example of a voltage sequence for grain debris.

[0415] The system's microcontroller transmits the acquired voltage sequence data to the cloud server via the communication module. On the cloud server, a machine learning algorithm is called to classify the waveform types. The first step is to determine whether the waveform is triggered by pests or grain debris. If it is triggered by pests, the second step is to classify the pest waveforms and give the probability of belonging to a specific pest species.

[0416] The process of the pest infestation detection algorithm is as follows: Figure 32 As shown, the raw data first undergoes data preprocessing, and then passes through a feature extraction and classification network to classify the waveform data.

[0417] Then, raw data was collected and labeled: the actual warehouse data came from a data collection experiment conducted in July 2021 at warehouse No. 12 of the Lu'an Modern Logistics Center in Anhui Province. A monitoring system consisting of 51 electronic probe traps was installed in the experiment. During the experiment, the actual number of pests trapped in the traps and the amount of data detected by the traps were manually checked and recorded eight times: July 24-27, July 27-30, July 30-August 3, August 5-6, August 6-9, August 9-10, August 10-11, and August 11-13. Furthermore, because the experimental warehouse was located in a medium-temperature, high-humidity grain storage area, one of my country's seven major grain storage ecological zones, and it was during the summer, a large number of booklice were present in the warehouse. However, booklice are not within the effective monitoring range of the traps. Therefore, booklice were a primary consideration in the raw data collection and labeling process.

[0418] The raw data obtained during the experiment are shown in Table 3.

[0419] Table 3

[0420]

[0421]

[0422] When manually labeling the raw data, the distinction is made between the waveforms of data collected when the trap captures pests and the waveforms of data collected when the trap is accidentally triggered. A comparison of several typical data types with normal pest data is provided. Figure 33 and Figures 34(a) to 34(d) As shown.

[0423] Based on the data collected from the electronic probe traps in the actual storage area, during the manual data annotation process, the normal pest detection waveforms with obvious characteristics were classified into Category 1, the waveforms triggered by booklice (Figure 34(a)) into Category 2, and the double-straight-line waveforms generated by repeated pest triggering (Figure 34(b)) into Category 3. Data waveforms generated by accidental triggering in other situations (Figure 34(c) and Figure 34(d)) were classified into Category 4 because they were extreme cases, had small data volumes, or lacked obvious patterns like Figure 34(d).

[0424] Based on the above classification rules, data was filtered and manually labeled. The overview of the constructed real-warehouse pest detection waveform dataset is shown in Table 4.

[0425] Table 4

[0426] Waveform categories Annotated data volume 1 9891 2 7712 3 18502 4 1339

[0427] Model training process: The waveform feature fusion classification network was trained using the real-time warehouse detection dataset obtained from the real-time warehouse data collection experiment. The training environment is shown in Table 5.

[0428] Table 5

[0429]

[0430] During training, the parameters of the soft min layer are first initialized. The initialization method involves applying a sliding window method to the original data in the dataset to obtain subsequences with the same length as each block in the soft min layer. Then, k-means clustering is applied to these subsequences, and the cluster centers obtained are used as the initial parameters of the soft min layer. Next, the dataset is divided into training and validation sets in a 7:3 ratio, and the network is trained on the training set with a learning rate of 0.01 and 30 training epochs.

[0431] After classifying the pest waveforms using the above identification algorithm, the probability of belonging to four categories can be obtained. The category with the highest probability is selected as the identification result, where category 1 is considered a normal waveform, and categories 2, 3, and 4 are considered abnormal waveforms. For the waveforms in category 1, the pest species will be identified.

[0432] In this application example, the process of the pest type identification algorithm is as follows:

[0433] Dataset Establishment for Pest Species Identification: The data in this dataset consists of waveform datasets generated from actual detection of eight common stored grain pests using electronic probe traps. Pests were manually thrown into the traps during data collection. The data volume for each pest category is shown in Table 6.

[0434] Table 6

[0435] Pest categories Data quantity Longhorn Flat Rice Thief 871 Turkish flatbread 885 Sawmiller 912 Red-eared Valley Beetle 785 Miscellaneous Valley Thief 1122 grain beetle 984 Rice Elephant 876 Corn Elephant 865

[0436] After transforming the original data into a two-dimensional multi-channel optimized recursive graph, its image size is (112, 112, 3). During training, the InceptionV3 network uses the cross-entropy loss function; the optimizer is the Adam optimizer; the initial learning rate is 0.001, decreasing by 10% every 5000 iterations; the input batch size is 16; and the training set to test set ratio is 8:2. The equipment and environment used for network training are shown in Table 7.

[0437] Table 7

[0438]

[0439]

[0440] In addition, to address the overfitting issue during network training, the pest species identification dataset underwent manual data cleaning to remove data with excessively large intra-class discrepancies. Taking the long-horned flat grain beetle (CF) as an example, examples of the retained and removed data are shown below. Figure 35 and Figure 36 As shown.

[0441] After data cleaning, the following two methods can be used to augment the data in the dataset:

[0442] (1) Shift the original data 5 units to the left or right, and after the shift, fill the missing sampling points on both sides with the sequence endpoint values ​​to obtain new data.

[0443] (2) Flip the two channels of the original data to obtain a new set of data.

[0444] Furthermore, since the red flour beetle (TC) and the mixed flour beetle (TCH) among the pests to be classified are too similar in appearance, and the basic resolution of the detection method using infrared photoelectric sensors is already low, this method merges the two pests, the red flour beetle (TC) and the mixed flour beetle (TCH), into a single class called the flour beetle for network training in order to further reduce overfitting and network complexity.

[0445] In summary, this application presents a novel algorithm for counting pests in stored grain based on an electronic probe trap. The algorithm first classifies the waveform of the raw data detected by the trap, specifically including:

[0446] (1) A waveform dataset of real-world data collected by the trap was established, and the raw data was manually divided into four typical waveforms based on their causes.

[0447] (2) A trap detection waveform feature extraction and classification network that integrates global and local features was designed to classify waveform data within the dataset.

[0448] The counting algorithm then counts the number of various waveforms collected by the trap and uses linear regression to predict the number of pests trapped by the trap.

[0449] This application also proposes a method for identifying stored grain pests based on an electronic probe trap. The method first converts the raw pest data collected by the trap into two-dimensional images using an optimized multi-channel recursive graph algorithm. Then, it uses the InceptionV3 network in the classic convolutional neural network structure to extract features and classify the converted two-dimensional images, thus realizing the identification of eight common stored grain pests.

[0450] This application example analyzes infrared waveforms triggered by pests, enabling not only the monitoring of pest occurrences but also the identification of pest species, achieving real-time and effective monitoring of stored grain pests. Furthermore, the system records historical data and generates reports, providing crucial information for pest management and playing a significant role in the monitoring and control of stored grain pests.

[0451] In addition, the following alternatives may also be found in the embodiments and application examples mentioned above in this application:

[0452] 1. At the data acquisition end, increase the number of infrared photodiodes in the trap, such as from the existing two pairs of infrared photodiodes to three pairs, four pairs, etc., to improve the dimensionality of the original pest data collected by the trap. Afterwards, the counting and species identification of pests trapped by the trap can still be realized according to the idea of ​​this invention.

[0453] 2. When classifying waveforms in the data collected from the trap's storage area, different time-series feature extraction methods, such as wavelet analysis, transformer, and LSTM, or different classifiers, such as SVM, XGBoost, and KNN, can be used to achieve waveform classification for the trap. Furthermore, when performing linear regression on the number of waveforms in different categories, different regression algorithms, such as logistic regression and ridge regression, can be used.

[0454] 3. When identifying pest species trapped by traps, different methods are used to transform the one-dimensional raw time-series data collected by the traps into two-dimensional images, such as: Gram angle field (GAF) method, Markov transfer field method, graphic difference method, relative position matrix method, etc. Then, different deep learning networks are used to extract features and classify the two-dimensional images, such as: ResNet, VGG, DenseNet, etc.

[0455] Those skilled in the art will understand that the exemplary components, systems, and methods described in conjunction with the embodiments disclosed herein can be implemented in hardware, software, or a combination of both. Whether implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application. When implemented in hardware, it can be, for example, electronic circuits, application-specific integrated circuits (ASICs), appropriate firmware, plug-ins, function cards, etc. When implemented in software, the elements of this application are programs or code segments used to perform the required tasks. The programs or code segments can be stored in a machine-readable medium or transmitted over a transmission medium or communication link via data signals carried on a carrier wave.

[0456] It should be clarified that this application is not limited to the specific configurations and processes described above and shown in the figures. For the sake of brevity, detailed descriptions of known methods are omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of this application is not limited to the specific steps described and shown. Those skilled in the art can make various changes, modifications, and additions, or change the order of steps, after understanding the spirit of this application.

[0457] In this application, features described and / or illustrated for one embodiment may be used in the same or similar manner in one or more other embodiments, and / or combined with or in place of features of other embodiments.

[0458] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Various modifications and variations can be made to the embodiments of this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. An insect trap, characterized in that, The insect trap is used to collect target infrared data containing a dual-channel discrete time series inside it, and send the target infrared data to a cloud server. The insect trap is also used to capture images of pests that fall into the trap when or after receiving an image acquisition control command sent by the cloud server. The insect trap includes an insect trapping body, which includes a control module, a trapping module, and a monitoring module. The trapping module includes a first tube and an insect-collecting funnel installed inside the first tube. The insect-collecting funnel is coaxially arranged with the first tube. In a first direction, one end of the insect-collecting funnel is the first end and the other end is the second end. The diameter of the insect-collecting funnel gradually narrows from the first end to the second end. An insect passage hole is provided at the second end. The monitoring module includes a second tube, and an infrared data acquisition component and an insect-swatting component, both installed inside the second tube. The second tube is coaxially arranged with and connected to the first tube. Both the infrared data acquisition component and the insect-swatting component are communicatively connected to the control module. The infrared data acquisition component is used to acquire target infrared data containing a dual-channel discrete time series inside the insect trap according to the instructions of the control module. The insect-swatting component is used to acquire images of pests falling into the insect trap according to the instructions of the control module. The infrared data acquisition component includes a first infrared emitting tube and a first infrared receiving tube. The insect-passing hole is located close to the infrared data acquisition component. The projection of the insect-passing hole in a first plane is located within the range of light emitted by the first infrared emitting tube that the first infrared receiving tube can receive. The first plane is perpendicular to the first direction, and the axis of the first infrared emitting tube and the axis of the first infrared receiving tube are both located in this plane. The first direction is the axial direction of the first tube. The infrared data acquisition component further includes a second infrared receiver and a second infrared emitter coaxially arranged. The axis of the second infrared receiver and the axis of the second infrared emitter are both located in the first plane. The first infrared emitter and the first infrared receiver are coaxially arranged. The axis of the first infrared receiver is the first axis, and the axis of the second infrared receiver is the second axis. The intersection of the first axis and the second axis is located at the center of the projection of the insect hole in the first plane, and the projection of the insect hole in the first plane is located within the range of light emitted by the second infrared emitter that the second infrared receiver can receive. The insect-catching assembly includes a camera and a shooting tube. A shooting chamber is formed inside the shooting tube. An inlet for the shooting chamber is formed at one end of the shooting tube in the first direction, and an outlet for the shooting chamber is formed at the other end of the shooting tube. The insect passage hole communicates with the inlet for the shooting chamber. In the first direction, the infrared data acquisition assembly is located between the insect collecting funnel and the inlet for the shooting chamber. The camera is positioned close to the inlet for the shooting chamber and is fixed to the inner wall of the shooting chamber. The camera is communicatively connected to the control module to capture images into the shooting chamber.

2. The insect trap according to claim 1, characterized in that, The first axis is perpendicular to the second axis.

3. The insect trap according to claim 2, characterized in that, The distance between the first infrared emitting tube and the projection of the insect passage hole in the first plane is greater than the distance between the first infrared receiving tube and the projection of the insect passage hole in the first plane, and the distance between the second infrared emitting tube and the projection of the insect passage hole in the first plane is greater than the distance between the second infrared receiving tube and the projection of the insect passage hole in the first plane.

4. The insect trap according to claim 1, characterized in that, The diameter of the shooting chamber gradually decreases from the entrance to the exit.

5. The insect trap according to claim 1, characterized in that, The monitoring module further includes a cleaning component installed inside the second tube. The cleaning component includes a housing, an insect-carrying platform, and a drive unit. The housing is connected to the imaging tube, the drive unit is installed on the housing, and the insect-carrying platform is installed on the drive unit. The drive unit is used to drive the insect-carrying platform to reciprocate between a first position and a second position. In the first position, the insect-carrying platform engages with the imaging chamber outlet, and in the second position, a gap is formed between the insect-carrying platform and the imaging chamber outlet. The drive unit is communicatively connected to the control module.

6. The insect trap according to claim 5, characterized in that, The driving component includes a driving component body and a first output shaft mounted on the driving component body. The driving component is a rotary driving component. The first output shaft is coaxially arranged with the second tube. The cleaning assembly also includes a threaded rod mounted on the first output shaft. A threaded hole is formed on the housing. The threaded rod engages with the threaded hole.

7. The insect trap according to claim 6, characterized in that, The cleaning component further includes a trigger and a first limit switch and a second limit switch, both installed on the housing. The first limit switch and the second limit switch are both used to communicate with the control module. The trigger is installed on the drive body and is used to move with the insect-carrying platform so that the trigger contacts the first limit switch when the insect-carrying platform is in a first position, and the trigger contacts the second limit switch when the insect-carrying platform is in a second position.

8. The insect trap according to claim 6, characterized in that, The drive unit further includes a second output shaft, the first output shaft and the second output shaft are coaxially arranged, and the first output shaft and the second output shaft are located at opposite ends of the drive unit body in the first direction, and the insect-carrying platform is fixed to the second output shaft.

9. The insect trap according to claim 8, characterized in that, The insect-carrying platform includes a top surface facing the imaging chamber. A first convex ridge and a second convex ridge are formed on the top surface. The length directions of the first convex ridge and the second convex ridge are both parallel to the top surface, and the first convex ridge and the second convex ridge intersect perpendicularly. A convex cone is also formed at the intersection of the first convex ridge and the second convex ridge.

10. The insect trap according to claim 8, characterized in that, The insect-catching body also includes a collection tube coaxially arranged with the first tube. The collection tube is detachably connected to the end of the second tube away from the first tube, and a collection cavity communicating with the first tube is formed inside the collection tube.

11. The insect trap according to any one of claims 1 to 10, characterized in that, The insect-catching body also includes a thread tube, which is coaxially arranged with the first tube. One end of the thread tube extends to the end of the first tube away from the second tube, and the other end of the thread tube extends to the insect-collecting funnel. A thread-passing hole is formed on the insect-collecting funnel, and the thread tube communicates with the thread-passing hole.

12. A pest infestation identification system, characterized in that, include: A cloud server that is interconnected and the insect trap as described in any one of claims 1 to 11; The cloud server is used to execute the pest ingress identification method; The pest identification method includes: Receive target infrared data containing a dual-channel discrete time series collected from the current insect trap; Based on a preset time series classification model, the waveform recognition type corresponding to the discrete time series of the dual channels in the target infrared data is obtained. If the waveform recognition type is a pest detection waveform, it is confirmed that a pest has fallen into the insect trap and the insect trap is controlled to collect images of the pest that has fallen into it. The method of obtaining the waveform recognition type corresponding to the discrete time series of the dual channels in the target infrared data based on the preset time series classification model includes: Global features are extracted from the discrete time series of the two channels in the target infrared data to obtain the global features of the target infrared data; The discrete time series of the two channels and the global features are input into a preset time series classification model so that the time series classification model extracts the local features corresponding to each of the discrete time series of the two channels, and fuses each of the local features with the global features to obtain the waveform type identification result data of the target infrared data. The waveform type identification result data includes the probability of different waveform identification types, and the waveform identification types include pest detection waveforms and at least one non-pest detection waveform. The step of performing global feature extraction on the discrete time series of the dual channels in the target infrared data to obtain the global features of the target infrared data includes: The discrete time series of the dual channels in the target infrared data are respectively processed by shifting the waveform downward with a minimum value of 0 to obtain two preprocessed time series corresponding to the target infrared data; Based on a preset effective threshold, effective sampling points are selected in the two preprocessed time series respectively to form the reaction zone corresponding to each of the two preprocessed time series; The global features of the target infrared data are determined based on the reaction regions corresponding to the two preprocessed time series.

Images