Insect killing lamp set insect bottle and method for disposing insect based on image recognition

Abstract

The application provides an image recognition-based insect disposal system and method in a lamp-trap bottle, and belongs to the technical field of pest control and remote monitoring. The system comprises a lamp, a lamp-trap bottle, an image acquisition module, a classification collection module and a control module. The bottom of the lamp-trap bottle is provided with a transparent plate and a light supplement lamp which can be turned over. The image acquisition module acquires images of pests on the transparent plate. The control module adjusts the brightness of the light supplement lamp according to the image quality parameters of the preview images. The classification collection module comprises a driving motor and first and second collection bins. The control module controls the driving motor to rotate forward or reversely according to the identification results, so that the pests which can be reused or cannot be reused fall into the corresponding collection bins. A horizontal rod wrapped with rubber cotton is arranged on the turning path of the transparent plate. The vibration generated by the turning and collision helps the pests to fall off, and the impact noise is absorbed. The application realizes the full-automatic operation of pest trapping, image recognition, classification collection and pest situation early warning.

Description

Technical Field

[0001] This invention belongs to the field of pest control and remote monitoring technology, specifically relating to a pest disposal system and method based on image recognition for insecticidal lamps and insect collection bottles. Background Technology

[0002] Insecticidal lamps are important tools in the field of crop pest monitoring and control. Their basic principle is to attract phototactic pests using light of specific wavelengths, then collect them in a collection device using a high-voltage electric grid or wind suction. In recent years, with the rapid development of computer vision and deep learning technologies, intelligent insect monitoring lamps based on image recognition have gradually become a research hotspot. These devices use an image acquisition module within the collection device to photograph the captured pests, and then use image recognition models to automatically identify the pest species and count their numbers. This not only significantly reduces the workload and error rate of traditional manual classification and counting, but also provides data support for pest forecasting. Existing patents disclose technical solutions for applying image recognition technology to insecticidal lamps, such as using cameras to collect pest images and using artificial intelligence models for automatic identification and early warning; using flashlights and camera components to achieve automatic photography and insect counting; and devices that collect insects by size using multi-layer electric grids and automatically classify and store them after image recognition. These existing technologies have, to a certain extent, achieved the functions of pest image recognition and automatic counting.

[0003] However, existing insecticidal lamps with image recognition capabilities still have the following shortcomings. First, the image quality of pest images is difficult to guarantee effectively. The insect collection bottle of an insecticidal lamp is usually a closed or semi-closed structure with insufficient ambient light. Most existing solutions simply turn on supplementary lighting during image acquisition, lacking precise control over the intensity of the supplementary light. When the thickness of the pest accumulation layer in the collection bottle changes, a fixed supplementary light brightness often leads to overexposure or underexposure of the image, directly affecting the clarity of the pest outline and the accuracy of the recognition algorithm. Second, after completing image recognition, existing insecticidal lamps usually collect all pests into the same container, failing to differentiate them according to pest species. However, in actual agricultural production, not all pests are harmful organisms—some pests (such as armyworms and beet armyworms, etc.) are rich in protein and can be reused as poultry feed or organic fertilizer; while other pests contain toxins or have stingers (such as tussock moths and tussock moths), and cannot be reused, requiring separate treatment. Collecting the two types of pests together not only wastes the insect resources that can be reused, but also increases the difficulty of subsequent sorting and processing.

[0004] To address the aforementioned problems, this invention provides an image recognition-based insecticidal lamp insect collection bottle pest disposal system. Summary of the Invention

[0005] This invention provides a system and method for treating pests in insect collection bottles of insecticidal lamps based on image recognition, aiming to solve the technical problems mentioned in the background art.

[0006] In a first aspect, the present invention provides an image recognition-based insecticidal lamp insect collection bottle pest disposal system, including an insecticidal lamp; The insect collecting bottle has a bottom that is a rotatable transparent plate, an insect temporary storage and identification area is formed above the transparent plate, and a supplementary light is provided below the transparent plate. An image acquisition module is located above the insect collection bottle and is used to acquire images of pests in the pest temporary storage and identification area. The sorting and collection module includes a drive motor disposed at the side end of the transparent plate, and a first collection compartment and a second collection compartment disposed at the bottom of the transparent plate, wherein the output shaft of the drive motor is fixedly connected to the transparent plate; The control module is connected to the image acquisition module, the fill light, and the drive motor, respectively. The control module is configured as follows: Receive pest images acquired by the image acquisition module and identify the pest species and quantity; The brightness of the supplementary light is controlled according to the preview image quality parameters of the pest image; Based on the pest species, determine whether the pest is reusable or not, and control the drive motor to rotate forward or backward: when it is determined to be a reusable pest, control the drive motor to rotate forward by a preset angle, so that the reusable pest falls into the first collection chamber; when it is determined to be a non-reusable pest, control the drive motor to rotate backward by a preset angle, so that the non-reusable pest falls into the second collection chamber.

[0007] Furthermore, the transparent plate includes a circular transparent plate body and two mounting shaft segments arranged along the diameter direction of the transparent plate body. The two mounting shaft segments are respectively arranged on opposite sides of the transparent plate body. One of the mounting shaft segments is provided with a keyway that cooperates with the output shaft of the drive motor so that the transparent plate rotates synchronously with the output shaft of the drive motor. The other mounting shaft segment is rotatably supported in the support hole in the side wall of the insect collecting bottle.

[0008] Furthermore, the transparent plate body is made of transparent glass or transparent acrylic material, and the surface of the transparent plate body facing the pest temporary storage and identification area is provided with an anti-adhesion coating.

[0009] Furthermore, the supplementary light includes a ring-shaped LED array arranged around the rotation axis of the transparent plate, the ring-shaped LED array being located below the transparent plate and uniformly arranged along the circumference of the transparent plate.

[0010] Furthermore, the ring-shaped LED array includes multiple supplementary lighting zones divided along the circumference, each of which is connected to the control module and can independently adjust its brightness; the control module adjusts the brightness of the supplementary lighting zone corresponding to each image region according to the image quality parameters of different image regions in the preview image.

[0011] Furthermore, the preview image quality parameters include average image brightness, brightness uniformity, overexposed area ratio, and underexposed area ratio; the control module compares the preview image quality parameters with the corresponding preset quality range, and increases or decreases the brightness of the fill light according to the comparison result.

[0012] Furthermore, the sorting and collection module also includes a crossbar, which is located below and parallel to the rotation axis of the transparent plate. The two ends of the crossbar are fixedly connected to the side wall of the insect collection bottle, and the outer surface of the crossbar is covered with rubber and plastic cotton. During the process of the drive motor driving the transparent plate to flip in the corresponding direction to the preset angle, the transparent plate collides with the crossbar, so that the transparent plate vibrates.

[0013] Furthermore, the control module stores a pest classification database, which includes pest species information and corresponding reusable or non-reusable attributes for each pest species. The control module queries the pest classification database based on the identified pest species to determine the corresponding utilization attribute, and after controlling the transparent plate to complete a rotation in the corresponding direction, controls the drive motor to drive the transparent plate back to the initial position that forms the pest temporary storage and identification area.

[0014] Furthermore, a low-temperature drying plate is provided at the bottom of the first collection chamber and the second collection chamber, the low-temperature drying plate is connected to a temperature controller, and the temperature controller is electrically connected to the control module.

[0015] Secondly, the present invention provides a method for treating pests in an insect-collecting bottle of an insecticidal lamp based on image recognition, applicable to the insect-collecting bottle system for treating pests in an insecticidal lamp based on image recognition described in any of the above-mentioned embodiments, comprising the following steps: Step 1: The insecticidal lamp attracts and kills pests. The pests fall into the pest temporary storage and identification area through the inlet at the top of the insect collection bottle and accumulate on the transparent plate. Step 2: Turn on the fill light and illuminate the transparent plate with the initial brightness. Control the image acquisition module to acquire the preview image of the pest in the pest temporary storage and identification area. Calculate the image quality parameters of the preview image. The image quality parameters include the average brightness of the image, brightness uniformity, overexposed area ratio, and underexposed area ratio. Step 3: Dynamically adjust the output power of the supplementary light according to the image quality parameters until the image quality parameters reach a preset qualified threshold; after the image quality parameters reach the qualified threshold, control the image acquisition module to capture the target image and perform image recognition on the target image to determine the type and quantity of pests; Step 4: Determine whether the identified pest is a reusable or non-reusable pest: If it is a reusable pest, control the drive motor to rotate forward by a preset angle, causing the transparent plate to flip and the pest to fall into the first collection chamber; if it is a non-reusable pest, control the drive motor to rotate backward by a preset angle, causing the transparent plate to flip in the opposite direction and the pest to fall into the second collection chamber. Step 5: Control the drive motor to rotate back to the initial position, so that the transparent plate returns to a horizontal state, waiting for the next round of pests to fall in and then identifying and dealing with them.

[0016] The present invention has the following beneficial effects: First, this invention effectively solves the problems of insufficient ambient light inside the insect collection bottle and overexposure or underexposure caused by changes in the thickness of the insect accumulation layer, by using a supplementary light lamp set under the transparent plate in conjunction with the image acquisition module, and by using a control module to adjust the brightness of the supplementary light lamp in a closed loop according to the image quality parameters of the preview image. This ensures that the image acquisition module always captures target images under optimal lighting conditions, significantly improving the clarity, contrast and sharpness of the insect images, thereby greatly improving the accuracy and reliability of the subsequent image recognition model in identifying insect species and counting quantities, enabling the entire system to maintain stable and consistent recognition performance under different insect densities.

[0017] Secondly, this invention sets up a first collection chamber and a second collection chamber below the transparent plate, and uses a control module to determine whether the pests are reusable or non-reusable based on image recognition results. Then, it controls the drive motor to rotate forward or backward by a preset angle, causing the transparent plate to flip in different directions. This allows reusable pests (such as lepidopteran insects rich in protein that can be fed to poultry) and non-reusable pests (such as harmful organisms containing toxins or with talons) to be collected into different collection chambers. This achieves automatic classification and collection of pests, which not only avoids the waste of reusable insect resources, allowing them to be used as feed or organic fertilizer, but also reduces the labor costs of subsequent classification and processing, thus having good economic benefits and environmental value.

[0018] Third, this invention wraps rubber-plastic cotton around a crossbar parallel to the rotation axis of the transparent plate. During the rotation of the transparent plate, the edge of the transparent plate collides with the crossbar, generating vibration. This vibration causes the dead insects adhering to the surface of the transparent plate to completely detach. At the same time, the rubber-plastic cotton layer effectively absorbs the impact noise generated by the collision, avoiding the sharp sound generated by traditional rigid collisions that would drive away and interfere with phototactic pests in the surrounding area. This ensures the complete transfer of pests from the insect collection bottle to the collection chamber, preventing insect residue from affecting the accuracy of the next round of image acquisition and classification. It also maintains the continuous attraction effect of the insecticidal lamp on field pests, improving the system's continuous operation capability and overall trapping efficiency.

[0019] Fourth, this invention, through the control module, further acquires the image recognition confidence level after the initial adjustment of the supplementary light brightness, and continues to fine-tune the supplementary light brightness and re-acquire the image for recognition when the confidence level is lower than a preset threshold, until the confidence level reaches the standard. This forms a dual closed-loop control of supplementary light adjustment and recognition verification, which further ensures that the recognition results used for each classification process have sufficient credibility and effectively reduces the risk of misidentification and misclassification caused by fluctuations in the quality of individual images.

[0020] Fifth, this invention achieves rapid and accurate matching of pest species information through a pest classification database pre-stored in the control module. At the same time, after completing one flip, the control module drives the transparent plate to automatically return to the initial horizontal position, preparing for the next round of pest storage and identification. This forms a complete automated operation cycle of "trapping - storage - identification - classification - reset". The entire system is compact, reliable, and highly suitable for unattended pest monitoring and classification in the field. Attached Figure Description

[0021] Figure 1 The structural block diagram of the insecticidal lamp insect collection bottle pest disposal system based on image recognition provided by the present invention; Figure 2 The front view of the structure of the insecticidal lamp insect collection bottle pest disposal system based on image recognition provided by the present invention; Figure 3 The side view of the structure of the insecticidal lamp insect collection bottle pest disposal system based on image recognition provided by the present invention; Figure 4 This is a schematic diagram of the transparent plate structure provided by the present invention; Figure 5 The flowchart illustrates the method for treating pests in an insecticidal lamp collection bottle based on image recognition provided by this invention.

[0022] The components represented by each label in the attached diagram are listed below: 1-Insecticidal lamp, 2-Insect collection bottle, 201-Transparent plate, 2011-Transparent plate body, 2012-Mounting shaft section, 2013-Keyway, 202-Pest temporary storage and identification area, 3-Supplemental light, 4-Image acquisition module, 5-Classification and collection module, 501-Drive motor, 502-First collection chamber, 503-Second collection chamber, 504-Horizontal bar, 6-Control module. Detailed Implementation

[0023] To further understand the content of this invention, a detailed description of the invention is provided in conjunction with the accompanying drawings and embodiments. The specific embodiments described herein are for illustrative purposes only and are not intended to limit the invention. It should also be noted that, for ease of description, only the parts relevant to the invention are shown in the accompanying drawings.

[0024] The application scenarios of this invention are as follows: it can be applied to farmland, orchards, tea gardens, forest areas, gardens, facility agriculture and pest monitoring stations, etc., and can be used in conjunction with insecticidal lamps to automatically identify, classify and collect and dispose of attracted pests, so as to realize the resource utilization of pests and the classification and treatment of harmful pests, and improve the efficiency of pest monitoring and green prevention and control.

[0025] Firstly, see [the following] Figures 1-4This embodiment provides an image recognition-based insecticidal lamp insect collection bottle pest disposal system, including: an insecticidal lamp 1, an insect collection bottle 2, a supplementary light 3, an image acquisition module 4, a classification and collection module 5, and a control module 6. The insect collecting bottle 2 is installed below the insecticidal lamp 1 to temporarily collect pests that fall into the collecting bottle 2 after being attracted and killed by the insecticidal lamp 1. A flip-out transparent plate 201 is provided at the bottom of the collecting bottle 2, and a pest temporary storage and identification area is formed above the transparent plate 201 for identifying the type and number of pests through the image acquisition module 4. A supplementary light 3 is provided below the transparent plate to provide illumination for pest image acquisition. The image acquisition module 4 is located above the collecting bottle 2 to acquire images of pests in the pest temporary storage and identification area and send them to the control module 6. The classification and collection module 5 includes a drive motor 501, a first collection chamber 502, a second collection chamber 503, and a crossbar 504. The drive motor drives the transparent plate to rotate forward or backward, so that the pests on the transparent plate fall into the first collection chamber 502 or the second collection chamber 503 respectively. The crossbar 504 is used to contact the transparent plate 201 during the flipping process to generate vibration, so as to promote the detachment of pests. The control module 6 is connected to the image acquisition module 4, the supplementary light 3, and the drive motor. It includes an image processing unit, a supplementary light brightness control unit, a pest classification and judgment unit, a drive motor control unit, and a data storage unit. It is used to complete functions such as pest image recognition, supplementary light brightness adjustment, pest classification and judgment, transparent plate flipping control, and pest classification database storage, thereby realizing automatic identification, classification, collection, and disposal of pests.

[0026] In this embodiment, the insecticidal lamp 1 includes a lamp holder, an insect-attracting light source, and a high-voltage grid. The insect-attracting light source is installed below the lamp holder, and the high-voltage grid is arranged around the periphery of the insect-attracting light source. An insect-collecting bottle 2 is fixedly connected below the lamp holder. The top of the insect-collecting bottle 2 has an insect inlet, and the side wall of the insect-collecting bottle 2 is detachably connected to the lamp holder by threads or snaps. The insect-attracting light source uses an LED lamp column with a wavelength of 365nm to 380nm. Light waves in this wavelength range have a selective attraction effect on phototactic pests. The high-voltage grid is a cylindrical fence structure, coaxially arranged around the insect-attracting light source. The high-voltage grid is electrically connected to a high-voltage generator and is used to kill pests attracted by the insect-attracting light source. A funnel-shaped insect-receiving tray is arranged below the high-voltage grid. The larger end of the insect-receiving tray faces the high-voltage grid, and the smaller end is connected to the insect inlet at the top of the insect-collecting bottle 2. Pests killed by the high-voltage grid are collected by the insect-receiving tray and fall into the transparent plate 201 inside the insect-collecting bottle 2 through the insect inlet. The lamp holder is equipped with a hanging component at the top, which is used to hang the insecticidal lamp 1 on a field support or lamp post.

[0027] The insect collecting bottle 2 has a cylindrical structure and is made of a dark, opaque material. The top of the bottle has an insect inlet that connects to the small opening of the insect-collecting tray of the insecticidal lamp 1. The insect inlet is funnel-shaped, wider at the top and narrower at the bottom, to guide the killed insects into the bottle. A transparent plate 201 is located at the bottom of the bottle. The transparent plate 201 is made of circular transparent glass or transparent acrylic, with a thickness of 3mm to 8mm. The upper surface of the transparent plate 201 forms a temporary insect storage and identification area 202. The surface of the transparent plate 201 facing the temporary insect storage and identification area 202 is coated with a polytetrafluoroethylene (PTFE) anti-adhesion coating to reduce the probability of insect body fluids adhering to the transparent plate surface. Similarly, the surface of the funnel-shaped insect-collecting tray and the inner surface of the bottle can also be coated with this anti-adhesion coating.

[0028] The transparent plate 201 includes a transparent plate body 2011, a mounting shaft section 2012, and a keyway 2013 disposed at the end of one mounting shaft section. Specifically, the transparent plate 201 includes a circular transparent plate body 2011 and two mounting shaft sections 2012 disposed along the diameter direction of the transparent plate body 2011. The two mounting shaft sections 2012 are respectively disposed on opposite sides of the transparent plate body 2011 and are integrally formed with the transparent plate body 2011. One mounting shaft section 2012 is provided with a keyway 2013 that cooperates with the output shaft of the drive motor 501, so that the transparent plate 201 rotates synchronously with the output shaft of the drive motor 501, thereby realizing the fixed connection between the transparent plate 201 and the output shaft of the drive motor 501 through the keyway 2013. The other mounting shaft section 2012 is rotatably inserted into a support hole opened in the side wall of the insect collecting bottle 2. A self-lubricating bearing is embedded in the support hole to reduce the rotational friction between the mounting shaft section 2012 and the support hole. A gap of 0.5mm to 1.5mm is left between the transparent plate 201 and the side wall of the insect collecting bottle 2 to ensure that the transparent plate 201 does not interfere with the side wall of the insect collecting bottle 2 during the flipping process.

[0029] Image acquisition module 4 is positioned above insect collection bottle 2. Image acquisition module 4 includes a visible light camera with its lens facing vertically downwards. Image acquisition module 4 is mounted on the top of insect collection bottle 2 via a fixed bracket. The lens of image acquisition module 4 is located below and outside the small opening of the insect receiving tray, with the lens's field of view avoiding obstruction by the small opening wall of the insect receiving tray and directly covering the upper surface of transparent plate 201. Insects falling into the large opening of the insect receiving tray slide down the inner wall of the tray, passing through the small opening and the insect entrance before falling into the insect temporary storage and identification area 202. The lens of image acquisition module 4 is located outside the small opening of the insect receiving tray, its optical axis perpendicular to the upper surface of transparent plate 201 and spatially offset from the insect's falling path. Both operate independently and do not obstruct each other. The image acquisition module 4 is positioned below the small opening of the insect-collecting tray to prevent the lens from obstructing the passage through which pests fall into the insect inlet from the small opening of the tray. Simultaneously, the distance between the lens of the image acquisition module 4 and the transparent plate 201 ensures that the field of view of the visible light camera covers the entire upper surface of the transparent plate 201, allowing the image acquisition module 4 to completely capture images of pests within the temporary insect storage and identification area 202 after they fall into the transparent plate 201. The visible light camera is connected to the control module 6 via a data cable, transmitting the captured insect images to the control module 6. The top cap of the insect-collecting bottle 2 is detachably connected to the side wall of the insect-collecting bottle 2 via a snap-fit ​​or threaded connection, facilitating maintenance and cleaning of the visible light camera.

[0030] A supplementary light 3 is positioned below the transparent plate 201. The supplementary light 3 includes a ring-shaped LED array arranged around the rotation axis of the transparent plate 201. The ring-shaped LED array is located below the transparent plate 201 and is uniformly arranged along the circumference of the transparent plate 201. The ring-shaped LED array includes multiple supplementary lighting zones divided along the circumference. Each supplementary lighting zone is connected to the control module 6 and can independently adjust its brightness. The control module 6 receives pest images acquired by the image acquisition module 4 and identifies the pest species and quantity. Based on the image quality parameters of different image regions in the preview image, it adjusts the brightness of the supplementary lighting zones corresponding to each image region. The preview image quality parameters include average image brightness, brightness uniformity, overexposed area ratio, and underexposed area ratio. The control module 6 compares the preview image quality parameters with the corresponding preset quality range and increases or decreases the brightness of the supplementary light 3 based on the comparison result.

[0031] More specifically, after the image acquisition module 4 acquires a preview image of the pests on the transparent plate 201, the control module acquires the preview image and, based on the spatial correspondence between the lens field of view of the image acquisition module 4 and the upper surface of the transparent plate 201, divides the preview image into multiple image regions. Each image region corresponds spatially to a supplementary lighting zone. The image processing unit of the control module 6 is configured to receive the preview image acquired by the image acquisition module 4 and perform image quality parameter calculations on the preview image. These image quality parameters include average image brightness, brightness uniformity, overexposed area ratio, and underexposed area ratio.

[0032] The image average brightness is the average grayscale value of all pixels within the region. Brightness uniformity is the ratio of the standard deviation of the average brightness of each sub-block within the region to the global average brightness. Overexposed area percentage is the percentage of pixels with a grayscale value exceeding 240 out of the total number of pixels in that region. Underexposed area percentage is the percentage of pixels with a grayscale value below 15 out of the total number of pixels in that region. The control module judges various image quality parameters for each image region: when the average brightness of an image region is lower than the preset target brightness range lower limit of 100, the control module increases the PWM duty cycle of the corresponding supplementary lighting zone to increase brightness; when the average brightness is higher than the target brightness range upper limit of 180, the control module decreases the PWM duty cycle of the supplementary lighting zone to reduce brightness; when the overexposed area percentage exceeds 5%, the control module reduces the brightness of the corresponding supplementary lighting zone; when the underexposed area percentage exceeds 5%, the control module increases the brightness of the corresponding supplementary lighting zone; when the brightness uniformity is lower than 0.85, the control module compares the average brightness of each image region and prioritizes increasing the brightness of the supplementary lighting zone corresponding to the low-brightness region to reduce the brightness difference between regions. For example, when a large number of scarab beetles trapped in the field accumulate on the left side of the transparent plate 201, the corresponding area in the preview image shows low brightness and underexposed areas accounting for more than 5%. The control module determines that the brightness of the corresponding supplementary lighting zone is insufficient, and increases the PWM duty cycle of that supplementary lighting zone separately, while maintaining the original brightness of the supplementary lighting zones in other areas. This makes the outline of the pests clearly visible in the image captured after the supplementary lighting in that local area is enhanced. The control module 6 forms a closed-loop control for the brightness adjustment of each zone of the supplementary light 3. That is, after one adjustment, the preview image is re-acquired and the image quality parameters of each image area are recalculated. If there are still areas that deviate from the preset quality range, the brightness of the corresponding supplementary lighting zone is adjusted again until the image quality parameters of all image areas fall within the preset quality range.

[0033] In actual operation, after pests fall onto the transparent plate 201, their body size and aggregation patterns vary depending on the pest species. Smaller pests (such as rice planthoppers and aphids) may concentrate in the central area of ​​the transparent plate 201, while larger pests (such as lepidopteran moths) may be located at the edges. This uneven distribution leads to differences in brightness requirements in different areas of the preview image. By using independent dimming for each area, the control module can adjust the corresponding supplementary lighting zones according to the actual exposure state of each image area, ensuring that each area of ​​the transparent plate 201 receives appropriate supplementary lighting intensity. After adjusting the brightness of each supplementary lighting zone, the control module re-acquires the preview image and recalculates the image quality parameters of each image area. If there are still areas where the parameters do not meet the standards, iterative adjustments are continued until the image quality parameters of all image areas fall within the preset quality range. The local dimming function of the supplementary light 3 and the image recognition function support each other. Local dimming provides high-quality input images with uniform illumination and clear outlines for image recognition, while the pest distribution information fed back by image recognition guides the supplementary light 3 to adjust the supplementary light intensity of each area in a targeted manner. The interaction ensures that the system can stably obtain image data that meets the recognition requirements under different pest densities and pest distribution patterns. It solves the problem of local overexposure or underexposure of images under fixed brightness supplementary lighting mode caused by uneven thickness of pest accumulation layer, thereby improving the accuracy of pest species identification and quantity statistics.

[0034] Furthermore, after adjusting the brightness of the supplementary light 3, the control module 6 controls the image acquisition module 4 to acquire pest images again, and obtains the recognition confidence level based on the re-acquired pest images; when the recognition confidence level is lower than the preset confidence threshold, the brightness of the supplementary light 3 is adjusted again and pest images are re-acquired until the recognition confidence level reaches the preset confidence threshold.

[0035] Specifically, after adjusting the brightness of the supplementary light 3, the control module 6 performs recognition confidence verification and supplementary light iterative optimization. The control module 6 then controls the image acquisition module 4 to re-acquire pest images within the pest temporary recognition area 202 and inputs these images into the pre-trained pest recognition model. This model outputs the pest species recognition result and the corresponding recognition confidence score, expressed as a percentage, to characterize the reliability of the model for the current recognition result. The control module 6 compares the recognition confidence score with a preset confidence threshold, which is set to 85%. When the recognition confidence level is below 85%, the control module 6 determines that the current supplementary lighting conditions have not yet reached the ideal recognition state. Based on the current brightness of the supplementary light 3, it adjusts the output power of the supplementary light 3 in 5% increments. If the recognition confidence level is below the threshold and the previous adjustment direction was to increase brightness, it continues to adjust in the direction of increasing brightness; otherwise, it continues to adjust in the direction of decreasing brightness. After each adjustment, a new image of the pest is acquired and re-input into the recognition model to obtain a new recognition confidence level. This process is repeated until the recognition confidence level reaches above 85%. The control module 6 records the brightness value and corresponding recognition confidence level for each adjustment. When the number of consecutive adjustments reaches the preset maximum number of iterations, the control module 6 selects the brightness value corresponding to the highest recognition confidence level from the recorded data as the final supplementary lighting brightness.

[0036] The supplementary lighting brightness control unit of control module 6 records the brightness value and corresponding recognition confidence level for each adjustment. When the number of consecutive adjustments reaches the preset maximum number of iterations, the brightness value corresponding to the highest recognition confidence level is selected from the recorded data as the final supplementary lighting brightness. The pest classification management unit is configured to store a pest classification database, which is stored in non-volatile memory in the form of data tables. Each record in the data table includes a pest species name field, a pest characteristic code field, and a corresponding utilization attribute label field. The utilization attribute label is divided into reusable and non-reusable categories.

[0037] In actual operation, when the pests piled on the transparent plate 201 are small planthoppers, if the brightness of the supplementary light 3 is too high, the grayscale difference between the transparent wing veins of the planthoppers and the background of the transparent plate 201 will be excessively compressed. The pest recognition model will not fully extract the morphological features of this type of pest, and the output recognition confidence level may only be 70%. After the control module 6 judges that the confidence level is lower than the 85% threshold, it gradually reduces the brightness of the supplementary light 3 by 5% in each step. When the brightness is reduced to 75% of the initial brightness, the contrast between the wing vein texture of the planthoppers and the background is significantly enhanced. After re-acquiring the image, the confidence level output by the recognition model increases to 88%, which meets the threshold requirement. The control module 6 then stops adjusting and uses this set of images as the final recognition basis. The algorithmic features of the aforementioned recognition confidence verification and supplementary lighting iteration adjustment, along with the structural features of the supplementary light lamp 3 and the control module 6, functionally support each other and interact with each other. Together, they constitute a technical means of implementing closed-loop optimization of supplementary lighting parameters based on recognition feedback. This solves the technical problem that even after adjusting the lighting solely based on image quality parameters, there may still be insufficient confidence in the recognition of specific pest species, leading to low reliability of classification decisions. It achieves the technical effect of using the actual output credibility of the recognition model as the final evaluation criterion, ensuring that the supplementary lighting conditions and the recognition algorithm are adapted, and improving the overall recognition accuracy.

[0038] Furthermore, the classification and collection module 5 includes a drive motor 501 disposed at the side end of the transparent plate 201, and a first collection chamber 502 and a second collection chamber 503 disposed at the bottom of the transparent plate 201. When it is determined that the pests are reusable, the drive motor is controlled to rotate forward by a preset angle, so that the reusable pests fall into the first collection chamber 502; when it is determined that the pests are not reusable, the drive motor is controlled to rotate backward by a preset angle, so that the non-reusable pests fall into the second collection chamber 503. It also includes a crossbar 504, which is located below the rotation axis of the transparent plate 201 and is arranged parallel to the rotation axis. The two ends of the crossbar 504 are fixedly connected to the side wall of the insect collection bottle 2, and the outer surface of the crossbar 504 is covered with rubber and plastic cotton. During the process of the drive motor 501 driving the transparent plate 201 to rotate in the corresponding direction to the preset angle, the transparent plate 201 collides with the crossbar 504, so that the transparent plate 201 vibrates.

[0039] Specifically, one implementation of the classification and collection module 5 is as follows: The transparent plate 201 is a circular plate with mounting shaft sections 2012 on its two opposite ends in the diameter direction. One mounting shaft section 2012 has a concave keyway and is keyed to the output shaft of the drive motor 501. The other mounting shaft section 2012 is rotatably supported in the support hole in the side wall of the insect collection bottle 2. Thus, the transparent plate 201 rotates around its own central axis under the drive of the drive motor 501. The drive motor 501 is a stepper motor. The control module 6 sends forward rotation pulse signals or reverse rotation pulse signals to the drive motor 501 according to the image recognition results. The number of pulses corresponds to a preset rotation angle, which is set to 90°. The control module 6 internally stores a pest classification database, which contains pest species information and corresponding utilization attribute identifiers. The utilization attribute identifiers are divided into reusable and non-reusable categories.

[0040] When the control module 6 determines that the pest species is of the reusable type based on the image recognition result, it sends a forward rotation pulse signal to the drive motor 501. The drive motor 501 rotates forward and drives the transparent plate 201 to rotate 90° towards the direction of the first collection chamber 502. The pests on the transparent plate 201 slide down the surface of the transparent plate 201 under the action of gravity and fall into the first collection chamber 502. When the control module 6 determines that the pest species is of the non-reusable type, it sends a reverse rotation pulse signal to the drive motor 501. The drive motor 501 reverses and drives the transparent plate 201 to rotate 90° towards the direction of the second collection chamber 503. The pests on the transparent plate 201 slide down under the action of gravity and fall into the second collection chamber 503. A crossbar 504 is positioned below and parallel to the rotation axis of the transparent plate 201. The horizontal distance between the crossbar 504 and the rotation axis is less than the radius of the transparent plate 201. Both ends of the crossbar 504 are fixedly connected to the side walls of the insect collection bottle 2. The outer surface of the crossbar 504 is covered with a layer of rubber-plastic cotton. During the process of the transparent plate 201 rotating 90° under the drive of the drive motor 501, the edge of the transparent plate 201 periodically strikes the rubber-plastic cotton layer outside the crossbar 504, causing the transparent plate 201 to generate high-frequency, low-amplitude vibrations. This vibration is transmitted to the pests on the surface of the transparent plate 201, disrupting the adhesion between the pest corpses and the surface of the transparent plate 201, causing the pests adhering to the surface of the transparent plate 201 to detach from the transparent plate 201 and fall into the corresponding collection chamber. At the same time, the rubber-plastic cotton layer undergoes elastic deformation when the transparent plate 201 strikes, absorbing the impact kinetic energy to suppress the propagation of impact noise.

[0041] In actual operation, the carcasses of lepidopteran pests killed by the high-voltage power grid adhere to the surface of the transparent plate 201 due to the adhesion of their body fluids. They are difficult to remove by gravity alone, and the remaining insects can obstruct the image acquisition area for subsequent pests, leading to a decrease in recognition accuracy. The vibration generated by the collision between the transparent plate 201 and the crossbar 504 during the flipping process forces the adhered insects off. Simultaneously, the rubber-plastic layer suppresses impact noise, preventing sharp sounds from driving away phototactic pests in the field. After completing one flip, the control module 6 sends a reverse pulse signal to the drive motor 501, which drives the transparent plate 201 back to its initial horizontal position. The transparent plate 201 returns to its horizontal state, forming the pest temporary storage and recognition area 202, awaiting the next round of pests falling in. In this way, the vibration generated by the collision between the transparent plate 201 and the crossbar 504 during the flipping process, combined with the anti-adhesion coating on the surface of the transparent plate body 2011 facing the pest temporary storage and identification area 202, solves the technical problems of waste of reusable insect resources caused by mixed collection and the impact of insect adhesion residue on continuous operation.

[0042] Furthermore, the data storage unit of the control module 6 stores a pest classification database. This database includes pest species information and corresponding reusable or non-reusable attributes for each pest species. Specifically, the control module 6 receives pest images acquired by the image acquisition module 4, identifies the pest species, queries the pest classification database based on the species, determines the corresponding utilization attribute, and classifies the pest as reusable or non-reusable. It then controls the drive motor 501 to rotate forward or backward: when a reusable pest is identified, the drive motor rotates forward by a preset angle, causing the reusable pest to fall into the first collection chamber 502; when a non-reusable pest is identified, the drive motor rotates backward by a preset angle, causing the non-reusable pest to fall into the second collection chamber 503. After the transparent plate 201 completes one rotation in the corresponding direction, the drive motor 501 drives the transparent plate 201 back to its initial position, forming the pest temporary storage and identification area 202.

[0043] Specifically, for common phototactic pests in agricultural production, specimen images and morphological characteristic data of various pests are collected. Each pest species is labeled with a utilization attribute tag, which is divided into two categories: reusable and non-reusable. The reusable category includes lepidopteran insects such as armyworms, beet armyworms, and sugar beet armyworms. These insects are rich in protein and, after drying and crushing, can be used as an additive in poultry feed or as raw material for organic fertilizer. The non-reusable category includes toxic insects such as spotted stingers, tussock moth larvae with piercing stingers, and aphids with piercing-sucking mouthparts. These pests contain harmful substances or are physically damaging and are unsuitable for use as feed or fertilizer. The pest classification database is stored in the non-volatile memory of control module 6 in the form of data tables. Each record in the data table includes a pest species name field, a pest characteristic code field, and a corresponding utilization attribute tag field.

[0044] After performing image recognition, control module 6 outputs the identified pest species name or species code through the image recognition model of the pest classification judgment unit. Control module 6 uses this species name or species code as a query keyword to search and match in the pest classification database. If a matching record is found, control module 6 extracts the utilization attribute label field from the record and determines the corresponding rotation direction control command based on the category of the label: when the utilization attribute label is a reusable category, control module 6 generates a forward rotation control command and sends it to drive motor 501; when the utilization attribute label is a non-reusable category, control module 6 generates a reverse rotation control command and sends it to drive motor 501. If no matching record is found in the pest classification database, control module 6 marks the current identification result as an unknown species and classifies the unknown species by default as a non-reusable category. At the same time, it generates a log record to prompt the user to manually review the unidentified species and supplement the database label. For example, when the image recognition model outputs that the pest species is armyworm, the control module 6 queries the pest classification database using "armyworm" as the keyword. If the matching utilization attribute tag for armyworm is found to be of the reusable category, the control module 6 immediately sends a forward rotation pulse signal to the drive motor 501, causing the drive motor 501 to rotate the transparent plate 201 towards the first collection chamber 502. When the image recognition model outputs that the pest species is tarantula, the control module 6 finds that its utilization attribute tag is of the non-reusable category. The control module 6 sends a reverse rotation pulse signal to the drive motor 501, causing the drive motor 501 to rotate the transparent plate 201 towards the second collection chamber 503. After sending a rotation control command to the drive motor 501 and waiting for a preset time, the control module 6 sends a reverse rotation pulse signal to the drive motor 501. The drive motor 501 then rotates the transparent plate 201 around the rotation axis to a horizontal initial position. At this horizontal initial position, a pest temporary storage and recognition area 202 is formed on the upper surface of the transparent plate 201, used to hold newly fallen pests and await the next round of image acquisition. The storage and query matching algorithm features of the aforementioned pest classification database and the structural features of the flip direction control and reset control of the transparent plate 201 are functionally mutually supportive and interactive. Together, they constitute a technical means to drive the classification execution mechanism to perform corresponding actions based on the database query results. This solves the technical problem of how to convert the pest species information output by image recognition into specific classification execution actions, and can achieve the technical effect of automatic mapping between recognition results and classification actions, and accurate and reliable classification and disposal.

[0045] It should be noted that one implementation of the image recognition model for pest identification is as follows: The image recognition model adopts a classification model based on a deep convolutional neural network. Its network structure includes an input layer, multiple convolutional layers, multiple pooling layers, a fully connected layer, and an output layer connected in sequence. The convolutional layers are used to extract local features of the pest image, the pooling layers are used to reduce the dimensionality of the feature map and enhance the translation invariance of the model, the fully connected layers are used to map the extracted high-dimensional features to the pest category space, and the output layer uses the Softmax activation function to output the probability distribution of the image to be identified belonging to each type of pest. The model training process is as follows: Images of various phototactic pests captured in actual field operation scenarios using insecticidal lamps are collected, covering images under different lighting conditions, different packing densities, and different shooting angles, to construct a training dataset. Each image in the training dataset is labeled with its corresponding pest species label. The training dataset is divided into a training set and a validation set according to a set ratio. Images in the training set are input into a convolutional neural network, and the network parameters are iteratively optimized using the backpropagation algorithm and stochastic gradient descent optimizer, aiming to minimize the cross-entropy loss function between the model's prediction results and the labeled labels, until the model's recognition accuracy on the validation set no longer improves, at which point training stops. After training, the model parameters are permanently stored in the non-volatile memory of the control module 6. During actual recognition, the image processing unit preprocesses the pest images captured by the image acquisition module 4 by size normalization and inputs them into the model. The model calculates the probability values ​​of each type of pest, and the category with the highest probability value is taken as the recognition result. This highest probability value is output as the recognition confidence score. For images whose confidence level is lower than the preset confidence threshold, the control module 6 triggers the iterative dimming process of the fill light 3 to improve the quality of the input image before re-identification.

[0046] Furthermore, the data storage unit of control module 6 also pre-stores a spatiotemporal distribution map and an economic threshold database. The spatiotemporal distribution map is stored in the form of a data table. Each record in the data table includes a pest species field, a geographic region field, a time period field, and a corresponding historical occurrence density field. This spatiotemporal distribution map is generated based on statistical data from pest monitoring over the years and is used to characterize the occurrence patterns of different pest species in specific geographic regions and time periods. The economic threshold database is stored in the form of a data table. Each record includes a pest species field, a crop type field, and a corresponding economic damage threshold field. The economic damage threshold refers to the critical density value at which control measures must be taken to avoid economic losses when the pest population density reaches a certain level. After each pest species identification and quantity count, the pest classification management unit in control module 6 outputs the current identification result to the decision unit of the control module. The decision unit extracts the pest species name, quantity, and current timestamp from the identification result and queries the spatiotemporal distribution map and the economic threshold database using the species name as the key. When querying the spatiotemporal distribution map, the decision-making unit uses the current timestamp and the current geographical area of ​​the field as conditions to retrieve the historical occurrence density of the pest species within the corresponding time period and area as a reference baseline. The current identified quantity is compared with the historical occurrence density to determine whether the current pest situation deviates from the normal fluctuation range. When querying the economic threshold database, the decision-making unit uses the pest species and the current crop type as conditions to retrieve the corresponding economic damage threshold and compares the current cumulative quantity of that pest species in the collection bottle with the retrieved economic damage threshold. If the current cumulative quantity reaches or exceeds the economic damage threshold, the decision-making unit determines that an early warning is needed and executes the early warning triggering and disposal priority ranking. The logic for generating the disposal priority ranking is as follows: The decision unit extracts the hazard level coefficient of the pest species. The hazard level coefficient is pre-calibrated based on the pest's feeding amount, reproduction rate, and spread ability. Simultaneously, the decision unit reads the remaining capacity of the insect collection bottle 2. The remaining capacity is estimated as follows: the control module 6 records the cumulative number of times the transparent plate 201 is flipped and tilted each time, and accumulates the pest count statistics output by the image recognition module 4 in each round of recognition to obtain the total number of pests collected in the insect collection bottle 2. The remaining capacity is estimated by subtracting the product of the total number of collected pests and the average volume of a single pest from the total capacity of the insect collection bottle 2. The decision unit calculates the comprehensive disposal priority using the hazard level coefficient and the remaining capacity as input variables. A higher hazard level coefficient and a smaller remaining capacity correspond to a higher disposal priority. For example, when the cumulative number of rice stem borers reaches the economic hazard threshold and the remaining capacity of the insect collection bottle 2 is less than 20%, the decision unit determines the disposal priority to be the highest level. The decision-making unit generates corresponding disposal suggestions based on the disposal priority. The disposal suggestions include immediately cleaning the collection bin and increasing the working time of the insect-attracting light source of insecticidal lamp 1. The disposal suggestions are then output to the remote terminal via wireless communication for the user to refer to and implement.

[0047] Furthermore, low-temperature drying plates are respectively provided at the bottom of the first collection chamber 502 and the second collection chamber 503. The low-temperature drying plate uses a semiconductor cooling chip as its core component. The cold end of the semiconductor cooling chip is attached to the inner surface of the bottom of the collection chamber, while the hot end is provided with heat sinks and exposed to the outside of the collection chamber to achieve directional heat conduction and dissipation. The low-temperature drying plate is connected to a temperature controller, which is electrically connected to the control module 6. The control module 6 adjusts the working state of the low-temperature drying plate through the temperature controller. A first threshold is preset in the control module 6. This first threshold represents the capacity warning value of the total number of pests collected in the insect collection bottle 2. This capacity warning value is estimated based on the total volume of the insect collection bottle 2 and the average volume of a single pest, for example, set as the number of pests corresponding to 80% of the total volume of the insect collection bottle 2. After each round of image recognition, the decision unit in the control module 6 accumulates the number of pests identified in that round to the total number of collected pests counter and compares the accumulated result with the first threshold. When the accumulated result reaches or exceeds the first threshold, the decision unit determines that the amount of pests accumulated in the collection chamber has reached the critical density requiring the initiation of anti-corrosion treatment, and then sends an activation signal to the temperature controller. In response to this activation signal, the temperature controller applies a working voltage to the semiconductor cooling chip of the low-temperature drying plate. The cold end of the semiconductor cooling chip begins to cool, lowering and maintaining the bottom inner surface temperature of the first collection chamber 502 and the second collection chamber 503 between 35°C and 40°C. Because 35°C to 40°C is below the denaturation temperature of most pest proteins, it will not damage the nutritional value of the protein in the insects as a feed additive; simultaneously, this temperature range is higher than the dew point temperature inside the collection chamber, effectively reducing the relative humidity at the bottom of the collection chamber, inhibiting the growth of microorganisms in the moisture attached to the insect surface and body fluids, and delaying the decomposition process of the pest carcasses. The temperature sensor integrated into the temperature controller collects the surface temperature of the low-temperature drying plate in real time and feeds it back to the temperature controller. The temperature controller uses a PID control algorithm to adjust the working current of the semiconductor cooling chip, stabilizing the surface temperature of the low-temperature drying plate within the target range of 35°C to 40°C. When the control module 6 determines that the transparent plate 201 has completed its tilting and the pests in the collection chamber have been emptied, the decision unit resets the counter for the total number of collected pests to zero and sends a stop signal to the temperature controller. The temperature controller then cuts off the working voltage of the semiconductor cooling chip, and the low-temperature drying plate returns to the ambient temperature.

[0048] In actual operation, under the high temperature and humidity of summer, lepidopteran pests collected in the insect collection bottle 2 will begin to show mold and rotten smell after being left in the natural state for more than 48 hours. The ammonia and hydrogen sulfide gases produced by the rotting will interfere with the insect-attracting effect of the insecticidal lamp 1. At the same time, the rotten insects mixed with newly fallen pests make it difficult for the image recognition module 4 to accurately extract the morphological features of the target pests. By using a low-temperature drying plate to maintain the temperature at the bottom of the collection chamber at 35°C to 40°C and reducing the relative humidity in the collection chamber to below 40%, the moisture on the surface of the pests is effectively evaporated, and the growth environment of microorganisms is inhibited. This solves the problem of pests accumulating in the collection chamber for a long time, causing rotting and deterioration, which affects the accuracy of recognition and the working environment. It can achieve the technical effects of delaying pest decay, ensuring the clarity of recognition images, and extending the equipment maintenance cycle.

[0049] Secondly, such as Figure 5 The present embodiment provides a method for treating pests in an insecticidal lamp collection bottle based on image recognition, applicable to any of the above-mentioned insecticidal lamp collection bottle pest treatment systems based on image recognition, characterized by including the following steps: Step 1: The insecticidal lamp 1 attracts and kills pests. The pests fall into the pest temporary storage and identification area 202 through the top entrance of the insect collection bottle 2 and accumulate on the transparent plate 201. Step 2: Turn on the fill light 3 and illuminate the transparent plate 201 with the initial brightness. Control the image acquisition module 4 to acquire the preview image of the pest in the pest temporary storage and identification area 202. Calculate the image quality parameters of the preview image. The image quality parameters include the average brightness of the image, brightness uniformity, overexposed area ratio, and underexposed area ratio. Step 3: Dynamically adjust the output power of the supplementary light 3 according to the image quality parameters until the image quality parameters reach the preset qualified threshold; after the image quality parameters reach the qualified threshold, control the image acquisition module 4 to capture the target image and perform image recognition on the target image to determine the type and quantity of pests; Step 4: Determine whether the identified pest is a reusable or non-reusable pest: If it is a reusable pest, control the drive motor 501 to rotate forward by a preset angle, causing the transparent plate 201 to flip and the pest to fall into the first collection chamber 502; if it is a non-reusable pest, control the drive motor 501 to rotate backward by a preset angle, causing the transparent plate 201 to flip in the opposite direction and the pest to fall into the second collection chamber 503. Step 5: Control the drive motor 501 to rotate back to the initial position, so that the transparent plate 201 returns to a horizontal state, waiting for the next round of pests to fall in and then identifying and dealing with them.

[0050] More specifically, in step one, the insect-attracting light source of the insecticidal lamp 1 emits ultraviolet light with a wavelength of 365nm to 380nm. Phototactic pests in the field are attracted to this light source and fly towards the insecticidal lamp 1. Upon contact with the high-voltage grid, they are killed by high-voltage electric shock. The dead pests are collected and guided by the insect-collecting tray, and fall through the pest entrance at the top of the insect-collecting bottle 2 onto the transparent plate 201 inside the insect-collecting bottle 2, where they accumulate in the pest temporary storage and identification area 202 formed on the upper surface of the transparent plate 201. For example, when operating in rice fields, the insect-attracting light source can attract and kill lepidopteran pests such as rice leaf rollers and rice stem borers. The dead pests fall onto the transparent plate 201 to await further processing.

[0051] In step two, control module 6 sends an initial PWM duty cycle signal to fill light 3, and fill light 3 illuminates transparent plate 201 at the initial brightness. Control module 6 controls image acquisition module 4 to acquire preview images of pests in pest temporary storage and identification area 202, and the preview image is transmitted to image processing unit of control module 6. After receiving the preview image, image processing unit divides the preview image into multiple image regions according to the spatial correspondence between the lens field of view and the upper surface of transparent plate 201, and calculates image quality parameters for each region. Image quality parameters include average image brightness, brightness uniformity, overexposed area percentage, and underexposed area percentage; wherein, average image brightness is the average gray value of all pixels in the region, brightness uniformity is the ratio of the standard deviation of the average brightness of each sub-block in the region to the global average brightness, overexposed area percentage is the percentage of pixels with a gray value greater than 240 out of the total number of pixels in the region, and underexposed area percentage is the percentage of pixels with a gray value less than 15 out of the total number of pixels in the region. For example, when pests accumulate in the central area of ​​the transparent plate 201 while there are no pests in the edge area, the central area has low brightness due to light absorption caused by the accumulation of pests. The image processing unit calculates that the average brightness of the image in the central area is lower than the lower limit of the preset target range by 100, and the underexposed area accounts for more than 5%.

[0052] In step three, the fill light brightness control unit of control module 6 receives the image quality parameters of each image region calculated by the image processing unit, compares each parameter of each image region with the corresponding preset quality range, and dynamically adjusts the output power of fill light 3 according to the comparison results. When the average brightness of an image region is lower than the lower limit of the target brightness range by 100, the fill light brightness control unit increases the PWM duty cycle of the fill light zone corresponding to that image region; when the average brightness of the image is higher than the upper limit of the target brightness range by 180, the PWM duty cycle of the fill light zone is decreased; when the proportion of overexposed areas exceeds 5%, the brightness of the corresponding fill light zone is reduced; when the proportion of underexposed areas exceeds 5%, the brightness of the corresponding fill light zone is increased; when the brightness uniformity is lower than 0.85, the brightness of the fill light zone corresponding to the low brightness area is increased first. After completing the brightness adjustment of each fill light zone, the fill light brightness control unit triggers the image acquisition module 4 again to acquire preview images and recalculate the image quality parameters of each image region, repeating the above process until the image quality parameters of all image regions fall within the preset quality range. After the image quality parameters reach the acceptable threshold, control module 6 controls image acquisition module 4 to capture the target image. The image processing unit inputs the target image into a pre-trained pest identification model. This model uses a convolutional neural network to extract features and classify the pest image, outputting the pest species identification result and the corresponding identification confidence level. The supplementary light brightness control unit also performs identification confidence level verification: when the identification confidence level is lower than the preset confidence level threshold of 85%, the supplementary light brightness control unit adjusts the output power of the supplementary light 3 in steps of 5%. After each step adjustment, the image is re-acquired and re-identified until the identification confidence level reaches 85% or higher. The species and quantity of pests are determined based on this identification result. For example, when the pests piled on the transparent plate 201 are planthoppers, the grayscale difference between the transparent wing veins of the planthoppers and the background is insufficient when the supplementary light is insufficient, and the recognition confidence is only 70%. After the brightness control unit of the supplementary light is stepped to adjust, the brightness is increased to 125% of the initial brightness, the wing vein texture is clearly distinguishable, and the recognition confidence is increased to 88%, which meets the threshold requirement.

[0053] In step four, the pest classification management unit of control module 6 receives the pest species identification result output by the image processing unit, and performs a search and matching in the pest classification database using the species name as the query keyword to extract the corresponding utilization attribute tag. If the utilization attribute tag is a reusable category, the drive motor control unit of control module 6 sends a forward rotation pulse signal to drive motor 501, causing drive motor 501 to rotate forward and rotate transparent plate 201 towards the direction of first collection chamber 502 by a preset angle; if the utilization attribute tag is a non-reusable category, the drive motor control unit sends a reverse rotation pulse signal to drive motor 501, causing drive motor 501 to rotate in reverse and rotate transparent plate 201 towards the direction of second collection chamber 503 by a preset angle. The preset rotation angle is set to 90°. During the process of the transparent plate 201 being flipped to a preset angle, the edge of the transparent plate 201 periodically strikes the rubber and plastic cotton layer outside the crossbar 504, causing the transparent plate 201 to generate high-frequency, low-amplitude vibrations. These vibrations are transmitted to the pests on the surface of the transparent plate 201, disrupting the adhesion between the pest corpses and the surface of the transparent plate 201. This causes the pests to slide down the surface of the transparent plate 201 under the combined action of gravity and vibration and fall into the corresponding collection chamber. The rubber and plastic cotton layer undergoes elastic deformation upon impact, absorbing the impact kinetic energy to suppress the propagation of impact noise. For example, when the identification result is armyworm, the pest classification management unit finds that the utilization attribute label corresponding to the armyworm is a reusable category. The drive motor 501 rotates forward, causing the transparent plate 201 to flip towards the first collection chamber 502. With the assistance of the vibration generated by the transparent plate 201 hitting the crossbar 504, all the armyworms fall into the first collection chamber 502. When the identification result is spotted stinger, the utilization attribute label is a non-reusable category. The drive motor 501 rotates in reverse, causing the transparent plate 201 to flip towards the second collection chamber 503.

[0054] In step five, after sending a flip control command and waiting for a preset time, the drive motor control unit sends a reverse pulse signal to the drive motor 501. The drive motor 501 drives the transparent plate 201 to rotate around the axis to a horizontal initial position. At this horizontal initial position, a pest temporary storage and identification area 202 is formed on the upper surface of the transparent plate 201. After the transparent plate 201 returns to a horizontal state, the system waits for the next round of pests to fall in and performs identification and treatment, thus forming a complete automated operation cycle of "trapping—temporary storage—identification—classification—reset". For example, after classifying and dumping armyworms, the transparent plate 201 returns to a horizontal state. The next rice stem borer to be killed falls onto the transparent plate 201, and the system begins a new round of image acquisition and quality adjustment.

[0055] Furthermore, after step five, the decision-making unit of control module 6 also performs an early warning judgment based on the economic damage threshold. After each round of image recognition, the decision-making unit uses the name of the currently identified pest species as a keyword to search and match in the pre-stored economic threshold database to obtain the economic damage threshold of that pest species under the current crop type. The economic threshold database pre-stores the economic damage thresholds of different pest species for different crops. This threshold represents the critical density value at which the pest population density reaches the point where control measures are needed to avoid economic losses. The decision-making unit compares the current cumulative number of that pest species in the insect collection bottle 2 with the economic damage threshold. If the current cumulative number reaches or exceeds the economic damage threshold, the decision-making unit triggers an early warning signal and sends the early warning signal to a remote terminal through the wireless communication module. The content of the early warning signal includes the pest species name, the current cumulative number, the economic damage threshold, and the suggested treatment time window. For example, when working in rice paddies, if the cumulative number of rice stem borers identified reaches the economic damage threshold of 20 borers per 100 rice plants, the decision-making unit will generate an early warning signal containing information that the number of rice stem borers has exceeded the threshold, the number exceeding the threshold, and a recommendation to apply the corresponding pesticide within 3 days, and send it to the remote terminal for farmers' reference.

[0056] Furthermore, the control module 6 also performs the activation determination of the low-temperature drying plate in the collection chamber. The control module 6 has a preset first threshold, which is estimated based on the total volume of the insect collection bottle 2 and the average volume of a single pest, for example, set as the number of pests corresponding to 80% of the total volume of the insect collection bottle 2. After each round of image recognition, the decision unit in the control module 6 accumulates the number of pests identified in that round to a counter of the total number of collected pests, and compares the accumulated result with the first threshold. When the accumulated result reaches or exceeds the first threshold, the decision unit determines that the amount of pests accumulated in the collection chamber has reached the critical density requiring the initiation of anti-corrosion treatment, and then sends an activation signal to the temperature controller. The temperature controller responds to the activation signal by controlling the low-temperature drying plate to start operation. The low-temperature drying plate uses a semiconductor cooling chip, the cold end of which is attached to the inner surface of the bottom of the collection chamber. After activation, the cold end of the semiconductor cooling chip cools down, causing the temperature of the inner surface of the bottom of the first collection chamber 502 and the second collection chamber 503 to decrease and be maintained between 35°C and 40°C. This temperature range is higher than the dew point temperature inside the collection chamber, which can reduce the relative humidity at the bottom of the collection chamber to inhibit the reproduction activity of microorganisms, thereby delaying the decomposition process of pest corpses.

[0057] The image recognition-based insecticidal lamp collection bottle pest disposal system and method provided in this embodiment utilizes a control module to adjust the brightness of the supplementary light in a closed loop based on the image quality parameters of the preview image, and performs iterative optimization of the supplementary light based on the recognition confidence level. This ensures that the image acquisition module can obtain clear, high-contrast pest images under different insect densities and distribution patterns, significantly improving the accuracy of the image recognition model in identifying pest species and counting quantities. By setting up a first collection chamber and a second collection chamber, and coordinating the forward and reverse rotation of the drive motor to control the directional flipping of the transparent plate, reusable and non-reusable pests are collected separately, realizing the utilization of insect resources. The system utilizes classified pests to avoid resource waste caused by mixed collection. By placing crossbars covered with rubber-plastic cotton along the flipping path of the transparent plate, the collision vibration during the flipping process causes adhered insects to detach completely, while simultaneously suppressing the impact noise that drives away phototactic pests in the field, ensuring the system's continuous operation and trapping efficiency. Through the storage and query matching of the pest classification database, automatic mapping between identification results and classification actions is achieved, ensuring the accuracy and reliability of classification and disposal. Joint analysis of spatiotemporal distribution maps and economic threshold databases enables pest infestation early warning and priority ranking for disposal, providing a decision-making basis for precision pesticide application and green pest control. The entire system is compact and reliable, forming a complete automated operation cycle of "trapping—temporary storage—identification—classification—reset," and can be widely applied to unattended pest monitoring and classification disposal operations in large-scale agricultural planting areas and organic agricultural product production bases.

[0058] The above description is merely a specific embodiment of this specification. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, modules, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here. It should be understood that the scope of protection of this specification is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this specification, and these modifications or substitutions should all be covered within the scope of protection of this specification.

Claims

1. An image recognition-based insecticidal lamp insect collection bottle pest treatment system, comprising an insecticidal lamp (1), characterized in that, Also includes: The insect collection bottle (2) has a bottom of a flip-up transparent plate (201), a pest temporary storage and identification area (202) is formed above the transparent plate (201), and a supplementary light (3) is provided below the transparent plate (201). The image acquisition module (4) is set above the insect collection bottle (2) and is used to acquire images of pests in the pest temporary storage and identification area (202); The classification and collection module (5) includes a drive motor (501) disposed at the side end of the transparent plate (201), and a first collection compartment (502) and a second collection compartment (503) disposed at the bottom of the transparent plate (201). The output shaft of the drive motor (501) is fixedly connected to the transparent plate (201). The control module (6) is connected to the image acquisition module (4), the fill light (3), and the drive motor (501) respectively. The control module (6) is configured as follows: Receive pest images acquired by the image acquisition module (4) and identify the pest species and quantity; The brightness of the supplementary light (3) is controlled according to the preview image quality parameters of the pest image; Based on the pest species, determine whether the pest is a reusable pest or not, and control the drive motor (501) to rotate forward or backward: when it is determined to be a reusable pest, control the drive motor to rotate forward by a preset angle, so that the reusable pest falls into the first collection chamber (502); when it is determined to be a non-reusable pest, control the drive motor to rotate backward by a preset angle, so that the non-reusable pest falls into the second collection chamber (503).

2. The insecticidal lamp insect collection bottle pest treatment system based on image recognition according to claim 1, characterized in that, The transparent plate (201) includes a circular transparent plate body (2011) and two mounting shaft segments (2012) arranged along the diameter direction of the transparent plate body (2011). The two mounting shaft segments (2012) are respectively arranged on opposite sides of the transparent plate body (2011). One of the mounting shaft segments (2012) is provided with a keyway (2013) that cooperates with the output shaft of the drive motor (501) so that the transparent plate (201) rotates synchronously with the output shaft of the drive motor (501). The other mounting shaft segment (2012) is rotatably supported in the support hole of the side wall of the insect collecting bottle (2).

3. The insecticidal lamp insect collection bottle pest treatment system based on image recognition according to claim 2, characterized in that, The transparent plate body (2011) is made of transparent glass or transparent acrylic material, and the surface of the transparent plate body (2011) facing the pest temporary storage identification area (202) is provided with an anti-adhesion coating.

4. The insecticidal lamp insect collection bottle pest treatment system based on image recognition according to claim 1, characterized in that, The supplementary light (3) includes a ring-shaped LED array arranged around the rotation axis of the transparent plate (201). The ring-shaped LED array is located below the transparent plate (201) and is uniformly arranged along the circumference of the transparent plate (201).

5. The insecticidal lamp insect collection bottle pest treatment system based on image recognition according to claim 4, characterized in that, The ring-shaped LED array includes multiple supplementary lighting zones divided along the circumference. Each supplementary lighting zone is connected to the control module (6) and its brightness can be adjusted independently. The control module (6) adjusts the brightness of the supplementary lighting zones corresponding to each image region according to the image quality parameters of different image regions in the preview image.

6. The insecticidal lamp insect collection bottle pest treatment system based on image recognition according to claim 1, characterized in that, The preview image quality parameters include average image brightness, brightness uniformity, overexposed area ratio, and underexposed area ratio; the control module (6) compares the preview image quality parameters with the corresponding preset quality range, and increases or decreases the brightness of the fill light (3) according to the comparison result.

7. The insecticidal lamp insect collection bottle pest treatment system based on image recognition according to claim 1, characterized in that, The classification and collection module (5) also includes a crossbar (504), which is located below the rotation axis of the transparent plate (201) and parallel to the rotation axis. The two ends of the crossbar (504) are fixedly connected to the side wall of the insect collection bottle (2). The outer surface of the crossbar (504) is covered with rubber and plastic cotton. During the process of the drive motor (501) driving the transparent plate (201) to rotate to the preset angle in the corresponding direction, the transparent plate (201) collides with the crossbar (504) to make the transparent plate (201) vibrate.

8. The insecticidal lamp insect collection bottle pest treatment system based on image recognition according to claim 1, characterized in that, The control module (6) stores a pest classification database, which includes pest species information and reusable or non-reusable attributes corresponding to each pest species. The control module (6) queries the pest classification database according to the identified pest species, determines the corresponding utilization attribute, and after controlling the transparent plate (201) to complete a rotation in the corresponding direction, controls the drive motor (501) to drive the transparent plate (201) to return to the initial position of forming the pest temporary identification area (202).

9. The insecticidal lamp insect collection bottle pest treatment system based on image recognition according to claim 1, characterized in that, The bottom of the first collection chamber (502) and the second collection chamber (503) are respectively provided with low-temperature drying plates, the low-temperature drying plates are connected to a temperature controller, and the temperature controller is electrically connected to the control module (6).

10. A method for treating pests in an insect-collecting bottle of an insecticidal lamp based on image recognition, applied to the insect-collecting bottle system for treating pests in an insecticidal lamp based on image recognition as described in any one of claims 1-9, characterized in that, Includes the following steps: Step 1: The insecticidal lamp (1) attracts and kills pests. The pests fall into the pest temporary storage and identification area (202) through the top entrance of the insect collection bottle (2) and accumulate on the transparent plate (201); Step 2: Turn on the fill light (3) and illuminate the transparent plate (201) with the initial brightness. Control the image acquisition module (4) to acquire the preview image of the pest in the pest temporary storage and identification area (202). Calculate the image quality parameters of the preview image. The image quality parameters include the average brightness of the image, brightness uniformity, overexposed area ratio and underexposed area ratio. Step 3: Dynamically adjust the output power of the supplementary light (3) according to the image quality parameters until the image quality parameters reach the preset qualified threshold; after the image quality parameters reach the qualified threshold, control the image acquisition module (4) to capture the target image and perform image recognition on the target image to determine the type and quantity of pests; Step 4: Determine whether the identified pest is a reusable or non-reusable pest: If it is a reusable pest, control the drive motor (501) to rotate forward by a preset angle, causing the transparent plate (201) to flip and the pest to fall into the first collection chamber (502); if it is a non-reusable pest, control the drive motor (501) to rotate backward by a preset angle, causing the transparent plate (201) to flip in the opposite direction and the pest to fall into the second collection chamber (503). Step 5: Control the drive motor (501) to rotate back to the initial position, so that the transparent plate (201) returns to a horizontal state, and waits for the next round of pests to fall in and is identified and dealt with.

Images