Pest and disease camera and regional pest and disease control system

Abstract

The technical scheme of the invention discloses a pest and disease damage camera which is characterized by comprising a high-precision image acquisition module, a dynamic environment adaptation module, a central processing module and a real-time data transmission module. Another technical scheme of the invention is to provide a regional pest control system which is characterized by comprising the pest camera, a server and an unmanned aerial vehicle with an image acquisition function. According to the invention, 20 million-pixel multispectral imaging (including visible light, near-infrared and thermal imaging wavebands) is adopted, a lightweight AI algorithm (YOLOv7 model) is combined, 14 types of diseases and insect pests (the accuracy rate is greater than or equal to 90%) can be identified in real time, the weak light imaging quality is guaranteed through dynamic LED light supplement (the wavelength is 380-850 nm), grading early warning (ventilation and humidity reduction, targeted spraying and chemical prevention and treatment) is triggered through 5G transmission of data, and the quality of the disease and insect pests is guaranteed. The system is linked with a temperature and humidity sensor and a CO2 sensor, hidden insect pests such as leaf miners are early warned 48 hours in advance, and the prediction error is smaller than or equal to 8%.

Description

Technical Field

[0001] This invention relates to a pest and disease monitoring device and a regional pest and disease control system using the pest and disease monitoring device. Background Technology

[0002] Traditional methods for monitoring agricultural pests and diseases include field surveys and methods using traps. Field surveys are one of the most common methods in agricultural production. By regularly inspecting farmland and observing crop growth and the extent of damage, farmers and agricultural technicians can promptly identify and detect pests and diseases. Traps are widely used tools for pest and disease monitoring. They are usually made of special materials and carry specific attractants or baits. By placing traps in farmland or around crops, they can attract and capture large numbers of pests and diseases, thus revealing the density and types of pests and diseases.

[0003] Traditional methods for monitoring agricultural pests and diseases rely on manual inspections and chemical reagent testing, which suffer from low efficiency and poor timeliness. With the development of the Internet of Things and image recognition technology, intelligent monitoring systems based on multispectral imaging and AI algorithms are gradually being applied. However, early equipment has shortcomings such as insufficient recognition accuracy (<80%) and poor adaptability to low light. Summary of the Invention

[0004] The purpose of this invention is to combine current technology with edge computing (computing power ≥ 4 TOPS) and multi-sensor fusion (thermal imaging / CO2 / humidity) to achieve millimeter-level positioning and early warning of pests and diseases, supporting the needs of precision plant protection in smart agriculture.

[0005] To achieve the above objectives, the technical solution of this invention discloses a pest and disease camera, characterized in that it includes a high-precision image acquisition module, a dynamic environment adaptation module, a central processing module, and a real-time data transmission module, wherein:

[0006] The high-precision image acquisition module, connected to the central processing module, is used to acquire visible light and multispectral images of the target plant and send the acquired images to the central processing module.

[0007] The dynamic environment adaptation module, connected to the central processing module, is used for exposure synchronization when the high-precision image acquisition module acquires the visible light image, so that the imaging noise is ≤3% in low light environment, and is used to obtain temperature and humidity data to compensate for the visible light image.

[0008] The central processing module runs a lightweight YOLOv7 disease detection algorithm and a temperature and humidity compensation algorithm, among which:

[0009] The temperature and humidity compensation algorithm is used to compensate for the temperature and humidity of the visible light images synchronously acquired by the visible light camera based on the temperature and humidity data collected by the temperature and humidity compensation unit, so as to eliminate the impact of fog and high humidity environment on image quality.

[0010] The lightweight YOLOv7 disease detection algorithm takes temperature and humidity compensated visible light images and multispectral images as input, uses multi-frame image fusion technology to eliminate leaf shading interference, and identifies various diseases and pests.

[0011] The real-time data transmission module, connected to the central processing module, is used to upload the identification results output by the lightweight YOLOv7 disease detection algorithm, along with the corresponding visible light images and multispectral images, to the cloud platform via 5G or narrowband IoT.

[0012] Preferably, the high-precision image acquisition module includes a 20-megapixel visible light camera and a multispectral imaging module, wherein the visible light camera has a resolution of ≥3840×2160, and the multispectral imaging module includes visible light, near-infrared and thermal imaging bands.

[0013] Preferably, the high-precision image acquisition modules are arranged at longitudinal intervals of 10-15 meters along the greenhouse cultivation rack, and the visible light camera is installed at a vertical downward angle of 60°.

[0014] Preferably, the dynamic environment adaptation module includes a dimmable LED fill light strip with a wavelength of 380-850nm and a temperature and humidity compensation unit, wherein the light intensity of the dimmable LED fill light strip is adjustable in the range of 500-1500μmol / m 2 / s, under the control of the central processing module, the exposure of the visible light camera in the high-precision image acquisition module is synchronized with the PWM dimming;

[0015] The temperature and humidity compensation unit is used to collect the temperature and humidity data.

[0016] Preferably, the central processing module uses an NPU chip with a computing power of 4 TOPS.

[0017] Another technical solution of the present invention is to provide a regional pest and disease control system, characterized in that it includes the above-mentioned pest and disease camera, server, and drone with image acquisition function, wherein:

[0018] Based on the plants in the greenhouse, the entire greenhouse is divided into different areas, and multiple pest and disease cameras are evenly set up in each area. Each pest and disease camera is an edge computing node.

[0019] The server triggers a three-tiered response mechanism based on data uploaded from edge computing nodes: a primary warning response mechanism, a secondary warning response mechanism, and a senior warning response mechanism.

[0020] When the area affected by pests and diseases is less than 5%, the primary early warning response mechanism is triggered, which will send an alarm to the mobile phone of the operator and link the external ventilation system to reduce humidity.

[0021] When the area affected by pests and diseases is 5%-15%, an intermediate early warning response mechanism is triggered, which is linked with an external targeted spraying robot, or a drone is used to track and drive the insects to sticky traps. Once the insects are captured, an alarm is simultaneously sent to the operators. The operators then work with the drone to drive the insects away and treat them.

[0022] When the area affected by pests and diseases exceeds 15%, an advanced early warning response mechanism is triggered, which automatically generates a chemical control plan and pushes it to the fertilizer applicator for pesticide application.

[0023] Preferably, the server triggers the three-level response mechanism based on the following method:

[0024] The server obtains image data collected by various pest and disease cameras, and first captures the movement trajectory of bumblebees and the presence of workers based on humidity, temperature, carbon supplementation, bumblebee release in the greenhouse;

[0025] The server performs a large-scale scan of the captured leaves, and records and captures the movement trajectories of other insects or moving objects. The server pre-stores data on the migratory habits and movement trajectories of potential pests, as well as leaf comparison data. The server analyzes the types and numbers of insects in each area from the images transmitted by the pest cameras, and compares and predicts their possible paths and speeds.

[0026] The server generates ring-cut strip image data based on the original leaf condition.

[0027] The server combines the obtained circumferential strip image data with the movement trajectories of insects or other objects to judge and predict pests and diseases, and triggers the three-level response mechanism based on the prediction results.

[0028] This invention employs 20-megapixel multispectral imaging (including visible light, near-infrared, and thermal imaging bands) combined with a lightweight AI algorithm (YOLOv7 model) to identify 14 types of pests and diseases in real time (accuracy ≥90%). Dynamic LED supplemental lighting (wavelength 380-850nm) ensures low-light imaging quality. Data is transmitted via 5G to trigger tiered early warnings (ventilation and dehumidification → targeted spraying → chemical control), and is linked with temperature, humidity, and CO2 sensors to provide early warnings of hidden pests such as leaf miners up to 48 hours in advance, with a prediction error ≤8%. Attached Figure Description

[0029] Figure 1 This is a schematic diagram of a pest and disease camera disclosed in an embodiment of the present invention;

[0030] Figure 2 This is a schematic diagram of a regional pest control system using a pest camera disclosed in an embodiment of the present invention. Detailed Implementation

[0031] The present invention will be further illustrated below with reference to specific embodiments. It should be understood that these embodiments are for illustrative purposes only and are not intended to limit the scope of the invention. Furthermore, it should be understood that after reading the teachings of this invention, those skilled in the art can make various alterations or modifications to the invention, and these equivalent forms also fall within the scope defined by the appended claims.

[0032] like Figure 1 As shown, one aspect of this invention discloses a pest and disease camera, including a high-precision image acquisition module, a dynamic environment adaptation module, a central processing module, and a real-time data transmission module.

[0033] A high-precision image acquisition module is used to acquire visible light and multispectral images of the target plant and send the acquired images to a central processing module to support micron-level identification of leaf lesions and insect eggs. In this embodiment of the invention, the high-precision image acquisition module includes a 20-megapixel visible light camera and a multispectral imaging module. The visible light camera has a resolution ≥3840×2160, and the multispectral imaging module includes visible light, near-infrared, and thermal imaging bands. The high-precision image acquisition modules are arranged at 10-15 meter intervals along the longitudinal direction of the greenhouse cultivation rack, with the visible light camera installed at a vertical tilt angle of 60°. The coverage area of ​​a single high-precision image acquisition module can reach 80-120m². 2 .

[0034] The dynamic environment adaptation module is connected to the central processing module and includes a dimmable LED supplemental lighting strip with a wavelength of 380-850nm and a temperature and humidity compensation unit. The light intensity of the dimmable LED supplemental lighting strip is adjustable in the range of 500-1500μmol / m². 2 Under the control of the central processing module, the system synchronizes the exposure of the visible light camera in the high-precision image acquisition module with PWM dimming to ensure that the imaging noise is ≤3% in low-light environments. The temperature and humidity compensation unit is used to collect temperature and humidity data and send it to the central processing module.

[0035] The central processing module uses an NPU chip with a computing power of 4 TOPS, on which run a lightweight YOLOv7 disease detection algorithm and a temperature and humidity compensation algorithm, among which:

[0036] The temperature and humidity compensation algorithm is used to compensate for the temperature and humidity of the visible light images synchronously acquired by the visible light camera based on the temperature and humidity data collected by the temperature and humidity compensation unit, so as to eliminate the impact of fog and high humidity environment on image quality.

[0037] The lightweight YOLOv7 disease detection algorithm uses temperature and humidity compensated visible light images and multispectral images as input. It employs multi-frame image fusion technology to eliminate leaf shading interference, resulting in a positioning error of ≤1.5cm. 2 It can identify 14 common diseases and pests such as powdery mildew and aphids with an accuracy of ≥90% and a single frame processing time of ≤0.3 seconds.

[0038] The real-time data transmission module, connected to the central processing module, is used to upload the identification results output by the lightweight YOLOv7 disease detection algorithm, along with the corresponding visible light and multispectral images, to the cloud platform via 5G or narrowband Internet of Things (NB-IoT), with a latency of <200ms.

[0039] like Figure 1 The pest and disease camera shown can detect hidden pests such as leaf miners 48 hours in advance by fusing visible light and thermal imaging data (temperature anomaly detection accuracy ±0.5℃).17 When combined with an environmental sensor (CO2 concentration monitoring ±50ppm), it can achieve a disease occurrence probability prediction error of ≤8%.

[0040] like Figure 2 As shown, another aspect of the present invention is to provide a regional pest control system using the above-mentioned pest camera, including the above-mentioned pest camera, server, and drone with image acquisition function.

[0041] Based on the plants in the greenhouse, the entire greenhouse is divided into different areas, with multiple pest and disease cameras evenly distributed across each area. The number of cameras at locations such as the greenhouse access doors exceeds the number of plants, allowing for timely intervention at these access points. Each pest and disease camera is an edge computing node. The identification results obtained by the edge computing nodes are uploaded to a server on the cloud platform via a 5G-MEC edge gateway. Based on the identification results, a three-level response mechanism is triggered: a primary early warning response mechanism (pest and disease area < 5%), a secondary early warning response mechanism (pest and disease area 5%-15%), and a senior early warning response mechanism (pest and disease area > 15%).

[0042] Once the primary early warning response mechanism is triggered, the identification results will be pushed to the mobile phone and the external ventilation system will be activated to reduce humidity.

[0043] Once the intermediate early warning response mechanism is triggered, it will coordinate with external targeted spraying robots.

[0044] Once the advanced early warning response mechanism is triggered, a chemical control plan is automatically generated and pushed to the fertilizer applicator for pesticide application.

[0045] In a preferred embodiment of the present invention, the server may also trigger the above-mentioned three-level response mechanism based on the following method:

[0046] Images collected by various pest and disease cameras are transmitted to a server. The server first tracks the movement of bumblebees and the presence of personnel based on factors such as humidity, temperature, carbon supplementation, and bumblebee deployment in the greenhouse. Bumblebees are mainly active in the growing and fruiting areas. After visiting flowers, they leave brownish-red, rust-like spots on the tomato flowers, indicating pollination. Bumblebees exhibit concentrated activity in the greenhouse, frequently visiting open flowers with minimal interference from external light sources, forming regular flower-visiting paths. Bumblebees fly by flapping their wings back and forth at a high frequency (over 200 times per second) to generate lift. Their numbers generally correspond to the number of bumblebees released into the hive.

[0047] Secondly, the server performs a large-scale scan of the captured leaves, and records and captures the movement trajectories of other insects or moving objects. The server pre-stores data on the migratory habits and movement trajectories of potential pests, as well as leaf comparison data. The server analyzes the types and numbers of insects in each area from the images transmitted by the pest cameras, and predicts their possible paths and speeds.

[0048] Furthermore, the original leaf condition is generated into ring-sectioned strip image data on the server (a panoramic shot is taken, stretched into a long strip, and then cut from bottom to top, as tomatoes grow from bottom to top). As the tomato continues to grow, the position of the ring-sectioned image shifts upwards, creating the original image, i.e., the original data. Growth is then recorded as the growth shifts. When the tomato plants periodically form tops, periodic image data is created, i.e., periodic data. The server compares the original data with the periodic data to confirm the development and evolution of pests and diseases affecting the tomato.

[0049] The ring-cut strip image data obtained by the server is static coordinate data by default, while the movement trajectory of insects or other objects is motion data. The server combines the static coordinate data and motion data to judge and predict pests and diseases.

[0050] Based on the forecast results, drones can be used to track and drive insects to sticky traps, or targeted spraying robots can spray pressurized water jets to drive them away. Since no pesticides are used, an alarm will be triggered simultaneously upon insect capture. Workers will then coordinate with the drones to drive the insects away, primarily using sticky traps and water sprays to control the pests. In the event of a severe infestation of pests or viruses, an alarm will be triggered to the management center, allowing for adjustments to fertilizer and pesticide application to suppress the infestation.

[0051] The drones are primarily responsible for inspecting the air corridors and visitor walkways. In this embodiment of the invention, the original upper obstruction of the pest and disease camera has been eliminated, and it has been replaced with a separate pest and disease camera. The lower fixed-point camera provides a 360-degree panoramic view of the plant's condition for inspection, assisting the drone in inspecting the top canopy and crown. The drone can also avoid obstacles while observing the pest and disease situation above.

[0052] In one preferred embodiment of the present invention, the regional pest and disease control system adopts a federated learning framework, which collects 2000+ labeled samples from each greenhouse every week to update the lightweight YOLOv7 disease detection algorithm, thereby shortening the cycle for improving the accuracy of new disease identification to 7 days.

Claims

1. A pest and disease camera, characterized in that, It includes a high-precision image acquisition module, a dynamic environment adaptation module, a central processing module, and a real-time data transmission module, among which: The high-precision image acquisition module, connected to the central processing module, is used to acquire visible light and multispectral images of the target plant and send the acquired images to the central processing module. The dynamic environment adaptation module, connected to the central processing module, is used for exposure synchronization when the high-precision image acquisition module acquires the visible light image, so that the imaging noise is ≤3% in low light environment, and is used to obtain temperature and humidity data to compensate for the visible light image. The central processing module runs a lightweight YOLOv7 disease detection algorithm and a temperature and humidity compensation algorithm, among which: The temperature and humidity compensation algorithm is used to compensate for the temperature and humidity of the visible light images synchronously acquired by the visible light camera based on the temperature and humidity data collected by the temperature and humidity compensation unit, so as to eliminate the impact of fog and high humidity environment on image quality. The lightweight YOLOv7 disease detection algorithm takes temperature and humidity compensated visible light images and multispectral images as input, uses multi-frame image fusion technology to eliminate leaf shading interference, and identifies various diseases and pests. The real-time data transmission module, connected to the central processing module, is used to upload the identification results output by the lightweight YOLOv7 disease detection algorithm, along with the corresponding visible light images and multispectral images, to the cloud platform via 5G or narrowband IoT.

2. A pest and disease camera as described in claim 1, characterized in that, The high-precision image acquisition module includes a 20-megapixel visible light camera and a multispectral imaging module. The visible light camera has a resolution of ≥3840×2160, and the multispectral imaging module includes visible light, near-infrared, and thermal imaging bands.

3. A pest and disease camera as described in claim 1, characterized in that, The high-precision image acquisition modules are arranged at longitudinal intervals of 10-15 meters along the greenhouse cultivation rack, and the visible light camera is installed at a vertical downward angle of 60°.

4. A pest and disease camera as described in claim 2, characterized in that, The dynamic environment adaptation module includes a dimmable LED fill light strip with a wavelength of 380-850nm and a temperature and humidity compensation unit, wherein the light intensity of the dimmable LED fill light strip is adjustable in the range of 500-1500μmol / m². 2 / s, under the control of the central processing module, the exposure of the visible light camera in the high-precision image acquisition module is synchronized with the PWM dimming; The temperature and humidity compensation unit is used to collect the temperature and humidity data.

5. A pest and disease camera as described in claim 1, characterized in that, The central processing module uses an NPU chip with a computing power of 4 TOPS.

6. A regional pest and disease control system, characterized in that, Includes the pest and disease camera, server, and drone with image acquisition function as described in claim 1, wherein: Based on the plants in the greenhouse, the entire greenhouse is divided into different areas, and multiple pest and disease cameras as described in claim 1 are evenly set in each area, with each pest and disease camera being an edge computing node. The server triggers a three-tiered response mechanism based on data uploaded from edge computing nodes: a primary warning response mechanism, a secondary warning response mechanism, and a senior warning response mechanism. When the area affected by pests and diseases is less than 5%, the primary early warning response mechanism is triggered, which will send an alarm to the mobile phone of the operator and link the external ventilation system to reduce humidity. When the area affected by pests and diseases is 5%-15%, an intermediate early warning response mechanism is triggered, which is linked with an external targeted spraying robot, or a drone is used to track and drive the insects to sticky traps. Once the insects are captured, an alarm is simultaneously sent to the operators. The operators then work with the drone to drive the insects away and treat them. When the area affected by pests and diseases exceeds 15%, an advanced early warning response mechanism is triggered, which automatically generates a chemical control plan and pushes it to the fertilizer applicator for pesticide application.

7. A regional pest and disease control system as described in claim 6, characterized in that, The server triggers the three-level response mechanism based on the following method: The server obtains image data collected by various pest and disease cameras, and first captures the movement trajectory of bumblebees and the presence of workers based on humidity, temperature, carbon supplementation, bumblebee release in the greenhouse; The server performs a large-scale scan of the captured leaves, and records and captures the movement trajectories of other insects or moving objects. The server pre-stores data on the migratory habits and movement trajectories of potential pests, as well as leaf comparison data. The server analyzes the types and numbers of insects in each area from the images transmitted by the pest cameras, and compares and predicts their possible paths and speeds. The server generates ring-cut strip image data based on the original leaf condition. The server combines the obtained circumferential strip image data with the movement trajectories of insects or other objects to judge and predict pests and diseases, and triggers the three-level response mechanism based on the prediction results.

Images