A long-term observation device for farmland soil animal diversity facing environmental changes

Abstract

The application discloses an environment change-oriented farmland soil animal diversity long-term observation device and belongs to the technical field of agricultural ecology monitoring. The application comprises the following: an observation point, which is a fixed area divided in a farmland; a biological information sensing module, which is arranged around the observation point; an environmental factor monitoring module, which is arranged in the soil of the observation point; a physical structure monitoring module, which is arranged in the soil of the observation point; a visual collection module, which is arranged in correspondence with the observation point, and which is used for performing preliminary soil animal target detection, classification identification and counting on collected soil profile images; and a control module, which is electrically connected with the biological information sensing module, the environmental factor monitoring module, the physical structure monitoring module and the visual collection module. The application synchronously monitors in-situ animal image recognition, soil physicochemical properties and microenvironment factors at a millisecond level, and performs deep correlation analysis, thereby providing data support for analyzing the driving mechanism of environmental changes on soil animals.

Description

Technical Field

[0001] This invention belongs to the field of agricultural ecological monitoring technology, and more specifically, it relates to a long-term observation device for farmland soil animal diversity in response to environmental changes. Background Technology

[0002] Soil animals are an important component of farmland ecosystems, playing a crucial role in material cycling, energy flow, soil structure improvement, and pest and disease control. Their community composition and dynamic changes are sensitive indicators of soil health and environmental change. However, traditional methods for observing soil animals (such as hand-picking and trapping) suffer from drawbacks such as high labor intensity, destructiveness, poor timeliness, difficulty in continuous observation, and inability to synchronously correlate with environmental factors. This severely restricts our dynamic understanding of subsurface ecological processes under environmental changes (such as climate change and alterations in agricultural management practices), and fails to meet the demands of precision agriculture and biodiversity conservation for high spatiotemporal resolution and standardized data.

[0003] In recent years, some soil biological observation devices based on image technology have begun to emerge, but most of them have limited functions, or are only for surface organisms, or cannot achieve in-situ profile observation of specific soil layers, and lack the ability to synchronously monitor and intelligently integrate with multi-dimensional environmental factors. Summary of the Invention

[0004] To solve the above-mentioned technical problems, the present invention adopts the following technical solution: A long-term monitoring device for farmland soil animal diversity in response to environmental change includes: Observation points are fixed areas demarcated within farmland. A bioinformation sensing module is placed around the observation point to indirectly sense soil microbial activity. An environmental factor monitoring module is installed in the soil at the observation point to monitor key soil physicochemical properties. A physical structure monitoring module is installed in the soil at the observation point to indirectly assess changes in soil porosity and aggregate structure. A visual acquisition module is configured corresponding to the observation point. The visual acquisition module is capable of taking pictures at regular intervals or triggered by events. The visual acquisition module is equipped with a lightweight deep learning model to perform preliminary soil animal target detection, classification, recognition, and counting on the acquired soil profile images. The control module is electrically connected to the bioinformation sensing module, the environmental factor monitoring module, the physical structure monitoring module, and the visual acquisition module. The control module is responsible for data storage, management, and in-depth analysis to obtain the composition and dynamics of farmland soil animal communities and their relationship with environmental factors, and to visualize them.

[0005] Furthermore, the bio-information sensing module is a near-infrared spectral sensor or a soil respiration sensor; the environmental factor monitoring module includes at least a soil temperature and humidity sensor, a soil pH sensor, and a soil electrical conductivity sensor; and the physical structure monitoring module is a miniature soil tensiometer or a dielectric sensor.

[0006] Furthermore, the visual acquisition module is a high-definition camera encapsulated in a protective cover, and a multi-band adjustable LED fill light system arranged coaxially or in a ring with the high-definition camera; the high-definition camera has timed or triggered shooting and automatic switching function of the fill light system.

[0007] Furthermore, the multi-band adjustable LED supplementary lighting system includes at least a visible white light band and a near-infrared light band, so that the high-definition camera can switch to use different bands of supplementary lighting in different observation modes.

[0008] Furthermore, the shooting triggering modes of the high-definition camera include: Timed trigger mode, taking pictures at preset time intervals; The environmental event trigger mode triggers shooting based on whether the soil temperature and / or soil moisture data collected by the environmental factor monitoring module reaches a preset threshold. The motion detection trigger mode automatically triggers shooting when a moving target is detected in the field of view by analyzing the differences between consecutive image frames.

[0009] Furthermore, the lightweight deep learning model carried by the visual acquisition module is a target detection model based on the YOLO series or MobileNet-SSD architecture that has undergone pruning and quantization processing, and its model input size has been optimized to a fixed size that matches the resolution of the high-definition camera.

[0010] Furthermore, the lightweight deep learning model employs a data augmentation strategy to enhance the detection of small targets in soil animals during the training phase. Specifically, the acquired raw images are preprocessed and then input into the lightweight deep learning model for inference, generating recognition results containing target bounding boxes, categories, and confidence levels. Subsequently, the recognition results undergo non-maximum suppression post-processing, and the number of each category is counted. Finally, a recognition data package containing statistical results, environmental data snapshots, and labeled thumbnails is sent to the control module, which performs deep analysis to obtain and visualize the relationship between farmland soil animals and environmental factors.

[0011] Furthermore, the lightweight deep learning model receives remote update instructions through a cloud platform and completes online model upgrades.

[0012] Compared with the prior art, the present invention has the following beneficial effects: The long-term observation device for farmland soil animal diversity oriented to environmental change provided by this invention can synchronously monitor in situ animal image recognition and soil physicochemical and microenvironmental factors at the millisecond level, and perform in-depth correlation analysis, providing data support for analyzing the driving mechanism of environmental change on soil animals.

[0013] Meanwhile, long-term dynamic data can be used to assess the impact of different farming practices, fertilization treatments, and climate events on soil biodiversity, providing key data support for the formulation of protective agricultural measures and the development of bioindicator evaluation systems, which is of great significance for protecting agricultural underground biodiversity and enhancing ecosystem service functions. Attached Figure Description

[0014] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0015] Figure 1 This is a schematic diagram of the structure of a long-term monitoring device for farmland soil animal diversity in response to environmental changes, provided by the present invention. Detailed Implementation

[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] Example 1

[0018] refer to Figure 1 A long-term monitoring device for farmland soil animal diversity in response to environmental change includes: Observation points are fixed areas designated within farmland. The bioinformation sensing module is placed around the observation point to indirectly sense the activity of soil microorganisms. The environmental factor monitoring module is installed in the soil at the observation point to monitor key soil physicochemical properties. The physical structure monitoring module is installed in the soil at the observation point to indirectly assess changes in soil porosity and aggregate structure. The visual acquisition module is set up in correspondence with the observation point. The visual acquisition module can take pictures at regular intervals or triggered by events. The visual acquisition module is equipped with a lightweight deep learning model to perform preliminary soil animal target detection, classification and recognition and counting on the acquired soil profile images. The control module is electrically connected to the bioinformation sensing module, environmental factor monitoring module, physical structure monitoring module, and visual acquisition module. The control module is responsible for data storage, management, and in-depth analysis, obtaining the composition and dynamics of farmland soil animal communities and their relationship with environmental factors, and displaying them visually.

[0019] Specifically, the observation points are set up as follows: In a typical farmland plot, three repeating observation points are divided according to the grid method. Each observation point can be set as a fixed area of ​​1m×1m, permanently marked with a stainless steel frame, and the frame is buried 5cm deep to avoid interfering with cultivation. The observation points need to cover different areas of the farmland (such as the middle of the plot, the edge, the irrigation area, etc.) and avoid obvious human interference sources, such as fertilizer strips, farm machinery roads, etc.

[0020] In this embodiment, the bio-information sensing module is a near-infrared spectroscopy sensor or a soil respiration sensor; the environmental factor monitoring module includes at least a soil temperature and humidity sensor, a soil pH sensor, and a soil electrical conductivity sensor; and the physical structure monitoring module is a miniature soil tensiometer or a dielectric sensor.

[0021] Specifically, the near-infrared spectral sensor has a wavelength range of 1300-2500nm, a resolution of 10nm, a spectral acquisition time of ≤1 second, and is installed 10cm around the observation point, 20cm above the ground surface; the soil respiration sensor has a measurement range of 0-5000ppmCO2, a detection principle of infrared absorption, and is installed at the center of the observation point, 5cm below the soil surface.

[0022] Among them, the soil temperature and humidity sensor, soil pH sensor and soil conductivity sensor are installed by drilling holes with special soil drill bit to avoid soil compaction. The spacing between the sensors must be ≥15cm to prevent mutual interference.

[0023] The miniature soil tensiometer has a measurement range of 0-100 kPa, a ceramic head diameter of 5 mm, and is installed at the center of the observation point at a depth of 20 cm. The dielectric sensor operates at a frequency of 100 MHz and measures the dielectric constant ε. Its installation position is the same as that of the miniature soil tensiometer.

[0024] In this embodiment, the visual acquisition module is a high-definition camera encapsulated in a protective cover, and a multi-band adjustable LED fill light system arranged coaxially or in a ring with the high-definition camera; the high-definition camera has timed or triggered shooting and automatic switching function of the fill light system.

[0025] In this embodiment, the multi-band adjustable LED supplementary lighting system includes at least the visible white light band and the near-infrared light band, so that the high-definition camera can switch to use different bands of supplementary lighting in different observation modes.

[0026] The supplemental lighting strategy is as follows: During the day (light intensity > 5000 lux): only near-infrared supplemental lighting is used to avoid visible light interfering with natural behavior; Nighttime (illuminance < 100 lux): Dual-band supplemental lighting, white light ensures image clarity, and near-infrared enhances target features.

[0027] In this embodiment, the shooting triggering mode of the high-definition camera includes: Timed trigger mode, taking pictures at preset time intervals; The environmental event trigger mode triggers shooting based on whether the soil temperature and / or soil moisture data collected by the environmental factor monitoring module reaches a preset threshold. The motion detection trigger mode automatically triggers shooting when a moving target is detected in the field of view by analyzing the differences between consecutive image frames.

[0028] In this embodiment, the lightweight deep learning model carried by the visual acquisition module is a target detection model based on the YOLO series or MobileNet-SSD architecture that has undergone pruning and quantization processing. Its model input size has been optimized to a fixed size that matches the resolution of the high-definition camera.

[0029] Preferably, based on the YOLOv5s architecture, specific optimizations are performed for small target detection of soil animals; the pruning optimization adopts structured pruning, removing redundant convolution channels, reducing the number of model parameters by 40% and improving the inference speed by 35%; the quantization processing is INT8 quantization.

[0030] Specifically, the lightweight deep learning model employs a data augmentation strategy to enhance the detection of small targets in soil animals during the training phase. The acquired raw images are preprocessed and then input into the lightweight deep learning model for inference, generating recognition results that include target bounding boxes, categories, and confidence levels. The recognition results are then post-processed with non-maximum suppression, and the number of each category is counted. Finally, the recognition data package, containing statistical results, environmental data snapshots, and labeled thumbnails, is sent to the control module. The control module performs in-depth analysis to obtain and visualize the relationship between farmland soil animals and environmental factors.

[0031] In this embodiment, the lightweight deep learning model receives remote update instructions through a cloud platform and completes online model upgrades.

[0032] The long-term observation device for farmland soil animal diversity oriented to environmental change provided by this invention collects information through multi-dimensional information perception and fusion (including biological information perception, synchronous monitoring of environmental factors and changes in soil physical structure), and automatically photographs, identifies and counts active soil animals in specific soil profiles by combining deep learning image recognition. This enables automated, continuous and standardized observation of farmland soil animal community composition, dynamics and their relationship with environmental factors, providing continuous dynamic data for underground ecological processes under environmental change, which is of great significance for promoting sustainable agricultural development and protecting underground biodiversity.

[0033] The technical solutions of the present invention have been fully described above. It should be noted that the specific embodiments of the present invention are not limited to the above description. All technical solutions formed by those skilled in the art based on the spirit and essence of the present invention by adopting equivalent transformations or equivalent transformations in terms of structure, method or function fall within the protection scope of the present invention.

Claims

1. A long-term monitoring device for farmland soil animal diversity in response to environmental changes, characterized in that, include: Observation points are fixed areas demarcated within farmland. A bioinformation sensing module is placed around the observation point to indirectly sense soil microbial activity. An environmental factor monitoring module is installed in the soil at the observation point to monitor key soil physicochemical properties. A physical structure monitoring module is installed in the soil at the observation point to indirectly assess changes in soil porosity and aggregate structure. A visual acquisition module is configured corresponding to the observation point. The visual acquisition module is capable of taking pictures at regular intervals or triggered by events. The visual acquisition module is equipped with a lightweight deep learning model to perform preliminary soil animal target detection, classification, recognition, and counting on the acquired soil profile images. The control module is electrically connected to the bioinformation sensing module, the environmental factor monitoring module, the physical structure monitoring module, and the visual acquisition module. The control module is responsible for data storage, management, and in-depth analysis to obtain the composition and dynamics of farmland soil animal communities and their relationship with environmental factors, and to visualize them.

2. The long-term monitoring device for farmland soil animal diversity oriented towards environmental change according to claim 1, characterized in that, The bioinformation sensing module is a near-infrared spectroscopy sensor or a soil respiration sensor; the environmental factor monitoring module includes at least a soil temperature and humidity sensor, a soil pH sensor, and a soil electrical conductivity sensor; the physical structure monitoring module is a miniature soil tensiometer or a dielectric sensor.

3. The long-term monitoring device for farmland soil animal diversity oriented towards environmental change according to claim 1, characterized in that, The visual acquisition module consists of a high-definition camera encapsulated in a protective cover and a multi-band adjustable LED fill light system arranged coaxially or in a ring with the high-definition camera; the high-definition camera has timed or triggered shooting and automatic switching functions of the fill light system.

4. The long-term monitoring device for farmland soil animal diversity oriented towards environmental change according to claim 3, characterized in that, The multi-band adjustable LED fill light system includes at least a visible white light band and a near-infrared light band, so that the high-definition camera can switch to use different bands of fill light in different observation modes.

5. The long-term monitoring device for farmland soil animal diversity oriented towards environmental change according to claim 4, characterized in that, The shooting trigger modes of the high-definition camera include: Timed trigger mode, taking pictures at preset time intervals; The environmental event trigger mode triggers shooting based on whether the soil temperature and / or soil moisture data collected by the environmental factor monitoring module reaches a preset threshold. The motion detection trigger mode automatically triggers shooting when a moving target is detected in the field of view by analyzing the differences between consecutive image frames.

6. The long-term monitoring device for farmland soil animal diversity oriented towards environmental change according to claim 4, characterized in that, The lightweight deep learning model carried by the visual acquisition module is a target detection model based on the YOLO series or MobileNet-SSD architecture that has undergone pruning and quantization processing. Its model input size has been optimized to a fixed size that matches the resolution of the high-definition camera.

7. The long-term monitoring device for farmland soil animal diversity oriented towards environmental change according to claim 6, characterized in that, The lightweight deep learning model employs a data augmentation strategy to enhance the detection of small targets in soil animals during the training phase. The acquired raw images are preprocessed and then input into the lightweight deep learning model for inference, generating recognition results containing target bounding boxes, categories, and confidence levels. The recognition results are then post-processed with non-maximum suppression, and the number of each category is counted. Finally, a recognition data package containing statistical results, environmental data snapshots, and labeled thumbnails is sent to the control module, which performs deep analysis to obtain and visualize the relationship between farmland soil animals and environmental factors.

8. The long-term monitoring device for farmland soil animal diversity oriented towards environmental change according to claim 7, characterized in that, The lightweight deep learning model receives remote update instructions through a cloud platform and completes online model upgrades.

Images