Soil crust separation and analysis system

Abstract

The invention relates to the technical field of soil crust separation and analysis, in particular to a soil crust separation and analysis system. The soil crust separation equipment acquires sample soil and acquires a side image corresponding to the sample soil; measuring the soil crust thickness corresponding to the sample soil; separating the sample soil to obtain a sample soil crust; acquiring surface feature data corresponding to the sample soil crust; acquiring crust spectral data corresponding to the sample soil crust; and performing data processing on the soil crust thickness, the side image, the surface feature data and the crust spectral data, determining soil state data corresponding to the sample soil, and outputting corresponding processing measures. The efficiency is improved, and the soil crust thickness can be accurately obtained. In addition, the system can ensure the accuracy of soil state data. Therefore, the accuracy of treatment measures on the sample soil can be ensured.

Description

Technical Field

[0001] The invention relates to the technical field of soil crust separation and analysis, and in particular to a soil crust separation and analysis system. Background Art

[0002] Soil crust refers to a hard thin layer formed on the soil surface due to water evaporation or physical action. Its formation is closely related to the moisture, particle composition, salt and organic matter content of the soil surface. Soil crusts are mainly divided into three categories: physical crusts, chemical crusts and biological crusts. Soil crusts can significantly affect the permeability, air permeability, water retention capacity and plant growth of the soil. Its formation will hinder water penetration and root development, restrict air circulation, and aggravate soil erosion. Therefore, accurate separation and stripping of soil samples with crusts is of great significance to agricultural soil science research. Since the thickness of soil crusts is usually in the submillimeter to millimeter range and is dynamically regulated by environmental factors (such as rainfall intensity and soil texture), its slight changes can significantly affect the water conductivity, erosion resistance and root penetration ability of the soil. At present, in the existing technology, manual separation or the use of traditional separation methods is often inefficient and cannot accurately obtain the thickness of the crust. Traditional measurement methods such as the mechanical contact method of vernier calipers are easy to destroy the brittle crust structure and cannot meet the needs of modern scientific research and agricultural applications. Therefore, how to separate and analyze soil crust has become a technical problem that needs to be solved urgently. Summary of the invention

[0003] In view of this, the present invention provides a soil crust separation and analysis system to solve the above technical problems.

[0004] In a first aspect, the present invention provides a soil crust separation and analysis system, comprising: a soil crust separation device, a crust spectrum acquisition device and an electronic device, wherein: The soil crust separation device is used to obtain sample soil, and during the lifting process of the sample soil, obtain a side image corresponding to the sample soil; determine the soil crust thickness corresponding to the sample soil according to the side image, and separate the sample soil to obtain the sample soil crust; and obtain surface feature data corresponding to the sample soil crust; A crust spectrum acquisition device is used to obtain crust spectrum data corresponding to the sample soil crust; The electronic device is used to process the soil crust thickness, side image, surface characteristic data and crust spectrum data to determine the soil state data corresponding to the sample soil; and output corresponding treatment measures according to the soil state data.

[0005] The soil crust separation and analysis system provided in the embodiment of the present application, the soil crust separation device, is used to obtain sample soil, and during the lifting process of the sample soil, the side image corresponding to the sample soil is obtained, so that the thickness of the soil crust corresponding to the sample soil can be determined according to the side image, and the accuracy of the measured thickness of the soil crust corresponding to the sample soil is ensured. Then, the sample soil is separated according to the thickness of the soil crust to obtain the sample soil crust corresponding to the sample soil, thereby realizing the accurate separation and stripping of the soil sample with crust. No manual separation or traditional separation method is required, which improves efficiency and can accurately obtain the thickness of the soil crust. It can meet the needs of modern scientific research and agricultural applications. In addition, the soil crust separation device obtains the surface feature data corresponding to the sample soil crust, so that the sample soil crust can be studied according to the surface feature data, thereby contributing to agricultural soil science research. The crust spectrum acquisition device is used to collect spectral data of the sample soil crust and obtain the crust spectrum data corresponding to the sample soil crust; and the crust spectrum data is transmitted to the electronic device, so that the electronic device can obtain the crust spectrum data corresponding to the sample soil crust. The electronic device is used to process the soil crust thickness, side image, surface feature data and crust spectrum data, determine the soil state data corresponding to the sample soil, ensure the accuracy of the soil state data corresponding to the determined sample soil, and thus contribute to agricultural soil science research. Then, the corresponding treatment measures are output according to the soil state data, so as to ensure the accuracy of the treatment measures made for the sample soil, and thus ensure the normal growth of plants.

[0006] In an optional embodiment, the electronic device is used to perform feature fusion on the soil crust thickness, surface feature data, crust spectrum data and side image to generate a target high-dimensional feature vector; Perform principal component analysis on the target high-dimensional feature vector and convert the target high-dimensional feature vector into a set of linearly independent target principal components through linear transformation; Feature extraction is performed on the target principal component to determine the soil state data corresponding to the sample soil.

[0007] The soil crust separation and analysis system provided in the embodiment of the present application performs feature fusion on soil crust thickness, surface feature data, crust spectral data and side image to generate a target high-dimensional feature vector, thereby ensuring the accuracy of the generated target high-dimensional feature vector. Principal component analysis is performed on the target high-dimensional feature vector, and the target high-dimensional feature vector is converted into a set of linearly independent target principal components through linear transformation, thereby ensuring the accuracy of the converted target principal components. Feature extraction is performed on the target principal components to determine the soil state data corresponding to the sample soil, thereby ensuring the accuracy of the determined soil state data corresponding to the sample soil.

[0008] In an optional embodiment, the electronic device is used to extract features from sample data including soil crust thickness, surface feature data, crust spectrum data, and side images, and convert the extracted features into corresponding feature vectors; Concatenate each feature vector to generate an initial high-dimensional feature vector; Based on a preset encoder, the initial high-dimensional feature vector is compressed to generate a compressed low-dimensional feature vector; Based on a preset decoder, the compressed low-dimensional feature vector is decoded to generate a target high-dimensional feature vector.

[0009] The soil crust separation and analysis system provided in the embodiment of the present application performs feature extraction on sample data including soil crust thickness, surface feature data, crust spectral data and side images, and converts the extracted features into corresponding feature vectors, thereby ensuring the accuracy of the feature vectors corresponding to the converted sample data including soil crust thickness, surface feature data, crust spectral data and side images. Then, each feature vector is spliced ​​to generate an initial high-dimensional feature vector, thereby ensuring the accuracy of the generated initial high-dimensional feature vector. Then, based on a preset encoder, the initial high-dimensional feature vector is compressed to generate a compressed low-dimensional feature vector, thereby ensuring that the generated compressed low-dimensional feature vector has learned the features of the initial high-dimensional feature vector and reduced the dimension. Based on a preset decoder, the compressed low-dimensional feature vector is decoded to generate a target high-dimensional feature vector, thereby ensuring the accuracy of the generated target high-dimensional feature vector.

[0010] In an optional embodiment, the electronic device is further used to obtain environmental data of the sample soil within a preset time period in the future; the environmental data includes at least one of temperature, precipitation, light intensity, and wind speed; The environmental data, soil crust thickness, surface feature data, crust spectral data and side images are input into the preset soil crust prediction model, and the future soil crust data within the preset future time period corresponding to the sample soil is output.

[0011] The soil crust separation and analysis system and electronic device provided in the embodiments of the present application are also used to obtain environmental data of sample soil within a preset time in the future; the environmental data includes at least one of temperature, precipitation, light intensity, and wind speed; the environmental data and soil crust thickness, surface feature data, crust spectral data, and side image are input into a preset soil crust prediction model, and future soil crust data within a preset time in the future corresponding to the sample soil is output, thereby ensuring the accuracy of the future soil crust data within the preset time in the future corresponding to the output sample soil, so that the impact of soil crust on the soil ecosystem can be understood in advance based on the future soil crust data, and reasonable soil management strategies can be formulated.

[0012] In an optional embodiment, the preset soil crust prediction model includes an input gate, a memory unit, a forget gate, an output gate, and a generative adversarial network, an electronic device, and is used to fuse environmental data with soil crust thickness, surface feature data, crust spectrum data, and side images to generate initial fusion features; Input the initial fusion features into the preset soil crust prediction model according to the time series; The input gate extracts the target fusion feature from the current fusion feature in the initial fusion feature and adds it to the memory unit; The memory unit fuses the target fusion feature with the historical hidden state of the previous moment to generate an updated fusion feature; The forget gate determines whether to retain the fused features or discard the fused features from the updated fused features; The output gate outputs the current hidden state based on the retained fusion features; Calculate the correlation between the hidden state corresponding to each moment and the future soil crust data, and determine the attention weight information corresponding to the hidden state corresponding to each moment; According to the attention weight information, the hidden state corresponding to each moment is weighted summed to obtain the target hidden state; The target hidden state is input into the generative adversarial network and the future soil crust data is output.

[0013] The soil crust separation and analysis system and electronic device provided in the embodiment of the present application are used to fuse environmental data with soil crust thickness, surface feature data, crust spectral data and side images to generate initial fusion features, thereby ensuring the accuracy of the generated target fusion features. Then, the initial fusion features are input into the preset soil crust prediction model in time series; the input gate extracts the target fusion features from the current fusion features in the initial fusion features and adds them to the memory unit in the preset soil crust prediction model, thereby ensuring the accuracy of the target fusion features added to the memory unit. The memory unit fuses the target fusion features with the historical hidden state of the previous moment in the memory unit to generate an updated fusion feature, thereby ensuring the accuracy of the generated updated fusion feature. The forget gate in the preset soil crust prediction model determines the retained fusion features and the discarded fusion features from the updated fusion features in the memory unit, thereby ensuring the accuracy of the determined retained fusion features and discarded fusion features. The output gate in the preset soil crust prediction model outputs the current hidden state based on the retained fusion features, thereby ensuring the accuracy of the output current hidden state. Then, the correlation between the hidden state corresponding to each moment and the future soil crust data is calculated; according to each correlation, the attention weight information corresponding to the hidden state corresponding to each moment is determined, ensuring the accuracy of the attention weight information corresponding to the hidden state corresponding to each moment. Then, according to each attention weight information, the hidden state corresponding to each moment is weighted and summed to obtain the target hidden state, ensuring the accuracy of the obtained target hidden state. The target hidden state is input into the generative adversarial network in the preset soil crust prediction model, and the future soil crust data is output, ensuring the accuracy of the output future soil crust data, so that the impact of soil crust on the soil ecosystem can be understood in advance based on the future soil crust data, and reasonable soil management strategies can be formulated.

[0014] In an optional embodiment, the soil crust separation device includes a ring knife assembly, a laser ranging device, and a separation assembly; the ring knife assembly is installed below the soil crust separation and analysis system, the laser ranging device is installed above the ring knife assembly, and the separation assembly is slidably installed on a connecting rod between the ring knife assembly and the laser ranging device; wherein: A ring knife assembly is used to collect sample soil with crust; The laser ranging device is used to control the sample soil to be lifted upward in the vertical direction, and to obtain a side image during the lifting process of the sample soil; determine the thickness of the soil crust according to the side image; and obtain surface feature data corresponding to the sample soil crust; An electronic device, used for adjusting the height of the separation component according to the thickness of the soil crust, and controlling the separation component to separate the sample soil to obtain the sample soil crust; The separation component is used to separate the sample soil under the control of the electronic equipment to obtain the sample soil crust.

[0015] The soil crust separation and analysis system provided by the embodiment of the present application, the soil crust separation device includes a ring knife assembly, a laser ranging device, and a separation assembly, wherein the ring knife assembly is installed below the soil crust separation and analysis system, the laser ranging device is installed above the ring knife assembly, and the separation assembly is slidably installed on the connecting rod between the ring knife assembly and the laser ranging device; the laser ranging device and the separation assembly are respectively connected to the electronic device, wherein: the ring knife assembly is used to collect sample soil with crust. The laser ranging device is used to control the sample soil to be lifted upward in the vertical direction, and during the lifting process of the sample soil, the side image corresponding to the sample soil is obtained; the thickness of the soil crust corresponding to the sample soil is determined according to the side image; the accuracy of the thickness of the soil crust corresponding to the measured sample soil is ensured. And the surface feature data corresponding to the sample soil crust is obtained; so that the sample soil crust can be studied according to the surface feature data, thereby contributing to agricultural soil science research. The electronic device is used to adjust the height of the separation assembly according to the thickness of the soil crust, control the separation assembly to separate the sample soil, and obtain the sample soil crust corresponding to the sample soil; the separation assembly is used to separate the sample soil under the control of the electronic device to obtain the sample soil crust. This enables accurate separation and stripping of soil samples with crusts. There is no need for manual separation or the use of traditional separation methods, which improves efficiency and can accurately obtain the thickness of soil crusts. It can meet the needs of modern scientific research and agricultural applications.

[0016] In an optional embodiment, the laser distance measuring device includes: a camera assembly, a lifting platform, a laser distance measuring assembly and a digital display assembly; the laser distance measuring assembly is installed above the ring knife assembly, the lifting platform is installed below the ring knife assembly, and the digital display assembly is installed on the outside of the lifting track corresponding to the lifting platform; the camera assembly is installed on the outside of the lifting track corresponding to the lifting platform, and the lifting platform, the camera assembly and the digital display assembly are all connected to the electronic equipment; wherein: The lifting platform is used to drive the sample soil to be lifted upward from the ring knife assembly under the control of the electronic equipment after the ring knife assembly has collected the sample soil; Electronic equipment, used to control the camera assembly to collect side images in real time when the sample soil is lifted upward; The electronic device is also used to identify each side image and determine the contact interface between the sample soil crust and the lower soil in the sample soil; according to the identified contact interface, the lifting platform is controlled to drive the sample soil to be lifted upward in the vertical direction until the contact interface is flush with the upper surface of the ring knife assembly and stops; A laser distance measuring component is used to measure the lifting distance of the lifting platform and determine the lifting distance as the soil crust thickness corresponding to the sample soil; The laser ranging component is also used to collect surface feature data corresponding to the sample soil.

[0017] The soil crust separation and analysis system provided in the embodiment of the present application includes a lifting platform, which is used to drive the sample soil to be lifted upward from the ring knife assembly under the control of the electronic device after the ring knife assembly has collected the sample soil, so as to facilitate the separation of the sample soil and obtain the sample soil crust corresponding to the sample soil. The electronic device is used to control the camera assembly to collect the side image in real time when the sample soil is lifted upward, so as to ensure the accuracy of the acquired side image of the soil crust. The electronic device is also used to identify each side image, determine the contact interface between the sample soil crust and the lower soil in the sample soil, and ensure the accuracy of the contact interface between the sample soil crust and the lower soil in the determined sample soil. According to the identified contact interface, the lifting platform is controlled to drive the sample soil to be lifted upward in the vertical direction until the contact interface is flush with the upper surface of the ring knife assembly and stops, so as to facilitate the measurement of the soil crust thickness and facilitate the separation of the sample soil to obtain the sample soil crust corresponding to the sample soil. The laser ranging component is used to measure the lifting distance of the lifting platform, and determine the lifting distance as the soil crust thickness corresponding to the sample soil, so as to ensure the accuracy of the soil crust thickness corresponding to the determined sample soil. The laser ranging component is also used to collect surface feature data corresponding to the sample soil, so that the sample soil crust can be studied based on the surface feature data, thus contributing to agricultural soil science research.

[0018] In an optional embodiment, the laser ranging assembly includes: a laser sensor, a circular guide rail, a micro-seismic buffer assembly, a crossbar, a first motor and a steel frame; wherein: The diameter of the annular guide rail is the same as the size of the ring knife in the ring knife assembly and is used to carry the laser sensor; A micro-vibration buffer assembly is installed on the annular guide rail and is composed of a flexible silicone pad and a micro spring, and is used to reduce the interference of the vibration generated when the first motor drives the horizontal bar to slide on the laser sensor; The first motor is used to drive the horizontal bar to slide, so that the horizontal bar drives the laser sensor to slide along the steel frame; The laser sensor is used to emit a first infrared light and receive a first reflected light corresponding to the first infrared light before the lifting platform drives the sample soil to lift upward; and emit a second infrared light and receive a second reflected light corresponding to the second infrared light after the lifting platform stops; An electronic device for calculating the soil crust thickness based on the emission time of the first infrared light and the reception time of the first reflected light and the emission time of the second infrared light and the reception time of the second reflected light; The laser sensor is further used to emit visible light to the sample soil crust and generate a reflection image corresponding to the sample soil crust based on a visible reflection signal corresponding to the received visible light; The electronic device is used to identify the reflected image and determine the surface characteristic data corresponding to the soil crust of the sample, wherein the surface characteristic data includes at least one of roughness, particle distribution, and crack characteristics.

[0019] The soil crust separation and analysis system provided by the embodiment of the present application has a circular guide rail with the same diameter as the circular knife in the circular knife assembly, which is used to carry a laser sensor. The micro-seismic buffer assembly is installed on the circular guide rail and is composed of a flexible silicone pad and a micro spring, which is used to reduce the interference of the vibration generated when the first motor drives the horizontal bar to slide on the laser sensor. The first motor is used to drive the horizontal bar to slide, so that the horizontal bar can drive the laser sensor to slide along the steel frame. The laser sensor is used to emit a first infrared light and receive a first reflected light corresponding to the first infrared light before the lifting platform drives the sample soil to lift upward; after the lifting platform stops, it emits a second infrared light and receives a second reflected light corresponding to the second infrared light. The electronic device is used to calculate the soil crust thickness based on the emission time of the first infrared light and the reception time of the first reflected light and the emission time of the second infrared light and the reception time of the second reflected light, thereby ensuring the accuracy of the obtained soil crust thickness. The laser sensor is also used to emit visible light to the sample soil crust, and generate a reflection image corresponding to the sample soil crust based on the visible reflection signal corresponding to the received visible light, thereby ensuring the accuracy of the generated reflection image corresponding to the sample soil crust. The electronic device is used to identify the reflected image and determine the surface characteristic data corresponding to the soil crust of the sample, thereby ensuring the accuracy of the determined surface characteristic data corresponding to the soil crust of the sample.

[0020] In an optional embodiment, the electronic device is used to identify the reflection image based on a preset generative adversarial network to determine the roughness corresponding to the soil crust of the sample; and / or, to identify the reflection image based on a time series crack dynamic monitoring and analysis method to determine the crack characteristics corresponding to the soil crust of the sample; and / or, to identify the reflection image based on a preset target detection algorithm to determine the particle distribution corresponding to the soil crust of the sample.

[0021] The soil crust separation and analysis system provided in the embodiment of the present application recognizes the reflection image based on the preset generative adversarial network, determines the roughness corresponding to the sample soil crust, and ensures the accuracy of the roughness corresponding to the determined sample soil crust. The dynamic crack monitoring and analysis method based on the time series recognizes the reflection image, determines the crack characteristics corresponding to the sample soil crust, and ensures the accuracy of the crack characteristics corresponding to the determined sample soil crust. The reflection image is recognized based on the preset target detection algorithm, and the particle distribution corresponding to the sample soil crust is determined, which ensures the accuracy of the particle distribution corresponding to the determined sample soil crust.

[0022] In an optional embodiment, the separation component includes: a sliding component, a separation slide rail, a separation knife, a second motor, a second pressure sensor, and a processor; the sliding component is installed on a connecting rod between the ring knife component and the laser ranging device, the sliding component drives the separation slide rail to slide on the connecting rod, and the separation knife is installed on the separation slide rail. After the sliding component drives the separation slide rail to be fixed to a specific position according to the thickness of the soil crust, the second motor drives the separation knife to slide on the separation slide rail to separate the sample soil and obtain a sample soil crust corresponding to the sample soil; wherein: The second pressure sensor is used to collect resistance data during the cutting process when the separation knife separates the sample soil to obtain a sample soil crust corresponding to the sample soil, and transmit the resistance data to the processor and the electronic device; The processor is used to adjust the blade angle and cutting force of the separation knife according to the resistance data.

[0023] The soil crust separation and analysis system provided by the embodiment of the present application, the separation component includes: a sliding component, a separation slide rail, a separation knife, a second motor, a second pressure sensor, and a processor, wherein the sliding component is installed on the connecting rod between the ring knife component and the laser ranging device, the sliding component drives the separation slide rail to slide on the connecting rod, and the separation knife is installed on the separation slide rail. After the sliding component drives the separation slide rail to be fixed to a specific position according to the thickness of the soil crust, the second motor drives the separation knife to slide on the separation slide rail to separate the sample soil and obtain the sample soil crust corresponding to the sample soil; wherein: the second pressure sensor is used to collect resistance data during the cutting process when the separation knife separates the sample soil and obtains the sample soil crust corresponding to the sample soil, and transmits the resistance data to the processor and the electronic device. The processor is used to adjust the blade angle and cutting force of the separation knife according to the resistance data, so as to ensure the accuracy of adjusting the blade angle and cutting force of the separation knife. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0025] Figure 1 is a schematic structural diagram of a first soil crust separation and analysis system according to an embodiment of the present invention; Figure 2 is a schematic structural diagram of a second soil crust separation and analysis system according to an embodiment of the present invention; Figure 3 is a schematic structural diagram of a third soil crust separation and analysis system according to an embodiment of the present invention; Figure 4 is a schematic structural diagram of a fourth soil crust separation and analysis system according to an embodiment of the present invention; Figure 5 is a schematic structural diagram of a fifth soil crust separation and analysis system according to an embodiment of the present invention; in: The soil crust separation equipment is marked as 1; The ring knife assembly is marked as 11; The laser distance measuring device is marked as 12; The camera assembly is marked as 121; The lifting platform is marked as 122; The laser distance measuring component is marked as 123; The laser sensor is marked as 1231; The circular guide rail is marked 1232; The micro-seismic buffer assembly is marked as 1233; The horizontal bars are marked 1234; The first motor is marked as 1235; The steel frame is marked 1236; The digital display component is marked 124; The separated component is marked as 13; The sliding assembly is marked as 131; The separation rail is marked as 132; The separation knife is marked as 133; The second motor is marked as 134; The second pressure sensor is marked as 135; The processor is marked as 136; The crust spectrum collection device is marked as 2; Electronic devices are marked 3. DETAILED DESCRIPTION

[0026] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0027] The present invention provides a soil crust separation and analysis system, such as Figure 1 As shown, the soil crust separation and analysis system includes: a soil crust separation device 1, a crust spectrum acquisition device 2 and an electronic device 3. The soil crust separation device 1 and the crust spectrum acquisition device 2 are both connected to the electronic device 3 for communication, wherein: The soil crust separation device 1 is used to obtain sample soil, and during the lifting process of the sample soil, obtain a side image corresponding to the sample soil; determine the thickness of the soil crust corresponding to the sample soil according to the side image, and separate the sample soil to obtain the sample soil crust; and obtain surface feature data corresponding to the sample soil crust; The crust spectrum acquisition device 2 is used to obtain crust spectrum data corresponding to the soil crust of the sample; The electronic device 3 is used to process the soil crust thickness, side image, surface characteristic data and crust spectrum data to determine the soil state data corresponding to the sample soil; and output corresponding processing measures according to the soil state data.

[0028] Among them, soil status data is used to characterize the suitability of sample soil for plant growth.

[0029] Specifically, the soil crust separation device 1 is equipped with a special sampling device that can accurately collect sample soil with crust. Then, the soil crust separation device 1 controls the sample soil to be lifted upward in the vertical direction, and obtains the side image corresponding to the sample soil during the lifting process of the sample soil. Then, the side image is transmitted to the electronic device 3, and the electronic device 3 uses an image processing algorithm to measure the thickness of the soil crust. Specifically, the electronic device 3 can first pre-process the side image, such as noise reduction, contrast enhancement, etc., so that the boundary between the sample soil crust and the underlying soil is clearer. Then the electronic device 3 can use an edge detection algorithm, such as the Canny algorithm, to accurately identify the boundary between the sample soil crust and the underlying soil. By calculating the position difference of the dividing line in the image, and combining the scale of the side image and the actual measured sample soil lifting height, the thickness of the soil crust is accurately calculated.

[0030] Then, the soil crust separation device 1 separates the sample soil crust from the underlying soil according to the measured soil crust thickness. The surface characteristic data of the separated sample soil crust, such as surface roughness, particle distribution, etc., are obtained by laser scanning or other surface measurement technology. Specifically, the laser scanning device in the soil crust separation device 1 emits a laser beam, receives the light reflected from the soil crust surface, and constructs a three-dimensional model of the soil crust surface according to the time and angle information of the reflected light, and then analyzes and obtains the surface characteristic data.

[0031] The crust spectrum acquisition device 2 can be a high-resolution spectrometer. The crust spectrum acquisition device 2 can collect spectral data for the separated sample soil crust. The crust spectrum acquisition device 2 emits light of a specific wavelength range to illuminate the soil crust, and then receives the light reflected by the soil crust. The spectral characteristics of the soil crust are obtained by analyzing the intensity and wavelength distribution of the reflected light. Different chemical components will have specific absorption and reflection peaks on the spectrum. By identifying the position and intensity of these peaks, the content and type of various chemical components in the soil crust can be inferred. Then, the collected crust spectrum data is transmitted to the electronic device 3 in real time via wired or wireless transmission. Wired transmission can use interfaces such as USB and Ethernet to ensure the stability and high speed of data transmission; wireless transmission can use technologies such as Wi-Fi and Bluetooth to improve the flexibility of equipment use. After receiving the crust spectrum data, the electronic device 3 integrates it with the soil crust thickness, side image, surface feature data, etc., to prepare for subsequent data processing and analysis.

[0032] The electronic device 3 uses its powerful data processing capabilities to comprehensively process the soil crust thickness, side image, surface feature data, and spectral data. First, the data is cleaned and preprocessed to remove noise and outliers to ensure the accuracy and reliability of the data. Then, data fusion technology is used to fuse different types of data together to form a comprehensive sample soil feature data set. For example, the chemical composition information in the spectral data is combined with the surface feature data to gain a deeper understanding of the nature of the soil crust.

[0033] Then, the electronic device 3 uses a machine learning algorithm or establishes a mathematical model based on the fused data set to determine the soil state data corresponding to the sample soil. Through a large amount of historical data training, a mapping relationship between soil characteristics and suitable plant growth conditions is established. For example, using a support vector machine (SVM) algorithm, the soil is divided into different state data according to characteristics such as soil crust thickness, chemical component content, and surface roughness, and each state data corresponds to a different degree of suitability for plant growth.

[0034] Finally, the electronic device 3 outputs corresponding treatment measures according to the determined soil state data. For sample soils with higher soil grade and suitable for plant growth, protective agricultural measures are recommended, such as reasonable fertilization, water-saving irrigation, etc., to maintain the good state of the soil; for sample soils with lower soil grade and not suitable for plant growth, targeted improvement suggestions are provided, such as adding soil conditioners, deep plowing and loosening the soil, etc., to improve soil quality and increase its suitability for plant growth.

[0035] The soil crust separation and analysis system provided in the embodiment of the present application, the soil crust separation device 1, is used to obtain sample soil, and during the lifting process of the sample soil, the side image corresponding to the sample soil is obtained, so that the soil crust thickness corresponding to the sample soil can be determined according to the side image, and the accuracy of the soil crust thickness corresponding to the measured sample soil is ensured. Then, the sample soil is separated according to the thickness of the soil crust to obtain the sample soil crust corresponding to the sample soil, thereby realizing the accurate separation and stripping of the soil sample with crust. There is no need for manual separation or the use of traditional separation methods, which improves efficiency and can accurately obtain the thickness of the soil crust. It can meet the needs of modern scientific research and agricultural applications. In addition, the soil crust separation device 1 obtains the surface feature data corresponding to the sample soil crust, so that the sample soil crust can be studied according to the surface feature data, thereby contributing to agricultural soil science research. The crust spectrum acquisition device 2 is used to collect spectral data of the sample soil crust and obtain the crust spectrum data corresponding to the sample soil crust; and transmit the crust spectrum data to the electronic device 3, so that the electronic device 3 can obtain the crust spectrum data corresponding to the sample soil crust. The electronic device 3 is used to process the soil crust thickness, side image, surface feature data and crust spectrum data, determine the soil state data corresponding to the sample soil, ensure the accuracy of the determined soil state data corresponding to the sample soil, and thus contribute to agricultural soil science research. Then, the corresponding treatment measures are output according to the soil state data, so as to ensure the accuracy of the treatment measures made for the sample soil, and thus ensure the normal growth of plants.

[0036] In an optional embodiment of the present application, the electronic device 3 is used to perform feature fusion on the soil crust thickness, surface feature data, crust spectrum data and side image to generate a target high-dimensional feature vector; Perform principal component analysis on the target high-dimensional feature vector and convert the target high-dimensional feature vector into a set of linearly independent target principal components through linear transformation; Feature extraction is performed on the target principal component to determine the soil state data corresponding to the sample soil.

[0037] In an optional embodiment, the electronic device 3 is used to extract features from the sample data including soil crust thickness, surface feature data, crust spectrum data and side images, and convert the extracted features into corresponding feature vectors; Concatenate each feature vector to generate an initial high-dimensional feature vector; Based on a preset encoder, the initial high-dimensional feature vector is compressed to generate a compressed low-dimensional feature vector; Based on a preset decoder, the compressed low-dimensional feature vector is decoded to generate a target high-dimensional feature vector.

[0038] Specifically, the electronic device 3 can perform feature extraction on the sample data including soil crust thickness, surface feature data, crust spectrum data and side image, and convert the extracted features into corresponding feature vectors.

[0039] For the thickness of the soil crust, the electronic device 3 can use the thickness of the soil crust as a direct feature and convert it into a one-dimensional feature vector. For example, if the thickness of the soil crust is d, the corresponding feature vector can be expressed as [d]. For surface feature data, the surface feature data may include information such as roughness, particle distribution, and crack characteristics. For roughness, the electronic device 3 can obtain a value through a specific roughness calculation method; particle distribution can obtain a set of values ​​by counting the proportion of particles of different particle sizes; crack characteristics can be quantified by the length, width, number, etc. of the cracks. These quantized values ​​are combined into a feature vector. For example, the roughness is r, and the statistical value of the particle distribution is [p1, p2, ..., p n ], the crack feature quantization value is [c1,c2,c3], then the feature vector corresponding to the surface feature data is [r,p1,p2,...,p n ,c1,c2,c3]. For the crust spectrum data, the crust spectrum data records the reflectance of soil crust at different wavelengths and other information. The reflectance values ​​corresponding to each wavelength are combined into a feature vector. For example, the reflectances at m wavelengths are [s1,s2,...,s m ], then the eigenvector corresponding to the crust spectrum data is [s1,s2,...,s m ]. For the side image, image processing algorithms can be used to extract features, such as edge features, texture features, etc. The image is converted into a fixed-length feature vector through methods such as convolutional neural networks (CNN). Assume that the feature vector obtained after processing is [i1,i2,...,i k ].

[0040] Then, the electronic device 3 concatenates the feature vectors to generate an initial high-dimensional feature vector.

[0041] The preset encoder is usually a neural network model, which compresses the initial high-dimensional feature vector, removes redundant information, extracts more representative features, and generates a compressed low-dimensional feature vector. The preset encoder processes the input initial high-dimensional feature vector through a series of linear transformations and nonlinear activation functions. For example, the preset encoder may contain multiple fully connected layers, each of which performs a weighted summation on the input feature vector and performs a nonlinear transformation through an activation function (such as ReLU), and finally maps the high-dimensional feature vector to a low-dimensional space to obtain a compressed low-dimensional feature vector.

[0042] The preset decoder is also a neural network model, and its function is to restore the compressed low-dimensional feature vector to a high-dimensional feature vector. The structure of the preset decoder is similar to that of the preset encoder, but the processing process is reversed. The preset decoder receives the compressed low-dimensional feature vector and maps it back to the high-dimensional space through a series of linear transformations and nonlinear activation functions to generate the target high-dimensional feature vector. In this process, the preset decoder tries to retain the important information in the original feature vector as much as possible, while removing the noise that may be introduced during the compression process.

[0043] Then, the electronic device 3 first standardizes the target high-dimensional feature vector so that the mean of each feature is 0 and the standard deviation is 1. This can eliminate the impact of the dimensions between different features and ensure that each feature has equal importance in the principal component analysis. Suppose the target high-dimensional feature vector is X=[x1,x2,...,x l ], the standardized eigenvector is X std .

[0044] Next, the electronic device 3 calculates the covariance matrix C of the standardized target high-dimensional feature vector. The covariance matrix C reflects the correlation between the various features. std , the elements C of the covariance matrix C ij is the covariance of the i-th feature and the j-th feature, and the calculation formula is , where x ki is the i-th eigenvalue of the k-th sample, is the mean of the ith feature.

[0045] Perform eigenvalue decomposition on the covariance matrix C and obtain the eigenvalue and the corresponding eigenvector The eigenvalue represents the variance of each principal component. The larger the variance, the more information the principal component contains. Sort the eigenvalues ​​from large to small, and the corresponding eigenvectors are also arranged in the same order.

[0046] Select the first p principal components (p < l) according to the magnitudes of the eigenvalue such that the sum of the variances of these p principal components accounts for a certain threshold (e.g., 80% or 90%) of the total variance. These principal components are linear combinations of the original feature vectors and are linearly independent of each other. Project the standardized target high-dimensional feature vector onto these p principal components to obtain the target principal components.

[0047] Finally, the electronic device 3 inputs the target principal components into the input layer of a preset multi-layer perceptron model. After non-linear transformations through multiple hidden layers of the preset multi-layer perceptron model, the prediction result of the soil state data is finally obtained at the output layer. The MLP can learn the complex non-linear relationships between features and has good performance in dealing with complex soil grade classification problems.

[0048] The soil crust separation and analysis system provided by the embodiment of the present application extracts features from the sample data including soil crust thickness, surface feature data, crust spectral data, and side images, and converts the extracted features into corresponding feature vectors, ensuring the accuracy of the converted sample data including the feature vectors corresponding to the soil crust thickness, surface feature data, crust spectral data, and side images respectively. Then, splice the feature vectors to generate an initial high-dimensional feature vector, ensuring the accuracy of the generated initial high-dimensional feature vector. Then, based on a preset encoder, compress the initial high-dimensional feature vector to generate a compressed low-dimensional feature vector, ensuring that the generated compressed low-dimensional feature vector not only learns the features of the initial high-dimensional feature vector but also reduces the dimension. Based on a preset decoder, decode the compressed low-dimensional feature vector to generate a target high-dimensional feature vector, ensuring the accuracy of the generated target high-dimensional feature vector.

[0049] Then, perform principal component analysis on the target high-dimensional feature vector, and convert the target high-dimensional feature vector into a set of linearly independent target principal components through linear transformation, ensuring the accuracy of the converted target principal components. Extract features from the target principal components to determine the soil state data corresponding to the sample soil, ensuring the accuracy of the determined soil state data corresponding to the sample soil.

[0050] In an alternative embodiment of the present application, the electronic device 3 is further configured to obtain environmental data of the sample soil within a preset future time period; the environmental data includes at least one of temperature, precipitation, light intensity, and wind speed; Input the environmental data, the soil crust thickness, the surface feature data, the crust spectral data, and the side images into a preset soil crust prediction model, and output the future soil crust data of the sample soil within the preset future time period.

[0051] In an optional embodiment, the preset soil crust prediction model includes an input gate, a memory unit, a forget gate, an output gate, and a generative adversarial network, and the electronic device 3 is used to fuse the environmental data with the soil crust thickness, the surface feature data, the crust spectrum data, and the side image to generate an initial fusion feature; Input the initial fusion features into the preset soil crust prediction model according to the time series; The input gate extracts the target fusion feature from the current fusion feature in the initial fusion feature and adds it to the memory unit; The memory unit fuses the target fusion feature with the historical hidden state of the previous moment to generate an updated fusion feature; The forget gate determines the retained fusion features and the discarded fusion features from the updated fusion features in the memory unit; The output gate outputs the current hidden state based on the retained fusion features; Calculate the correlation between the hidden state corresponding to each moment and the future soil crust data; and determine the attention weight information corresponding to the hidden state corresponding to each moment; According to the attention weight information, the hidden state corresponding to each moment is weighted summed to obtain the target hidden state; The target hidden state is input into the generative adversarial network in the preset soil crust prediction model to output the future soil crust data.

[0052] Specifically, the electronic device 3 will obtain environmental data of the sample soil within a preset time in the future, and these environmental data include at least one of temperature, precipitation, light intensity, and wind speed. There are many ways to obtain these data, such as from the official database of the meteorological department, professional meteorological monitoring stations, or environmental sensors installed near the sample soil. These environmental data are crucial for predicting changes in soil crust, because the formation, development, and changes of soil crust are closely related to external environmental conditions. For example, the temperature will affect the evaporation rate of water in the soil, precipitation will directly change the humidity of the soil, light intensity will affect the temperature and microbial activity of the soil, and wind speed will affect the erosion and particle movement of the soil surface.

[0053] Then, the electronic device 3 fuses the acquired environmental data with the sample data (soil crust thickness, surface feature data, crust spectrum data, and side image) to generate an initial fusion feature. During the fusion process, different types of data may require different preprocessing. For example, environmental data is usually numerical data, and it may be necessary to normalize its value range to [0, 1] or [-1, 1] to eliminate the impact of dimensions between different data; while data such as side images may need to use image processing algorithms to extract features, and then fuse the extracted features with other data. The fusion method can use a simple splicing method to connect the feature vectors of different data in sequence, or a more complex feature fusion algorithm can be used, such as a fusion method based on a neural network, which trains a fusion network to learn the intrinsic relationship between different data, thereby generating a more representative initial fusion feature.

[0054] Next, the electronic device 3 inputs the initial fusion feature into the preset soil crust prediction model. The input gate in the preset soil crust prediction model extracts the target fusion feature from the current fusion feature in the initial fusion feature and adds it to the memory unit in the model. The input gate is like a filter, which uses a sigmoid activation function to determine which information can enter the memory unit. The output value of the sigmoid function is between 0 and 1. The closer the value is to 1, the more likely the information is to be selected to enter the memory unit, and the closer it is to 0, the more likely the information is to be discarded. In this way, the input gate can filter out information that is useful for predicting soil crust changes and pass it to the memory unit for further processing.

[0055] The memory unit in the preset soil crust prediction model fuses the target fusion feature with the historical hidden state of the previous moment in the memory unit to generate an updated fusion feature. The function of the memory unit is to store and transmit historical information. It can remember the impact of past inputs on the current state. When updating the fusion feature, the tanh activation function is usually used to transform the target fusion feature, and then weighted sum it with the historical hidden state of the previous moment. In this way, the current new information can be combined with the past historical information, so that the model can comprehensively consider the information at different time points for prediction.

[0056] The forget gate in the preset soil crust prediction model determines which fused features to keep and which to discard from the updated fused features in the memory unit. The forget gate also uses the sigmoid activation function, which determines which information in the memory unit needs to be retained and which information can be forgotten. As time goes by, some old information may no longer be helpful for the current prediction. The forget gate can remove this information from the memory unit to avoid information overload while retaining key information that is useful for prediction.

[0057] The output gate in the preset soil crust prediction model outputs the current hidden state based on the retained fusion features. The output gate also uses the sigmoid activation function to determine which information can be output from the memory unit as the current hidden state. The current hidden state contains the model's understanding and representation of the soil crust state at the current moment, which will serve as the basis for the next calculation.

[0058] Finally, the electronic device 3 calculates the correlation between the hidden state corresponding to each moment and the future soil crust data. The correlation can be calculated using various statistical methods, such as Pearson correlation coefficient, cosine similarity, etc. By calculating the correlation, the degree of influence of the hidden state at each moment on the future soil crust data can be understood.

[0059] The electronic device 3 determines the attention weight information corresponding to the hidden state corresponding to each moment according to the calculated correlation. The attention weight information reflects the importance of the hidden state at each moment in predicting future soil crust data. The softmax function is usually used to convert the correlation value into a probability distribution as the attention weight. The larger the weight, the greater the contribution of the hidden state at that moment to the prediction result.

[0060] The electronic device 3 performs weighted summation of the hidden states corresponding to each moment according to each attention weight information to obtain the target hidden state. Through weighted summation, the model can focus more on the hidden states at moments that are more important for predicting future soil crust data, thereby improving the accuracy of the prediction.

[0061] The target hidden state is input into the Generative Adversarial Network (GAN) in the preset soil crust prediction model to output the future soil crust data. The Generative Adversarial Network consists of a generator and a discriminator. The generator attempts to generate future soil crust data similar to the real soil crust data based on the target hidden state, and the discriminator is responsible for judging whether the generated data is real. Through adversarial training between the generator and the discriminator, the generator can continuously learn how to generate more realistic future soil crust data, and finally output the prediction results that meet the requirements, that is, the future soil crust data.

[0062] The soil crust separation and analysis system provided in the embodiment of the present application, the electronic device 3, is also used to obtain the environmental data of the sample soil within a preset time period in the future; the environmental data includes at least one of temperature, precipitation, light intensity, and wind speed; the environmental data is fused with the soil crust thickness, surface feature data, crust spectrum data, and side image to generate an initial fusion feature, thereby ensuring the accuracy of the generated target fusion feature. Then, the initial fusion feature is input into the preset soil crust prediction model in time series; the input gate in the preset soil crust prediction model extracts the target fusion feature from the current fusion feature in the initial fusion feature, and adds it to the memory unit in the preset soil crust prediction model, thereby ensuring the accuracy of the target fusion feature added to the memory unit. The memory unit fuses the target fusion feature with the historical hidden state of the previous moment in the memory unit to generate an updated fusion feature, thereby ensuring the accuracy of the generated updated fusion feature. The forget gate in the preset soil crust prediction model determines the retained fusion feature and the discarded fusion feature from the updated fusion feature in the memory unit, thereby ensuring the accuracy of the determined retained fusion feature and the discarded fusion feature. The output gate in the preset soil crust prediction model outputs the current hidden state based on the retained fusion features, ensuring the accuracy of the output current hidden state. Then, the correlation between the hidden state corresponding to each moment and the future soil crust data is calculated; according to each correlation, the attention weight information corresponding to the hidden state corresponding to each moment is determined, ensuring the accuracy of the attention weight information corresponding to the hidden state corresponding to each moment. Then, according to each attention weight information, the hidden state corresponding to each moment is weighted and summed to obtain the target hidden state, ensuring the accuracy of the obtained target hidden state. The target hidden state is input into the generative adversarial network in the preset soil crust prediction model, and the future soil crust data is output, ensuring the accuracy of the output future soil crust data, so that the impact of soil crust on the soil ecosystem can be understood in advance based on the future soil crust data, and reasonable soil management strategies can be formulated.

[0063] In an optional embodiment of the present application, Figure 2 As shown, the soil crust separation device 1 includes a ring knife assembly 11, a laser ranging device 12, and a separation assembly 13; the ring knife assembly 11 is installed below the soil crust separation and analysis system, the laser ranging device 12 is installed above the ring knife assembly 11, and the separation assembly 13 is slidably installed on the connecting rod between the ring knife assembly 11 and the laser ranging device 12; wherein: The ring knife assembly 11 is used for collecting sample soil with crust; wherein the inside of the ring knife assembly 11 is treated with a special hydrophilic polymer coating.

[0064] The laser ranging device 12 is used to control the sample soil to be lifted upward in the vertical direction, and to obtain a side image during the lifting process of the sample soil; determine the thickness of the soil crust according to the side image; and obtain surface feature data corresponding to the sample soil crust; The electronic device 3 is used to adjust the height of the separation component 13 according to the thickness of the soil crust, and control the separation component 13 to separate the sample soil to obtain the sample soil crust; The separation component 13 is used to separate the sample soil under the control of the electronic device 3 to obtain the sample soil crust.

[0065] In an optional embodiment, if Figure 3 As shown, the laser distance measuring device 12 includes: a camera assembly 121, a lifting platform 122, a laser distance measuring assembly 123 and a digital display assembly 124, wherein the laser distance measuring assembly 123 is installed above the ring knife assembly 11, the lifting platform 122 is installed below the ring knife assembly 11, and the digital display assembly 124 is installed on the outside of the lifting track corresponding to the lifting platform 122; the camera assembly 121 is installed on the outside of the lifting track corresponding to the lifting platform 122, and the lifting platform 122, the camera assembly 121 and the digital display assembly 124 are all connected to the electronic device 3; wherein: The lifting platform 122 is used to drive the sample soil to be lifted upward from the ring knife assembly 11 under the control of the electronic device 3 after the ring knife assembly 11 has collected the sample soil; The electronic device 3 is used to control the camera assembly 121 to collect side images in real time when the sample soil is lifted upward; The electronic device 3 is also used to identify each side image and determine the contact interface between the sample soil crust and the lower soil in the sample soil; according to the identified contact interface, the lifting platform 122 is controlled to drive the sample soil to be lifted upward in the vertical direction until the contact interface is flush with the upper surface of the ring knife assembly 11 and stops; The laser distance measuring component 123 is used to measure the lifting distance of the lifting platform 122 under the control of the electronic device 3, and determine the lifting distance as the soil crust thickness corresponding to the sample soil; The laser ranging component 123 is also used to collect surface feature data corresponding to the sample soil under the control of the electronic device 3.

[0066] In an optional embodiment, if Figure 4 As shown, the laser distance measuring assembly 123 includes: a laser sensor 1231, an annular guide rail 1232, a micro-seismic buffer assembly 1233, a horizontal bar 1234, a first motor 1235 and a steel frame 1236; wherein: The diameter of the annular guide rail 1232 is the same as the size of the annular cutter in the annular cutter assembly 11 and is used to carry the laser sensor 1231; The micro-vibration buffer assembly 1233 is installed on the annular guide rail 1232 and is composed of a flexible silicone pad and a micro spring, and is used to reduce the interference of the vibration generated when the first motor 1235 drives the horizontal bar 1234 to slide on the laser sensor 1231; The first motor 1235 is used to drive the horizontal bar 1234 to slide, so that the horizontal bar 1234 drives the laser sensor 1231 to slide along the steel frame 1236; The laser sensor 1231 is used to emit a first infrared light and receive a first reflected light corresponding to the first infrared light before the lifting platform 122 drives the sample soil to lift upward; and emit a second infrared light and receive a second reflected light corresponding to the second infrared light after the lifting platform 122 stops; An electronic device 3, used for calculating the soil crust thickness based on the emission time of the first infrared light and the reception time of the first reflected light and the emission time of the second infrared light and the reception time of the second reflected light; The laser sensor 1231 is also used to emit visible light to the sample soil crust, and generate a reflection image corresponding to the sample soil crust based on a visible reflection signal corresponding to the received visible light; The electronic device 3 is used to identify the reflected image and determine the surface characteristic data corresponding to the soil crust of the sample, wherein the surface characteristic data includes at least one of roughness, particle distribution, and crack characteristics.

[0067] In an optional embodiment, the electronic device 3 is used to identify the reflection image based on a preset generative adversarial network to determine the roughness corresponding to the soil crust of the sample; and / or, to identify the reflection image based on a time series crack dynamic monitoring and analysis method to determine the crack characteristics corresponding to the soil crust of the sample; and / or, to identify the reflection image based on a preset target detection algorithm to determine the particle distribution corresponding to the soil crust of the sample.

[0068] In an optional embodiment, if Figure 5 As shown, the separation component 13 includes: a sliding component 131, a separation slide rail 132, a separation knife 133, a second motor 134, a second pressure sensor 135, and a processor 136, wherein the sliding component 131 is installed on the connecting rod between the ring knife component 11 and the laser ranging device 12, the sliding component 131 drives the separation slide rail 132 to slide on the connecting rod, and the separation knife 133 is installed on the separation slide rail 132. After the sliding component 131 drives the separation slide rail 132 to be fixed to a specific position according to the thickness of the soil crust, the second motor 134 drives the separation knife 133 to slide on the separation slide rail 132 to separate the sample soil and obtain the sample soil crust corresponding to the sample soil; wherein: The second pressure sensor 135 is used to collect resistance data during the cutting process when the separation knife 133 separates the sample soil to obtain a sample soil crust corresponding to the sample soil, and transmit the resistance data to the processor 136 and the electronic device 3; The processor 136 is used to adjust the blade angle and cutting force of the separation knife 133 according to the resistance data.

[0069] Specifically, after the ring knife assembly 11 completes sample collection, the lifting platform 122 starts working. The electronic device 3 sends a command to the lifting platform 122, and the driving motor drives the lifting platform 122 to rise, thereby driving the sample soil to be lifted upward from the ring knife assembly 11. During the lifting process, the electronic device 3 controls the camera assembly 121 to continuously capture side images at a certain frame rate. The camera assembly 121 converts the image of the side of the sample soil crust into a digital image signal through the principle of optical imaging, and transmits it to the electronic device 3 in real time. The electronic device 3 uses an image processing algorithm to analyze the side image, identify the contact interface between the soil crust in the sample soil and the lower soil, and when the contact interface is flush with the upper surface of the ring knife assembly 11, the lifting platform 122 is controlled to stop lifting.

[0070] The specific identification process can be as follows: the electronic device 3 first performs noise reduction on the collected sample soil crust side view to eliminate the noise introduced by factors such as environmental interference and device sensor error. The Gaussian filtering algorithm is used to construct a Gaussian kernel function to perform weighted averaging on each pixel and its neighborhood in the sample soil crust side view. For example, for a 3×3 Gaussian kernel, the weight of the central pixel is the highest, and the weight of the surrounding pixels gradually decreases, so as to smooth the image, remove salt and pepper noise and Gaussian noise, etc., make the image clearer, and provide a good foundation for subsequent processing.

[0071] Then, the electronic device 3 converts the color side image into a grayscale side image to simplify the computational complexity. A weighted average method can be used to convert color pixels into grayscale values ​​by assigning different weights, such as 0.299, 0.587, and 0.114, according to the sensitivity differences of the human eye to the three primary colors of red, green, and blue. In order to highlight the boundary between the soil crust and the underlying soil, a histogram equalization algorithm is used to redistribute the grayscale values ​​of the image, making the grayscale distribution more uniform, enhancing the contrast, and making the contact interface easier to identify in the image.

[0072] Next, the electronic device 3 can use the Canny edge detection algorithm to extract edge information in the grayscale side image. The algorithm first calculates the gradient amplitude and direction of the grayscale side image, refines the edge by non-maximum suppression, and then uses dual threshold detection to determine the true edge. For example, a high threshold and a low threshold are set, and pixels with a gradient amplitude higher than the high threshold are determined as strong edges, and pixels below the low threshold are discarded. Pixels between the two are retained if they are connected to a strong edge, otherwise they are discarded, so as to accurately outline the potential boundary between the soil crust and the underlying soil.

[0073] To further enhance the edge features, the electronic device 3 may perform morphological operations. Specifically, the electronic device 3 uses a dilation operation to expand the edge through a structural element (such as a rectangle or a circle) to connect the broken edges; and then performs an erosion operation to remove isolated noise points and small burrs to make the edge more continuous and accurate. For example, selecting a 3×3 rectangular structural element for dilation and erosion operations can effectively optimize the edge features.

[0074] Finally, the electronic device 3 uses a threshold segmentation algorithm based on the image after edge detection and morphological operations, combined with the grayscale difference between the soil crust and the lower soil. For example, the Otsu algorithm can automatically calculate an optimal threshold value to divide the image into the foreground (soil crust) and the background (lower soil), thereby determining the contact interface. If the grayscale value of the soil crust is higher than that of the lower soil, pixels greater than the threshold are classified as soil crust, and pixels less than the threshold are classified as lower soil, and the boundary between the two is the contact interface.

[0075] Then, the laser distance measuring component 123 uses the propagation characteristics of light to measure the thickness of the soil crust. Before the lifting platform 122 drives the sample soil to lift upward, the laser sensor 1231 in the laser distance measuring component 123 emits a first infrared light, and the infrared light is reflected back after encountering the upper surface of the sample soil crust, and the laser sensor 1231 receives the first reflected light. The electronic device 3 calculates the first distance between the laser sensor 1231 and the upper surface of the sample soil crust before the sample soil is lifted by a formula based on the propagation speed of light, the emission time of the first infrared light, and the reception time of the first reflected light. Similarly, after the lifting platform 122 stops, the laser sensor 1231 emits a second infrared light and receives the second reflected light, and calculates the second distance between the laser sensor 1231 and the upper surface of the sample soil crust after the sample soil is lifted. Finally, the electronic device 3 subtracts the second distance from the first distance to obtain the soil crust thickness corresponding to the sample soil.

[0076] In addition, the laser sensor 1231 can also emit visible light to the sample soil crust. After the visible light is reflected on the surface of the soil crust, the laser sensor 1231 receives a visible reflection signal corresponding to the visible light, thereby generating a reflection image corresponding to the sample soil crust. For the reflection image, the electronic device 3 uses different algorithms for analysis.

[0077] Specifically, in order to improve the accuracy of subsequent analysis, the electronic device 3 can pre-process the reflected image. The noise in the image is removed by using a filtering algorithm. The median filter can effectively eliminate the salt and pepper noise, and the Gaussian filter is suitable for removing Gaussian noise to make the reflected image clearer. The contrast of the image is enhanced by grayscale transformation to make the texture features more obvious. For example, the histogram equalization method is used to redistribute the grayscale values ​​of the image, stretch the grayscale range, and highlight the details of the soil crust surface. In addition, an image defogging algorithm is used to remove image blur caused by environmental factors, restore the real texture information, and make the cracks more obvious in the image.

[0078] Then, the electronic device 3 inputs the reflection image corresponding to the sample soil crust whose roughness is to be determined into the trained preset generative adversarial network. Among them, the reflection image here needs to have the same preprocessing method as the training data to ensure the consistency of the data format and feature scale. The generator in the preset generative adversarial network generates a simulated soil crust image based on the characteristics of the input reflection image. This soil crust image is generated based on the roughness feature pattern learned by the generator during the training process, and it attempts to imitate the roughness performance of the real image.

[0079] The discriminator evaluates the input real reflection image and the simulated soil crust image generated by the generator, and outputs the probability value of the two images being real images. By comparing these two probability values ​​and the performance of the generator and the discriminator during the training process, the roughness of the sample soil crust can be inferred. For example, if the probability value of the image generated by the generator in the discriminator is close to the probability value of the real image, it means that the generator can simulate the roughness characteristics of the current sample well. Combined with the relationship between the roughness learned by the generator during the training process and the generated image, the roughness corresponding to the sample soil crust can be determined. In addition, some quantitative indicators such as mean square error (MSE) and peak signal-to-noise ratio (PSNR) can be used to measure the difference between the generated image and the real image, and further accurately determine the roughness of the soil crust.

[0080] Among them, the training process of the preset generative adversarial network includes: collecting a large number of soil crust reflection images with different roughness as training data. These images need to be preprocessed, such as normalization, to unify the pixel values ​​of the images into a specific range, such as [0, 1] or [-1, 1], to improve the training effect and stability. During the training process, the generator and the discriminator play an adversarial game. The generator strives to generate more realistic images to deceive the discriminator, while the discriminator strives to improve the discrimination ability and accurately identify the generated images. Specifically, the generator generates images based on random noise, and the discriminator judges the real image and the generated image. The loss function of the discriminator is based on its judgment results on the real image and the generated image, hoping to output 1 for the real image and 0 for the generated image. The loss function of the generator is based on the judgment results of the discriminator on the generated image, hoping that the discriminator will misjudge the generated image as a real image, that is, output a probability value close to 1. Through the back propagation algorithm, the parameters of the generator and the discriminator are updated respectively, so that their capabilities are continuously improved. During the training process, some optimization strategies can also be adopted, such as adjusting the learning rate and using regularization methods, to prevent the model from overfitting and improve the generalization ability of the model. Based on the preset generative adversarial network (GAN), through the adversarial training of the generator and the discriminator, the model learns the characteristic pattern of the surface roughness of the soil crust, thereby determining the roughness corresponding to the sample soil crust.

[0081] In addition, the electronic device 3 calculates the gradient amplitude and direction of the reflected image, and uses non-maximum suppression and dual threshold detection steps to accurately detect the edge of the crack. During the detection process, it is necessary to reasonably adjust the parameters of the algorithm, such as high and low thresholds, according to the characteristics of the image and actual needs, to ensure that the detected crack edge is complete and accurate.

[0082] For the detected cracks, the electronic device 3 can calculate characteristic parameters such as crack length, width, area, and number. The crack length can be obtained by counting and calculating the pixel points at the edge of the crack; the crack width can be measured in the direction perpendicular to the crack; the crack area is calculated by accumulating the pixel points in the crack area; and the number of cracks is obtained by marking and counting the detected cracks. Through these quantitative characteristic parameters, the state of soil crust cracks can be fully described.

[0083] In addition, the electronic device 3 can also input the reflected image into the trained target detection model. The convolution layer and the pooling layer in the target detection model extract the image features of the reflected image, and then perform target detection according to different algorithms. Taking YOLO as an example, the prediction results of each grid are calculated to obtain a series of bounding boxes and category probabilities. The bounding boxes with high overlap are removed by the non-maximum suppression (NMS) algorithm to retain the most accurate detection results.

[0084] According to the detected particle bounding box, the distribution density of the particles is calculated. The target detection model can divide the image into multiple sub-areas, count the number of particles in each sub-area, and obtain a spatial distribution density map of the particles. For the particle size distribution, the equivalent diameter or area of ​​each particle is calculated by the size of the bounding box, and a histogram of the particle size is drawn to analyze the proportion of particles of different sizes. In addition, the electronic device 3 can also introduce spatial autocorrelation analysis to study the spatial correlation of particle distribution and determine whether the particles are randomly distributed, clustered or uniformly distributed.

[0085] Finally, the electronic device 3 can calculate the specific position to which the separation component 13 needs to be moved based on the soil crust thickness measured by the laser ranging device 12. Then the sliding component 131 is controlled to drive the separation slide rail 132 to slide on the connecting rod to the specific position, completing the positioning preparation work before separation.

[0086] Then, the second motor 134 drives the blade of the separation knife 133 to slide on the separation slide rail 132 to cut and separate the sample soil. During the cutting process, the second pressure sensor 135 can collect resistance data during the cutting process in real time. The second pressure sensor 135 converts the cutting resistance into an electrical signal based on the piezoelectric effect or the strain gauge principle, and transmits the data to the processor 136 and the electronic device 3.

[0087] The processor 136 calculates the blade angle and cutting force adjustment value required by the separation knife 133 under the current cutting state through an algorithm based on the received resistance data. Then the actuator of the separation knife 133 is controlled to adjust the blade angle and cutting force to ensure that the separation process proceeds smoothly. If the resistance is too large, it may be that the blade angle is not appropriate or the cutting force is too large, causing the soil crust to be damaged. At this time, the processor 136 will reduce the cutting force and adjust the blade angle; if the resistance is too small, the cutting force may be insufficient, resulting in incomplete separation, and the processor 136 will increase the cutting force. The electronic device 3 can also monitor and adjust the entire separation process according to the pressure data to ensure the separation quality.

[0088] In an optional embodiment, the sliding assembly 131 of the separation assembly 13 can be equipped with a high-precision linear magnetic encoder. The linear magnetic encoder can accurately measure the position of the sliding module in real time, and its accuracy is higher than that of the traditional photoelectric encoder, which can reach ±0.01mm. This enables the cutting device to be more accurately positioned when automatically adjusting the cutting height according to the thickness of the crust, ensuring the accuracy of the cutting position and further improving the accuracy of soil crust separation.

[0089] The soil crust separation and analysis system provided by the embodiment of the present application, the soil crust separation device 1 includes a ring knife assembly 11, a laser ranging device 12, and a separation assembly 13, wherein the ring knife assembly 11 is installed below the soil crust separation and analysis system, the laser ranging device 12 is installed above the ring knife assembly 11, and the separation assembly 13 is slidably installed on the connecting rod between the ring knife assembly 11 and the laser ranging device 12; the laser ranging device 12 and the separation assembly 13 are respectively connected to the electronic device 3, wherein: the ring knife assembly 11 is used to collect sample soil with crust; wherein the inside of the ring knife assembly 11 is treated with a special hydrophilic polymer coating. The lifting platform 122 is used to drive the sample soil to be lifted upward from the ring knife assembly 11 under the control of the electronic device 3 after the ring knife assembly 11 has collected the sample soil, so as to facilitate the separation of the sample soil and obtain the sample soil crust corresponding to the sample soil. The electronic device 3 is used to control the camera assembly 121 to collect the side image in real time when the sample soil is lifted upward, so as to ensure the accuracy of the acquired side image of the soil crust. The electronic device 3 is also used to identify each side image, determine the contact interface between the sample soil crust and the lower soil in the sample soil, and ensure the accuracy of the contact interface between the sample soil crust and the lower soil in the determined sample soil. According to the identified contact interface, the lifting platform 122 is controlled to drive the sample soil to be lifted upward in the vertical direction until the contact interface is flush with the upper surface of the ring knife assembly 11 and stops, so as to facilitate the measurement of the soil crust thickness and the separation of the sample soil to obtain the sample soil crust corresponding to the sample soil. The laser ranging assembly 123 is used to measure the lifting distance of the lifting platform 122 under the control of the electronic device 3, and determine the lifting distance as the soil crust thickness corresponding to the sample soil, so as to ensure the accuracy of the soil crust thickness corresponding to the determined sample soil. The laser ranging assembly 123 is also used to collect surface feature data corresponding to the sample soil under the control of the electronic device 3, so that the sample soil crust can be studied based on the surface feature data, thereby contributing to agricultural soil science research.

[0090] The electronic device 3 is used to store the soil crust thickness and surface characteristic data, and adjust the height of the separation component 13 according to the soil crust thickness, and control the separation component 13 to separate the sample soil to obtain the sample soil crust corresponding to the sample soil.

[0091] The diameter of the annular guide rail 1232 is the same as the size of the annular knife in the annular knife assembly 11, and is used to carry the laser sensor 1231. The micro-seismic buffer assembly 1233, which is installed on the annular guide rail 1232, is composed of a flexible silicone pad and a micro spring, and is used to reduce the interference of the vibration generated when the first motor 1235 drives the horizontal bar 1234 to slide on the laser sensor 1231. The first motor 1235 is used to drive the horizontal bar 1234 to slide, so that the horizontal bar 1234 can drive the laser sensor 1231 to slide along the steel frame 1236. The diameter of the annular guide rail 1232 is the same as the size of the annular knife in the annular knife assembly 11, and is used to carry the laser sensor 1231. The micro-seismic buffer assembly 1233, which is installed on the annular guide rail 1232, is composed of a flexible silicone pad and a micro spring, and is used to reduce the interference of the vibration generated when the first motor 1235 drives the horizontal bar 1234 to slide on the laser sensor 1231. The first motor 1235 is used to drive the horizontal bar 1234 to slide, so that the horizontal bar 1234 can drive the laser sensor 1231 to slide along the steel frame 1236. The laser sensor 1231 is used to emit a first infrared light and receive a first reflected light corresponding to the first infrared light before the lifting platform 122 drives the sample soil to lift upward; after the lifting platform 122 stops, it emits a second infrared light and receives a second reflected light corresponding to the second infrared light. The electronic device 3 is used to determine the first distance between the laser sensor 1231 and the upper surface of the sample soil crust before the sample soil is lifted based on the emission time of the first infrared light and the reception time of the first reflected light, thereby ensuring the accuracy of the determined first distance. Based on the emission time of the second infrared light and the reception time of the second reflected light, the second distance between the laser sensor 1231 and the upper surface of the sample soil crust after the sample soil is lifted is determined, thereby ensuring the accuracy of the determined second distance. The soil crust thickness corresponding to the sample soil is calculated by subtracting the second distance from the first distance, thereby ensuring the accuracy of the obtained soil crust thickness. The laser sensor 1231 is also used to emit visible light to the sample soil crust, and generate a reflection image corresponding to the sample soil crust based on the visible reflection signal corresponding to the received visible light, thereby ensuring the accuracy of the generated reflection image corresponding to the sample soil crust. The electronic device 3 is used to identify the reflection image based on the preset generative adversarial network, determine the roughness corresponding to the sample soil crust, and ensure the accuracy of the roughness corresponding to the determined sample soil crust. The dynamic crack monitoring and analysis method based on the time series is used to identify the reflection image, determine the crack characteristics corresponding to the sample soil crust, and ensure the accuracy of the crack characteristics corresponding to the determined sample soil crust. The reflection image is identified based on the preset target detection algorithm, and the particle distribution corresponding to the sample soil crust is determined, thereby ensuring the accuracy of the particle distribution corresponding to the determined sample soil crust.

[0092] The separation component 13 includes: a sliding component 131, a separation slide rail 132, a separation knife 133, a second motor 134, a second pressure sensor 135, and a processor 136, wherein the sliding component 131 is installed on the connecting rod between the ring knife component 11 and the laser ranging device 12, the sliding component 131 drives the separation slide rail 132 to slide on the connecting rod, and the separation knife 133 is installed on the separation slide rail 132. After the sliding component 131 drives the separation slide rail 132 to be fixed to a specific position according to the thickness of the soil crust, the second motor 134 drives the separation knife 133 to slide on the separation slide rail 132 to separate the sample soil and obtain a sample soil crust corresponding to the sample soil; wherein: the second pressure sensor 135 is used to collect resistance data during the cutting process when the separation knife 133 separates the sample soil to obtain a sample soil crust corresponding to the sample soil, and transmit the resistance data to the processor 136 and the electronic device 3. The processor 136 is used to adjust the blade angle and cutting force of the separation knife 133 according to the resistance data, thereby ensuring the accuracy of adjusting the blade angle and cutting force of the separation knife 133.

[0093] Although the embodiments of the present invention have been described in conjunction with the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present invention, and such modifications and variations are all within the scope defined by the appended claims.

Claims

1. A soil crust separation and analysis system, characterized in that: include: Soil crust separation equipment, crust spectrum collection equipment and electronic equipment, including: The soil crust separation device is used to obtain sample soil, and during the lifting process of the sample soil, obtain a side image corresponding to the sample soil; determine the thickness of the soil crust corresponding to the sample soil according to the side image, and separate the sample soil to obtain the sample soil crust; and obtain surface feature data corresponding to the sample soil crust; The crust spectrum acquisition device is used to obtain crust spectrum data corresponding to the sample soil crust; The electronic device is used to process the soil crust thickness, the side image, the surface characteristic data and the crust spectrum data to determine the soil state data corresponding to the sample soil; and output corresponding processing measures according to the soil state data.

2. The soil crust separation and analysis system according to claim 1, characterized in that: The electronic device is used for Performing feature fusion on the soil crust thickness, the surface feature data, the crust spectrum data and the side image to generate a target high-dimensional feature vector; Performing principal component analysis on the target high-dimensional feature vector, and converting the target high-dimensional feature vector into a set of linearly independent target principal components through linear transformation; Feature extraction is performed on the target principal component to determine the soil state data corresponding to the sample soil.

3. The soil crust separation and analysis system according to claim 2, characterized in that: The electronic device is used for Extracting features from the sample data including soil crust thickness, surface feature data, crust spectrum data, and side images, and converting the extracted features into corresponding feature vectors; Concatenate the feature vectors to generate an initial high-dimensional feature vector; Based on a preset encoder, compressing the initial high-dimensional feature vector to generate a compressed low-dimensional feature vector; Based on a preset decoder, the compressed low-dimensional feature vector is decoded to generate the target high-dimensional feature vector.

4. The soil crust separation and analysis system according to claim 1, characterized in that: The electronic device is further used to obtain environmental data of the sample soil within a preset time period in the future; the environmental data includes at least one of temperature, precipitation, light intensity, and wind speed; The environmental data, the soil crust thickness, the surface characteristic data, the crust spectrum data and the side image are input into a preset soil crust prediction model, and the future soil crust data within the future preset time period corresponding to the sample soil is output.

5. The soil crust separation and analysis system according to claim 4, characterized in that: The preset soil crust prediction model includes an input gate, a memory unit, a forget gate, an output gate, and a generative adversarial network, and the electronic device is used to fuse the environmental data with the soil crust thickness, the surface feature data, the crust spectrum data, and the side image to generate an initial fusion feature; Inputting the initial fusion features into the preset soil crust prediction model according to the time series; The input gate extracts a target fusion feature from the current fusion feature in the initial fusion feature, and adds the target fusion feature to the memory unit; The memory unit fuses the target fusion feature with the historical hidden state at the previous moment to generate an updated fusion feature; The forget gate determines to retain the fused features and to discard the fused features from the updated fused features; The output gate outputs the current hidden state based on the retained fusion feature; Calculate the correlation between the hidden state corresponding to each moment and the future soil crust data, and determine the attention weight information corresponding to the hidden state corresponding to each moment; According to each of the attention weight information, weighted summation is performed on the hidden states corresponding to each moment to obtain a target hidden state; The target hidden state is input into the generative adversarial network, and the future soil crust data is output.

6. The soil crust separation and analysis system according to claim 1, characterized in that: The soil crust separation device comprises a ring knife assembly, a laser ranging device, and a separation assembly; the ring knife assembly is installed below the soil crust separation and analysis system, the laser ranging device is installed above the ring knife assembly, and the separation assembly is slidably installed on a connecting rod between the ring knife assembly and the laser ranging device; wherein: The ring knife assembly is used to collect the sample soil with crust; The laser ranging device is used to control the sample soil to be lifted upward in the vertical direction, and to obtain the side image during the lifting process of the sample soil; determine the thickness of the soil crust according to the side image; and obtain surface feature data corresponding to the sample soil crust; The electronic device is used to adjust the height of the separation component according to the thickness of the soil crust, and control the separation component to separate the sample soil to obtain the sample soil crust; The separation component is used to separate the sample soil under the control of the electronic device to obtain the sample soil crust.

7. The soil crust separation and analysis system according to claim 6, characterized in that: The laser distance measuring device comprises: a camera assembly, a lifting platform, a laser distance measuring assembly and a digital display assembly; the laser distance measuring assembly is installed above the ring knife assembly, the lifting platform is installed below the ring knife assembly, and the digital display assembly is installed on the outside of the lifting track corresponding to the lifting platform; the camera assembly is installed on the outside of the lifting track corresponding to the lifting platform; wherein: The lifting platform is used to drive the sample soil to be lifted upward from the ring knife assembly under the control of the electronic device after the ring knife assembly has collected the sample soil; The electronic device is used to control the camera assembly to collect the side image in real time when the sample soil is lifted upward; The electronic device is further used to identify each of the side images to determine the contact interface between the sample soil crust and the lower soil in the sample soil; according to the identified contact interface, the lifting platform is controlled to drive the sample soil to be lifted upward in the vertical direction until the contact interface is flush with the upper surface of the ring knife assembly and stops; The laser distance measuring component is used to measure the lifting distance of the lifting platform, and determine the lifting distance as the soil crust thickness corresponding to the sample soil; The laser ranging component is also used to collect the surface feature data corresponding to the sample soil.

8. The soil crust separation and analysis system according to claim 7, characterized in that: The laser distance measurement assembly comprises: a laser sensor, a circular guide rail, a micro-seismic buffer assembly, a crossbar, a first motor and a steel frame; wherein: The diameter of the annular guide rail is the same as the size of the annular cutter in the annular cutter assembly and is used to carry the laser sensor; The micro-seismic buffer assembly is installed on the annular guide rail and is composed of a flexible silicone pad and a micro spring; The first motor is used to drive the horizontal bar to slide, so that the horizontal bar drives the laser sensor to slide along the steel frame; The laser sensor is used to emit a first infrared light and receive a first reflected light corresponding to the first infrared light before the lifting platform drives the sample soil to lift upward; and emit a second infrared light and receive a second reflected light corresponding to the second infrared light after the lifting platform stops; The electronic device is used to calculate the soil crust thickness based on the emission time of the first infrared light and the reception time of the first reflected light and the emission time of the second infrared light and the reception time of the second reflected light; The laser sensor is further used to emit visible light to the sample soil crust, and generate a reflection image corresponding to the sample soil crust based on a visible reflection signal corresponding to the received visible light; The electronic device is used to identify the reflected image and determine the surface feature data corresponding to the soil crust of the sample, wherein the surface feature data includes at least one of roughness, particle distribution, and crack characteristics.

9. The soil crust separation and analysis system according to claim 8, characterized in that: The electronic device is used for Recognize the reflection image based on a preset generative adversarial network to determine the roughness corresponding to the soil crust of the sample; and / or, The reflection image is identified based on a time series crack dynamic monitoring and analysis method to determine the crack characteristics corresponding to the sample soil crust; and / or, The reflection image is identified based on a preset target detection algorithm to determine the particle distribution corresponding to the soil crust of the sample.

10. The soil crust separation and analysis system according to claim 6, characterized in that: The separation component includes: a sliding component, a separation slide rail, a separation knife, a second motor, a second pressure sensor, and a processor; the sliding component is installed on a connecting rod between the ring knife component and the laser ranging device, the sliding component drives the separation slide rail to slide on the connecting rod, the separation knife is installed on the separation slide rail, after the sliding component drives the separation slide rail to be fixed to a specific position according to the thickness of the soil crust, the second motor drives the separation knife to slide on the separation slide rail to separate the sample soil and obtain the sample soil crust corresponding to the sample soil; wherein: The second pressure sensor is used for collecting resistance data during the cutting process when the separation knife separates the sample soil to obtain a sample soil crust corresponding to the sample soil, and transmitting the resistance data to the processor and the electronic device; The processor is used to adjust the blade angle and cutting force of the separation knife according to the resistance data.

Images