An intelligent cultivation control system for aerial strawberries

Abstract

The invention relates to the technical field of cultivation control, and discloses an intelligent cultivation control system for aerial strawberries. In a first preset time period, the inner roof of a target planting shed is evenly divided into M sub-areas, and an indication ribbon is hung on each sub-area; in the first preset time period, test control is performed in the target planting shed to obtain a corresponding hanging image of the indication ribbon and the falling point coordinates of dust, and a falling point prediction model is obtained based on training of several groups of hanging images and falling point coordinates; in a second preset time period, the coordinates of a target tourist in the target planting shed are obtained, and a warning range of the target tourist is constructed based on the coordinates; by controlling the rotation speed, horizontal inclination angle and vertical inclination angle of a fan in the target planting shed, it is prevented that the falling point coordinates of dust are located within the warning range of the target tourist when the hanging planting rack is ascended and descended.

Description

Technical Field

[0001] The invention relates to the technical field of cultivation control, and more specifically, to an intelligent cultivation control system for aerial strawberries. Background Art

[0002] With the development of agricultural technology and the integration of tourism, aerial strawberry cultivation has gradually emerged as a modern agricultural planting model. This model not only improves planting efficiency and saves space by planting strawberries on hanging racks, but also provides tourists with a good picking experience. However, there is a problem of mud or debris falling during the lifting process of the strawberry planting rack. These fallen objects will float down with the airflow in the greenhouse. If there are tourists at the landing point, it will cause the problem of contamination to tourists. Summary of the invention

[0003] The invention provides an aerial strawberry intelligent cultivation control system to solve the technical problems raised in the background technology.

[0004] The present invention provides an aerial strawberry intelligent cultivation control system, comprising:

[0005] An area division module is used to evenly divide the inner roof of the target planting shed into M sub-areas within a first preset time period, and hang an indicator ribbon on each sub-area;

[0006] The landing point acquisition module is used to perform test control in the target planting shed within a first preset time period to obtain the corresponding hanging image of the indicator ribbon and the landing point coordinates of the dust, specifically including:

[0007] T1: random control parameters of the speed, horizontal inclination and vertical inclination of the fan in the target planting shed;

[0008] T2: Randomly control the rise or fall of any hanging planting rack in the target planting shed;

[0009] T3: Obtain the hanging image of the indicator ribbon and the coordinates of the dust falling point corresponding to T1 and T2;

[0010] Repeat T1, T2 and T3 for a preset number of times to obtain several sets of hanging images and landing point coordinates;

[0011] A prediction module, used for training a landing point prediction model based on a plurality of sets of hanging images and landing point coordinates;

[0012] A control and adjustment module, used to obtain the coordinates of the target tourist in the target planting shed within a second preset time period, and to construct a warning range of the target tourist based on the coordinates;

[0013] Based on the falling point prediction model, by controlling the speed, horizontal inclination and vertical inclination of the fan in the target planting shed, the falling point coordinates of the dust are prevented from being within the warning range of the target tourists when the hanging planting racks are rising and falling.

[0014] Furthermore, obtaining the coordinates of the dust landing point also includes:

[0015] Allocate indicator dust of different colors to each hanging planting rack in the target planting shed, and obtain a ground image of the target planting shed when the indicator dust rises or falls with the hanging planting rack;

[0016] Establishing a landing point coordinate system based on the ground image, the landing point coordinate system includes an X-axis and a Y-axis;

[0017] Based on the color of the indicating dust in the ground image, the coordinates of the landing point corresponding to the target planting rack are determined.

[0018] Furthermore, the speed, horizontal inclination and vertical inclination of the fan in the target planting shed are randomly controlled, including:

[0019] Determine the maximum and minimum speeds of the fans in the target growing shed;

[0020] Determine the horizontal allowable activity angle of the fan in the target planting shed;

[0021] Determine the vertical allowable motion angle of the fan in the target growing shed;

[0022] The maximum speed, minimum speed, horizontal allowable activity angle and vertical allowable activity angle of the fan are taken as constraints, and the speed, horizontal inclination angle and vertical inclination angle of the fan are randomly generated within the constraints.

[0023] Furthermore, a landing point prediction model is obtained based on several sets of hanging images and landing point coordinates, including:

[0024] Identify the M hanging images of the i-th group, and obtain the angle parameters of the indicator ribbon in each hanging image, the angle parameters include the degree of the angle between the indicator ribbon and the binding end of the indicator ribbon and the vertical line of the ground, and the direction of the angle; perform real number coding on the rise and fall of the hanging planting rack, and mark the state code corresponding to the i-th group of hanging images;

[0025] Establish a hanging image feature vector of group i, the feature vector includes:

[0026] The degree of the angle corresponding to a hanging image of the i-th group is taken as the first element of the feature vector, the direction of the angle is encoded by a real number as the second element of the feature vector, and the state encoding is taken as the third element of the feature vector;

[0027] The M eigenvectors are stacked to obtain a feature matrix, and the feature matrix is ​​normalized to obtain a standard matrix;

[0028] The standard matrix of the i-th group is used as the training sample of the landing prediction model, and the landing point coordinates of the i-th group are used as the sample labels of the landing prediction model; the landing prediction model is trained based on the back propagation optimization of the mean square error loss function;

[0029] The landing point prediction model includes an input layer, a hidden layer and a classification output layer.

[0030] Furthermore, the hidden layer and the classification output layer include:

[0031] The calculation formula of the hidden layer is as follows:

[0032] ;

[0033] in, represents the hidden layer The hidden state output by the hidden units, express Activation function, represents the total number of hidden units in the hidden layer, express The index of Indicates The weight matrix of hidden units, Indicates The bias matrix of hidden units, represents the standard matrix of input, represents the hidden layer The hidden state output by the hidden units, Indicates that right Perform convolution operation;

[0034] The classification output layer includes a classifier, which is used to input the The hidden state of the hidden unit output, and the classification space of the classifier represents the predicted landing point coordinates.

[0035] Furthermore, the warning range of the target tourists is constructed based on the coordinates, including:

[0036] Obtain the picking route of the target planting shed and the position of the target tourist, and associate the position of the target tourist with the corresponding position of the picking route; obtain the time for the hanging planting racks in the target planting shed to rise or fall, and the average moving speed of tourists in the historical time period; determine the warning moving distance of the target tourist based on the average moving speed and the time for the hanging planting racks to rise or fall; construct a characteristic circle with the corresponding position of the target tourist on the picking route as the center and the warning moving distance as the radius, and use the range of the characteristic circle as the warning range of the target tourist.

[0037] Further, the dust drop coordinates when the hanging planting rack is raised and lowered are prevented from being within the warning range of the target visitors, including:

[0038] Obtain the position and warning range of the target visitor at the zth moment in the second preset time period, and determine the hanging planting rack that needs to be raised or lowered;

[0039] Acquire the hanging image of the indicating ribbon corresponding to the z-th moment in the second preset time period, and obtain the predicted landing point coordinates corresponding to the hanging planting rack based on the landing point prediction model;

[0040] If the predicted landing point coordinates coincide with the warning range of the target tourist, the hanging planting rack will not be raised or lowered, and the first plan will be activated;

[0041] The first option includes:

[0042] E1, randomly adjust the speed, horizontal inclination and vertical inclination of the fan, obtain the corresponding hanging image, and determine the predicted landing point coordinates corresponding to the hanging image;

[0043] E2, repeat E1 until the predicted landing point coordinates and the warning range do not overlap, then start the second plan;

[0044] The second solution includes: controlling the corresponding hanging planting rack to rise or fall.

[0045] The beneficial effect of the present invention is that dust pollution control in the process of aerial strawberry cultivation is achieved through an intelligent control system, ensuring the safety experience of tourists during the picking process. The system uses fan adjustment, hanging planting rack lifting control and dust landing prediction technology to accurately control the distribution of dust and prevent it from falling into the warning range of tourists, thereby improving the intelligence, safety and sustainability of agricultural planting and tourist experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 The present invention is a module diagram of an aerial strawberry intelligent cultivation control system. DETAILED DESCRIPTION

[0047] The subject matter described herein will now be discussed with reference to example embodiments. It should be understood that the discussion of these embodiments is only to enable those skilled in the art to better understand and implement the subject matter described herein, and the functions and arrangements of the elements discussed may be changed without departing from the scope of protection of the contents of this specification. Each example may omit, replace or add various processes or components as needed. In addition, the features described relative to some examples may also be combined in other examples.

[0048] like Figure 1 As shown, an aerial strawberry intelligent cultivation control system includes:

[0049] An area division module is used to evenly divide the inner roof of the target planting shed into M sub-areas within a first preset time period, and hang an indicator ribbon on each sub-area;

[0050] The landing point acquisition module is used to perform test control in the target planting shed within a first preset time period to obtain the corresponding hanging image of the indicator ribbon and the landing point coordinates of the dust, specifically including:

[0051] T1: random control parameters of the speed, horizontal inclination and vertical inclination of the fan in the target planting shed;

[0052] T2: Randomly control the rise or fall of any hanging planting rack in the target planting shed;

[0053] T3: Obtain the hanging image of the indicator ribbon and the coordinates of the dust falling point corresponding to T1 and T2;

[0054] Repeat T1, T2 and T3 for a preset number of times to obtain several sets of hanging images and landing point coordinates;

[0055] A prediction module, used for training a landing point prediction model based on a plurality of sets of hanging images and landing point coordinates;

[0056] A control and adjustment module, used to obtain the coordinates of the target tourist in the target planting shed within a second preset time period, and to construct a warning range of the target tourist based on the coordinates;

[0057] Based on the falling point prediction model, by controlling the speed, horizontal inclination and vertical inclination of the fan in the target planting shed, the falling point coordinates of the dust are prevented from being within the warning range of the target tourists when the hanging planting racks are rising and falling.

[0058] Specifically, the purpose of raising and lowering the hanging planting rack is mainly to adjust the position of the planting rack to ensure that the growth environment of the strawberry plants is optimized. By raising and lowering the hanging planting rack, the height of the planting rack can be adjusted according to the growth stage of the strawberry, so as to ensure that the plant can grow under the best light, temperature and humidity conditions. This dynamic adjustment helps to improve the growth efficiency of strawberries. By raising and lowering the planting rack, the system can adjust the position of the planting rack according to the height and picking habits of the tourists, making the picking process more convenient and comfortable. This helps to enhance the experience of tourists, and also prevents physical discomfort to tourists due to improper positioning of the planting rack. In addition, for ripe strawberries, it is convenient for tourists to pick them by lowering. For unripe strawberries, they are placed in the air to grow. Since the ripening time of strawberries on a planting rack is not the same, it is necessary to raise and lower the hanging planting rack.

[0059] In one embodiment of the present invention, obtaining the coordinates of the dust landing point also includes:

[0060] Allocate indicator dust of different colors to each hanging planting rack in the target planting shed, and obtain a ground image of the target planting shed when the indicator dust rises or falls with the hanging planting rack;

[0061] Establishing a landing point coordinate system based on the ground image, the landing point coordinate system includes an X-axis and a Y-axis;

[0062] Based on the color of the indicating dust in the ground image, the coordinates of the landing point corresponding to the target planting rack are determined.

[0063] Specifically, each hanging planting rack is assigned a specific color of indicator dust. The main function of these dusts is to serve as a marker to help the system track the position changes of dust during the lifting and lowering of the planting rack. By distinguishing by color, dust can be avoided from being confused with other environmental factors, so as to accurately collect the coordinates of the landing point. When the hanging planting rack rises or falls, the system will collect images of the ground of the target planting shed. The indicator dust in the image will change as the planting rack moves, and the system analyzes the distribution pattern of dust through these changes. These images provide the actual location data of dust on the ground. Based on the collected ground images, the system will establish a coordinate system for recording and calibrating the specific landing points of dust. This coordinate system includes X-axis and Y-axis, which provides the spatial position of the dust landing point. This enables the system to accurately calculate the landing point of each indicator dust. The system determines the corresponding landing point coordinates by identifying different colors of dust in the ground image. Each color represents a different hanging planting rack. By analyzing the color, the system can accurately match each dust landing point with the corresponding planting rack. By using different colors of dust as markers, the system can more clearly distinguish the sources of dust from different planting racks, avoid confusion, and thus improve the accuracy of predicted landing points. Ground images provide a real-time view of dust distribution, and the system can adjust the control parameters of fans or hanging planting racks based on these images, and adjust the dust landing points in time to ensure the safety of visitors. Through the combination of color and coordinate system, the system can accurately locate the landing point of dust in space, facilitate more precise control of the adjustment of fans and planting racks, and improve the intelligent management of the cultivation environment.

[0064] In one embodiment of the present invention, the coordinates of the landing point corresponding to the target planting rack are determined based on the color of the dust indicating dust in the ground image, which can be achieved by image processing technology:

[0065] Use image processing algorithms (such as OpenCV) to process ground images, and use color recognition technology to extract different colors of dust on the ground. This technology relies on color space conversion in computer vision, converting the RGB color space into other color spaces (such as HSV), and distinguishing different colors of dust by hue, saturation, and brightness.

[0066] In one embodiment of the present invention, the speed, horizontal inclination and vertical inclination of the fan in the target planting shed are randomly controlled, including:

[0067] Determine the maximum and minimum speeds of the fans in the target growing shed;

[0068] Determine the horizontal allowable activity angle of the fan in the target planting shed;

[0069] Determine the vertical allowable motion angle of the fan in the target growing shed;

[0070] The maximum speed, minimum speed, horizontal allowable activity angle and vertical allowable activity angle of the fan are taken as constraints, and the speed, horizontal inclination angle and vertical inclination angle of the fan are randomly generated within the constraints.

[0071] Specifically, the speed and inclination of the fan are randomly controlled to simulate different airflow conditions in order to analyze the distribution and landing point of dust, thereby optimizing the environmental control during the aerial strawberry cultivation process. The process first needs to determine the operating range of the fan, including the maximum and minimum speeds. This is to ensure that the fan does not exceed its technical performance range when working and can provide an adjustable airflow intensity. These maximum and minimum values ​​provide boundary conditions for subsequent random adjustments. The fan can not only adjust the speed, but also change the direction of the airflow by changing its horizontal (left and right) and vertical (up and down) inclination. The system determines the horizontal and vertical allowable activity angles of the fan to ensure that the fan can be freely adjusted within these angles to meet different airflow directions and wind speed requirements. Randomly generate the speed, horizontal inclination and vertical inclination of the fan within the constraints: After knowing the activity range of the maximum speed, minimum speed, horizontal inclination and vertical inclination of the fan, the system will randomly generate a set of speed and inclination parameters. This means that each time the fan is controlled, the speed and inclination will randomly change within the set range, simulating different environmental conditions and helping the system analyze the impact of fan adjustment on dust landing points.

[0072] Suppose we have a planting shed with multiple fans to adjust the air flow to ensure that strawberries can grow in a suitable environment. During the experiment, the system hopes to analyze the impact of fan adjustment on dust fall points: the maximum speed of the fan is 1000rpm and the minimum speed is 200rpm. The fan can adjust the speed arbitrarily between 200rpm and 1000rpm to simulate different wind speed conditions. The range of horizontal tilt angle is -45° to 45°, that is, the wind direction of the fan can be offset to the left or right by up to 45 degrees. The range of vertical tilt angle is -30° to 30°, that is, the wind direction of the fan can be offset up or down by up to 30 degrees. The system will randomly generate a set of fan parameters, such as: fan speed: 500rpm (a random value between the minimum speed of 200rpm and the maximum speed of 1000rpm), horizontal tilt angle: 20° (a random value between -45° and 45°), vertical tilt angle: 10° (a random value between -30° and 30°).

[0073] In one embodiment of the present invention, a landing point prediction model is obtained based on a plurality of sets of hanging images and landing point coordinates, including:

[0074] Identify the M hanging images of the i-th group, and obtain the angle parameters of the indicator ribbon in each hanging image, the angle parameters include the degree of the angle between the indicator ribbon and the binding end of the indicator ribbon and the vertical line of the ground, and the direction of the angle; perform real number coding on the rise and fall of the hanging planting rack, and mark the state code corresponding to the i-th group of hanging images;

[0075] Establish a hanging image feature vector of group i, the feature vector includes:

[0076] The degree of the angle corresponding to a hanging image of the i-th group is taken as the first element of the feature vector, the direction of the angle is encoded by a real number as the second element of the feature vector, and the state encoding is taken as the third element of the feature vector;

[0077] The M eigenvectors are stacked to obtain a feature matrix, and the feature matrix is ​​normalized to obtain a standard matrix;

[0078] The standard matrix of the i-th group is used as the training sample of the landing prediction model, and the landing point coordinates of the i-th group are used as the sample labels of the landing prediction model; the landing prediction model is trained based on the back propagation optimization of the mean square error loss function;

[0079] The landing point prediction model includes an input layer, a hidden layer and a classification output layer.

[0080] Specifically, a drop prediction model is trained through machine learning to predict the drop coordinates of dust or other markers, thereby optimizing the environmental control during aerial strawberry cultivation.

[0081] Each group of hanging images is identified, and the system extracts the angle parameters of the indicator ribbon in each hanging image. The angle parameters of the indicator ribbon include:

[0082] The angle between the indicator ribbon and the binding end of the indicator ribbon and the vertical line of the ground: This angle indicates the angle at which the indicator ribbon deviates from the vertical line in the image.

[0083] Angle direction: Indicates the direction in which the streamer is offset from the vertical line (for example, left or right).

[0084] These angle parameters help the system understand the effect of airflow on hanging graphics, as well as the relationship between the raising and lowering of the grow rack and air flow.

[0085] The rising and falling status codes of the hanging planting rack:

[0086] The rising and falling actions of each hanging planter are encoded with real numbers, and each set of hanging images is labeled with a state code. For example, if a planter is rising, its state may be represented by a real number (such as +1), and when it is falling, it may be represented by another real number (such as -1).

[0087] These state encodings provide additional information about the dynamic behavior of the grow racks and can help the model determine where dust falls under different states.

[0088] Training the drop prediction model:

[0089] The normalized feature matrix is ​​used as training samples, and the drop point coordinates of each set of hanging images are used as sample labels. These samples and labels are used to train the drop point prediction model.

[0090] The model is optimized using the mean square error loss function, which aims to minimize the difference between the predicted value and the actual value. Through the back-propagation algorithm, the model continuously adjusts its parameters so that the predicted drop point coordinates are closer and closer to the actual drop point coordinates.

[0091] Model Architecture:

[0092] The trained landing point prediction model consists of three parts:

[0093] Input layer: accepts standard matrices as input.

[0094] Hidden layer: Processes the input information through a non-linear activation function and extracts the hidden state in the data.

[0095] Classification output layer: outputs the predicted landing point coordinates based on the learning results of the model.

[0096] In one embodiment of the present invention, the hidden layer and the classification output layer include:

[0097] The calculation formula of the hidden layer is as follows:

[0098] ;

[0099] in, represents the hidden layer The hidden state output by the hidden units, express Activation function, represents the total number of hidden units in the hidden layer, express The index of Indicates The weight matrix of hidden units, Indicates The bias matrix of hidden units, represents the standard matrix of input, represents the hidden layer The hidden state output by the hidden units, Indicates that right Perform convolution operation;

[0100] The classification output layer includes a classifier, which is used to input the The hidden state of the hidden unit output, and the classification space of the classifier represents the predicted landing point coordinates.

[0101] In one embodiment of the present invention, a warning range of a target tourist is constructed based on the coordinates, including:

[0102] Obtain the picking route of the target planting shed and the position of the target tourist, and associate the position of the target tourist with the corresponding position of the picking route; obtain the time for the hanging planting racks in the target planting shed to rise or fall, and the average moving speed of tourists in the historical time period; determine the warning moving distance of the target tourist based on the average moving speed and the time for the hanging planting racks to rise or fall; construct a characteristic circle with the corresponding position of the target tourist on the picking route as the center and the warning moving distance as the radius, and use the range of the characteristic circle as the warning range of the target tourist.

[0103] Specifically, the target tourists construct a warning range in the aerial strawberry cultivation environment. The purpose is to ensure the safety of tourists by predicting the range of tourists' movement and preventing dust or other objects from falling into the range of tourists due to the lifting of the hanging planting rack. First, the system obtains the picking route of the target planting shed, that is, the activity path of tourists in the shed. In addition, the location of the target tourists needs to be obtained, and the location of the tourists is associated with the location of the picking path where they are located through this information. In this way, the specific location of the tourists in the shed can be located and their possible activity range can be determined. The system also needs to understand the rising or falling time of the hanging planting rack, that is, the time required for the planting rack to pass during the lifting process. At the same time, the system will evaluate the average mobility of tourists during this period based on the average movement speed of tourists in the historical time period. Based on the average movement speed of tourists and the lifting time of the planting rack, the system can calculate the warning movement distance. The warning movement distance refers to the maximum distance that tourists may move during the process of the planting rack rising or falling. This distance can help the system predict the range of tourists who may enter the dangerous area. Finally, the system takes the corresponding position of the target tourists on the picking path as the center of the circle and uses the warning movement distance as the radius to construct a characteristic circle. The range of this characteristic circle is the warning range for target tourists.

[0104] In one embodiment of the present invention, preventing the dust from falling within the warning range of the target visitor when the hanging planting rack is raised and lowered includes:

[0105] Obtain the position and warning range of the target visitor at the zth moment in the second preset time period, and determine the hanging planting rack that needs to be raised or lowered;

[0106] Acquire the hanging image of the indicating ribbon corresponding to the z-th moment in the second preset time period, and obtain the predicted landing point coordinates corresponding to the hanging planting rack based on the landing point prediction model;

[0107] If the predicted landing point coordinates coincide with the warning range of the target tourist, the hanging planting rack will not be raised or lowered, and the first plan will be activated;

[0108] The first option includes:

[0109] E1, randomly adjust the speed, horizontal inclination and vertical inclination of the fan, obtain the corresponding hanging image, and determine the predicted landing point coordinates corresponding to the hanging image;

[0110] E2, repeat E1 until the predicted landing point coordinates and the warning range do not overlap, then start the second plan;

[0111] The second solution includes: controlling the corresponding hanging planting rack to rise or fall.

[0112] Specifically, a control method is provided to ensure that the lifting and lowering of the hanging planting rack will not cause dust pollution to the target tourists through real-time monitoring and adjustment. The method predicts the dust landing point and adjusts the operation of the fan and the planting rack to prevent dust from falling into the warning range of the tourists. Get the tourist location and warning range:

[0113] In the second preset time period, the system first needs to obtain the real-time location of the target tourist and the corresponding warning range. The warning range is determined based on the tourist's moving speed and possible activity range, with the purpose of predicting the area where the tourist may be exposed to dust. The system determines whether the position of the planting rack needs to be adjusted based on the tourist's location and warning range. If the lifting process of the planting rack may cause dust to fall into the tourist's warning range, the system will initiate subsequent control measures. The system obtains the hanging image of the indicator ribbon corresponding to the current time point, and predicts the landing coordinates of dust or markers through image recognition and landing prediction models. This prediction helps determine whether the dust will fall into the tourist's warning range. If the dust landing coordinates predicted by the system coincide with the tourist's warning range (that is, the dust will fall into the tourist's activity range), the system will not allow the planting rack to continue to rise and fall to avoid pollution. At this time, the system will start the first plan to adjust other parameters. The first plan (adjusting fan parameters): In the first plan, the system changes the airflow direction in the shed by randomly adjusting the fan speed, horizontal inclination and vertical inclination, thereby affecting the distribution of dust. After each adjustment, the system will obtain a new hanging image and predict the dust landing coordinates again. This process is repeated until the predicted dust fall point coordinates do not overlap with the visitor's warning range. At this point, the system ensures that the dust does not fall into the visitor's activity range. Second plan (adjusting the rise and fall of the planting rack): Once the first plan successfully adjusts the fan parameters and ensures that the dust no longer falls into the warning range, the system will start the second plan, which is to let the hanging planting rack continue to rise or fall for normal operation.

[0114] In one embodiment of the present invention, it is also included that the parameters of the fan need to be adjusted at intervals of a preset time, specifically:

[0115] During a test period, stop the fan and record the time. Wait until the indicator ribbon is completely vertical to the ground and then stop recording the time to get the first stop time. Repeat multiple times and take the average of the first recorded time to get the airflow static time.

[0116] In a test period, after the fan is stopped for a period longer than the airflow static time, the fan is started and the airflow stabilization time is obtained based on the following calculation formula:

[0117] ;

[0118] in, Indicates the airflow stabilization time, Indicates the volume of the target planting shed, Indicates the power of the fan. Indicates the density of the air in the target growing shed. Represents the empirical coefficient.

[0119] The fan is randomly adjusted according to the airflow static time and airflow stable time. For example, the first adjustment of the fan parameters will continue for the airflow stable time, and when the second adjustment is to be made, the fan will be turned off after the airflow is static for more than a period of time.

[0120] The above describes an embodiment of the present embodiment, but the present embodiment is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present embodiment, ordinary technicians in this field can also make many forms, all of which are within the protection of the present embodiment.

Claims

1. An intelligent aerial strawberry cultivation control system, characterized in that: include: An area division module is used to evenly divide the inner roof of the target planting shed into M sub-areas within a first preset time period, and hang an indicator ribbon on each sub-area; The landing point acquisition module is used to perform test control in the target planting shed within a first preset time period to obtain the corresponding hanging image of the indicator ribbon and the landing point coordinates of the dust, specifically including: T1: random control parameters of the speed, horizontal inclination and vertical inclination of the fan in the target planting shed; T2: Randomly control the rise or fall of any hanging planting rack in the target planting shed; T3: Obtain the hanging image of the indicator ribbon and the coordinates of the dust falling point corresponding to T1 and T2; Repeat T1, T2 and T3 for a preset number of times to obtain several sets of hanging images and landing point coordinates; A prediction module, used for training a landing point prediction model based on a plurality of sets of hanging images and landing point coordinates; A control and adjustment module, used to obtain the coordinates of the target tourist in the target planting shed within a second preset time period, and to construct a warning range of the target tourist based on the coordinates; Based on the falling point prediction model, by controlling the speed, horizontal inclination and vertical inclination of the fan in the target planting shed, the falling point coordinates of the dust are prevented from being within the warning range of the target tourists when the hanging planting racks are rising and falling.

2. A kind of aerial strawberry intelligent cultivation control system according to claim 1, characterized in that, Get the coordinates of the dust's landing point, including: Allocate indicator dust of different colors to each hanging planting rack in the target planting shed, and obtain a ground image of the target planting shed when the indicator dust rises or falls with the hanging planting rack; Establishing a landing point coordinate system based on the ground image, the landing point coordinate system includes an X-axis and a Y-axis; Based on the color of the indicating dust in the ground image, the coordinates of the landing point corresponding to the target planting rack are determined.

3. A kind of aerial strawberry intelligent cultivation control system according to claim 1, it is characterized in that, The speed, horizontal inclination and vertical inclination of the fan in the target planting shed are randomly controlled, including: Determine the maximum and minimum speeds of the fans in the target growing shed; Determine the horizontal allowable activity angle of the fan in the target planting shed; Determine the vertical allowable motion angle of the fan in the target growing shed; The maximum speed, minimum speed, horizontal allowable activity angle and vertical allowable activity angle of the fan are taken as constraints, and the speed, horizontal inclination angle and vertical inclination angle of the fan are randomly generated within the constraints.

4. A kind of aerial strawberry intelligent cultivation control system according to claim 1, characterized in that, The landing point prediction model is trained based on several sets of hanging images and landing point coordinates, including: Identify the M hanging images of the i-th group, and obtain the angle parameters of the indicator ribbon in each hanging image, the angle parameters include the degree of the angle between the indicator ribbon and the binding end of the indicator ribbon and the vertical line of the ground, and the direction of the angle; perform real number coding on the rise and fall of the hanging planting rack, and mark the state code corresponding to the i-th group of hanging images; Establish a hanging image feature vector of group i, the feature vector includes: The degree of the angle corresponding to a hanging image of the i-th group is taken as the first element of the feature vector, the direction of the angle is encoded by a real number as the second element of the feature vector, and the state encoding is taken as the third element of the feature vector; The M eigenvectors are stacked to obtain a feature matrix, and the feature matrix is ​​normalized to obtain a standard matrix; The standard matrix of the i-th group is used as the training sample of the landing prediction model, and the landing point coordinates of the i-th group are used as the sample labels of the landing prediction model; the landing prediction model is trained based on the back propagation optimization of the mean square error loss function; The landing point prediction model includes an input layer, a hidden layer and a classification output layer.

5. A kind of aerial strawberry intelligent cultivation control system according to claim 4, characterized in that, The hidden layers and classification output layer include: The calculation formula of the hidden layer is as follows: ； in, represents the hidden layer The hidden state output by the hidden units, express Activation function, represents the total number of hidden units in the hidden layer, express The index of Indicates The weight matrix of hidden units, Indicates The bias matrix of hidden units, represents the standard matrix of input, represents the hidden layer The hidden state output by the hidden units, Indicates that right Perform convolution operation; The classification output layer includes a classifier, which is used to input the The hidden state of the hidden unit output, and the classification space of the classifier represents the predicted landing point coordinates.

6. The intelligent cultivation control system of aerial strawberries according to claim 4 is characterized in that: And build the warning range of target tourists based on the coordinates, including: Obtain the picking route of the target planting shed and the position of the target tourist, and associate the position of the target tourist with the corresponding position of the picking route; obtain the time for the hanging planting racks in the target planting shed to rise or fall, and the average moving speed of tourists in the historical time period; determine the warning moving distance of the target tourist based on the average moving speed and the time for the hanging planting racks to rise or fall; construct a characteristic circle with the corresponding position of the target tourist on the picking route as the center and the warning moving distance as the radius, and use the range of the characteristic circle as the warning range of the target tourist.

7. An aerial strawberry intelligent cultivation control system according to claim 6, characterized in that: Avoid the dust drop coordinates when the hanging planting rack is raised and lowered to be within the warning range of the target visitors, including: Obtain the position and warning range of the target visitor at the zth moment in the second preset time period, and determine the hanging planting rack that needs to be raised or lowered; Acquire the hanging image of the indicating ribbon corresponding to the z-th moment in the second preset time period, and obtain the predicted landing point coordinates corresponding to the hanging planting rack based on the landing point prediction model; If the predicted landing point coordinates coincide with the warning range of the target tourist, the hanging planting rack will not be raised or lowered, and the first plan will be activated; The first option includes: E1, randomly adjust the speed, horizontal inclination and vertical inclination of the fan, obtain the corresponding hanging image, and determine the predicted landing point coordinates corresponding to the hanging image; E2, repeat E1 until the predicted landing point coordinates and the warning range do not overlap, then start the second plan; The second solution includes: controlling the corresponding hanging planting rack to rise or fall.

Images