An intelligent integrated seed planting machine, a control method and a seed germination state recognition method

Abstract

This invention discloses an intelligent integrated vegetable growing machine, comprising a machine shell, an environmental detection module, an intelligent control module, a first execution component, a seed germination rate identification module, and a second execution component. The machine shell includes a base, a support rod, a mounting plate, a truncated cone-shaped planting rack, and a nutrient chamber. The base has a first slot for mounting the nutrient chamber, which can be pulled out relative to the base. The truncated cone-shaped planting rack is mounted on the base, and the truncated cone-shaped planting rack and the nutrient chamber are connected through a recovery port on the base. The base is connected to the mounting plate via the support rod. The intelligent control module drives the first execution component to operate based on first environmental information received from the environmental detection module, and drives the second execution component to operate based on second environmental information received from the environmental detection module and germination rate information identified by the seed germination rate identification module. This invention can intelligently control the light intensity, temperature, humidity, and nutrient solution level of the vegetable growing machine.

Description

Technical Field

[0001] This invention relates to an intelligent integrated vegetable growing machine, a control method, and a seed germination status recognition method, belonging to the field of agricultural technology based on artificial intelligence. Background Technology

[0002] Soilless cultivation, also known as hydroponics, refers to the cultivation of plants without soil, using artificially prepared nutrient solutions to supply them with the mineral nutrients they need, thereby enabling the plants to grow and develop.

[0003] As people's demands for quality of life increase, hydroponics-based vegetable growing machines are gradually emerging. However, existing machines are not highly intelligent, only capable of timed and remote control. Furthermore, current machines lack research on detecting seed germination rates and controlling temperature using insulating films, failing to meet the irrigation and temperature requirements for vegetables to grow from seed to mature plants. Therefore, it is essential to invent an intelligent, integrated vegetable growing machine for home cultivation to address these issues. Summary of the Invention

[0004] This invention provides an intelligent integrated vegetable growing machine and its control method, which at least solves the problems mentioned in the prior art, such as the low level of intelligence of existing vegetable growing machines and their inability to identify seed germination rates.

[0005] The technical solution of this invention is:

[0006] According to a first aspect of the present invention, an intelligent integrated vegetable growing machine is provided, comprising a machine body shell 1, an environmental detection module, an intelligent control module, a first execution component, a seed germination rate identification module, and a second execution component; the machine body shell 1 includes a base 2, a support rod 3, a mounting plate 4, a frustum-shaped planting rack 5, and a nutrient chamber 8; the base 2 is provided with a first slot for mounting the nutrient chamber 8, which can be pulled out relative to the base 2; the base 2 is provided with a second slot for mounting the intelligent control module; the frustum-shaped planting rack 5 is mounted on the base 2, and the frustum-shaped planting rack 5 and the nutrient chamber 8 are connected... The base 2 is connected to the base plate 4 via a recycling port 9. The base 2 is connected to the mounting plate 4 via a support rod 3 arranged along the axial direction of the truncated cone planting frame 5, and there is a gap between the mounting plate 4 and the truncated cone planting frame 5. The environmental detection module is used to transmit the detected first environmental information and second environmental information to the intelligent control module. The intelligent control module is used to drive the first execution component to act according to the received first environmental information. The intelligent control module is used to drive the second execution component to act according to the received second environmental information and the germination rate information identified by the seed germination rate identification module.

[0007] Furthermore, a detachable horizontal bar 24 is provided below the mounting plate 4; the base 2 is provided with a third slot on the side near the first slot, which serves as a placement port 19 for the liquid level sensor.

[0008] Furthermore, the environmental detection module includes a temperature and humidity sensor 15, a light sensor 16, a liquid level sensor 28, an LCD display 18, and a temperature sensor. The temperature and humidity sensor 15 is installed above the base 2 and covered by the truncated cone-shaped planting rack 5. It is used to detect the temperature and humidity inside the truncated cone-shaped planting rack 5, using the internal temperature as the first environmental information and the humidity as the second environmental information. The temperature sensor is used to detect the outdoor temperature information, using the outdoor temperature as the first environmental information. The light sensor 16 is installed on the side of the horizontal rod 24 away from the center, used to detect the light intensity, using the light intensity as the first environmental information. The liquid level sensor 28 is installed inside the liquid level sensor placement port 19, used to detect the nutrient solution level in the nutrient tank 8. The LCD display 18 is installed on the outside of one side support rod 3, used to display the detected first and second environmental information, and to provide a prompt when the nutrient solution level is insufficient.

[0009] Further, the first execution component includes:

[0010] The light strip 7 is installed on the inner side of the two side support rods 3, the inner side of the support rods 3 being the side closest to the truncated cone-shaped planting frame 5, so that the intelligent control module can drive the light strip 7 to turn on / off according to the received first environmental information;

[0011] The arc-shaped track 23 is installed on the side of the mounting plate 4 near the truncated cone-shaped planting rack 5. The arc-shaped track 23 is provided with a sliding plate to connect the heat preservation film. The intelligent control module drives the motor module according to the received first environmental information, and the motor module drives the sliding plate to move, and the movement of the sliding plate drives the heat preservation film to open / close.

[0012] Further, the second actuating component includes a liquid guide tube 10, a drip tube 11, a first water pump, and a second water pump; the drip tube 11 is installed inside the truncated cone-shaped planting rack 5, arranged in a circular pattern around the interior of each truncated cone; the liquid guide tube 10 is installed inside the truncated cone-shaped planting rack 5, connecting the drip tubes 11 of the upper and lower layers; the drip tube 11 has small holes corresponding to the planting trough 6; the intelligent control module is used to drive the first water pump / second water pump to operate based on the received second environmental information and the germination rate information identified by the seed germination rate identification module: when When the first water pump is activated, it draws nutrient solution from the nutrient tank 8 at one end and connects the other end to the liquid guide pipe 10. The drawn nutrient solution is then transferred to the drip pipe 11 through the liquid guide pipe 10. After the drip pipe 11 is filled with nutrient solution, pressure forces the nutrient solution out from the small hole of the drip pipe 11 and drips it into the planting trough 6. When the second water pump is activated, one end of the second water pump is responsible for drawing nutrient solution from the nutrient tank 8 and the other end of the second water pump is connected to a nozzle. The nozzle is installed on the top of the inside of the truncated cone-shaped planting rack 5 and sprays water into the planting trough 6.

[0013] Furthermore, the seed germination rate identification module includes a camera 17 and a host computer. The camera 17 is installed on the side of the mounting plate 4 near the truncated cone-shaped planting rack 5. The camera 17 captures images from the truncated cone-shaped planting rack 5 and transmits them to the host computer. The host computer uses a small target detection algorithm based on the improved YOLOv5 to monitor the germination rate.

[0014] According to a second aspect of the present invention, a control method for an intelligent integrated vegetable growing machine is provided, the method comprising the following steps:

[0015] Humidity control: The camera 17 in the seed germination rate identification module takes pictures, and the host computer calls the improved YOLOv5 small target detection algorithm to identify the target based on the obtained image information, and calculates the seed germination rate based on the target identification result; when the temperature and humidity sensor 15 detects that the current humidity is not within the preset humidity range and identifies that the current seed germination rate is greater than or equal to the preset ratio value, the intelligent control module drives the second water pump in the second execution component to operate; when the temperature and humidity sensor 15 detects that the current humidity is not within the preset humidity range and identifies that the current seed germination rate is less than the preset ratio value, the intelligent control module drives the first water pump in the second execution component to operate.

[0016] Temperature control: When the external temperature sensor detects that the outside temperature is not within the preset range, the insulation film in the first actuator is closed. Then, the temperature and humidity sensor 15 detects the temperature inside the vegetable growing machine. If the detected temperature is lower than the preset range, the intelligent control module drives the motor module in the first actuator to control the insulation film to raise the temperature until it reaches the preset range, at which point the temperature increase stops. If the temperature detected by the temperature and humidity sensor 15 is higher than the preset range, the insulation film is controlled to lower the temperature until it reaches the preset range, at which point the temperature decreases. When the temperature detected by the external temperature sensor remains within the preset range for a preset duration, the motor is controlled to open the insulation film.

[0017] Light control: If the light sensor 16 detects and determines that the light intensity is lower than the preset range, the intelligent control module drives the light strip 7 to increase its brightness until the detected light intensity is within the preset range, at which point the brightness adjustment of the light strip 7 stops; if the light sensor 16 detects and determines that the light intensity is higher than the preset range, the intelligent control module drives the light strip 7 to decrease its brightness until the detected light intensity is within the preset range, at which point the brightness adjustment of the light strip 7 stops.

[0018] Liquid level control: If the liquid level sensor 28 detects that the nutrient solution level is lower than the preset position, the liquid level sensor 28 will display a prompt message indicating that the nutrient solution level is insufficient.

[0019] Furthermore, the improved YOLOv5-based small object detection algorithm uses a simplified version of the YOLOv5s model. The simplified YOLOv5s model uses the standard YOLOv5s model as its framework, and the backbone network of the standard YOLOv5s model is modified as follows:

[0020] Layer 2: Reduce the number of repetitions of the C3 module in Layer 2 of the backbone network in the standard model from 3 to 2.

[0021] Layer 4: The C3 module of Layer 4 in the backbone network of the standard model is repeated 6 times, but only 3 times.

[0022] Layer 6: The C3 module of layer 6 in the backbone network of the standard model is repeated 4 times instead of 9 times.

[0023] Layer 8: The C3 module of layer 7 in the backbone network of the standard model is repeated twice instead of three times.

[0024] According to a third aspect of the present invention, a method for identifying seed germination status is provided, which uses a small object detection algorithm based on an improved YOLOv5 to identify seed germination status; wherein, the small object detection algorithm based on the improved YOLOv5 uses a simplified version of the YOLOv5s model, which uses the standard YOLOv5s model as a framework, and modifies the backbone network of the standard YOLOv5s model as follows:

[0025] Layer 2: Reduce the number of repetitions of the C3 module in Layer 2 of the backbone network in the standard model from 3 to 2.

[0026] Layer 4: The C3 module of Layer 4 in the backbone network of the standard model is repeated 6 times, but only 3 times.

[0027] Layer 6: The C3 module of layer 6 in the backbone network of the standard model is repeated 4 times instead of 9 times.

[0028] Layer 8: The C3 module of layer 7 in the backbone network of the standard model is repeated twice instead of three times.

[0029] The beneficial effects of this invention are:

[0030] Compared to existing hydroponic vegetable growing machines, this invention can intelligently control the light intensity, temperature, humidity, and nutrient solution level of the growing machine, thereby reducing the cost of manual planting, freeing people's hands, and enabling self-cultivation. It can also use a camera to identify seed germination rate and monitor vegetable growth in real time, improving the intelligence level of home-use hydroponic vegetable growing machines and increasing the survival rate of vegetables grown by the machine. Furthermore, this invention allows users to grow vegetables independently while they are away at work, requiring almost no manual operation, saving users time and energy, and realizing remote home-based self-cultivation of vegetables. Attached Figure Description

[0031] Figure 1 This is a simplified block diagram of the intelligent vegetable planting machine described in an embodiment of the present invention;

[0032] Figure 2 This is a schematic diagram of the overall structure of the intelligent vegetable growing machine according to an embodiment of the present invention;

[0033] Figure 3 This is a schematic diagram of the structure of the frustum-shaped planting rack according to an embodiment of the present invention;

[0034] Figure 4 This is a top view (with cover plate) of the base structure described in an embodiment of the present invention;

[0035] Figure 5 This is a top view (without cover plate) of the base structure described in an embodiment of the present invention;

[0036] Figure 6This is a bottom view of the overall structure of the intelligent vegetable planting machine described in an embodiment of the present invention;

[0037] Figure 7 This is a schematic diagram of a partial structure of the mounting plate according to an embodiment of the present invention;

[0038] Figure 8 This is a schematic diagram (with crossbars) of the mounting plate structure described in an embodiment of the present invention;

[0039] Figure 9 This is a schematic diagram of the mounting plate structure described in an embodiment of the present invention (without crossbars);

[0040] Figure 10 This is a schematic diagram of the transverse bar structure described in an embodiment of the present invention;

[0041] Figure 11 This is a schematic diagram of the ventilation device described in an embodiment of the present invention;

[0042] Figure 12 This is a schematic diagram of the internal structure of the truncated cone-shaped planting rack according to an embodiment of the present invention;

[0043] Figure 13 This is a flowchart illustrating the training and testing process of the small object detection algorithm based on the improved YOLOv5 as described in an embodiment of the present invention.

[0044] Figure 14 The YOLOv5S architecture described in this embodiment of the invention;

[0045] Figure 15 This is a diagram of the MobileNetV3 network structure as described in an embodiment of the present invention;

[0046] Figure 16 This is a structural diagram of the C3 module according to an embodiment of the present invention;

[0047] Figure 17 This is a test diagram of the actual application of Simplify_yolov5s as described in the embodiments of the present invention;

[0048] The labels in the diagram are as follows: 1. Outer shell; 2. Base; 3. Support rod; 4. Mounting plate; 5. Conical planting rack; 6. Planting trough; 7. LED strip; 8. Nutrient tank; 9. Recovery port; 10. Liquid guide tube; 11. Drip tube; 12. Fixing screw; 13. Root inlet; 14. Ventilation device; 15. Temperature and humidity sensor; 16. Light sensor; 17. Camera; 18. LCD screen; 19. Liquid level sensor placement port; 20. Planting rack buckle; 21. Central control compartment; 22. Wheels; 23. Circular track; 24. Horizontal rod; 25. Fixing plate; 26. Nutrient tank cover; 27. Central control compartment cover; 28. Liquid level sensor. Detailed Implementation

[0049] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. It should be noted that, unless otherwise specified, the embodiments and features in the embodiments of this application can be arbitrarily combined with each other.

[0050] refer to Figure 1-17 As shown, according to a first aspect of the present invention, an intelligent integrated vegetable growing machine is provided, including a machine body shell 1, an environmental detection module, an intelligent control module, a first execution component, a seed germination rate identification module, and a second execution component; the machine body shell 1 includes a base 2, a support rod 3, a mounting plate 4, a truncated cone-shaped planting rack 5, and a nutrient chamber 8; the base 2 is provided with a first slot for mounting the nutrient chamber 8, which can be pulled out relative to the base 2; the base 2 is provided with a second slot serving as a central control compartment 21 for mounting the intelligent control module; the truncated cone-shaped planting rack 5 is mounted on the base 2. The nutrient bin 8 is connected to the base 2 via a recycling port 9. The base 2 is connected to the mounting plate 4 via a support rod 3 arranged along the axial direction of the truncated cone planting frame 5, and there is a gap between the mounting plate 4 and the truncated cone planting frame 5. The environmental detection module is used to transmit the detected first environmental information and second environmental information to the intelligent control module. The intelligent control module is used to drive the first execution component to act according to the received first environmental information. The intelligent control module is used to drive the second execution component to act according to the received second environmental information and the germination rate information identified by the seed germination rate identification module.

[0051] Furthermore, a detachable horizontal bar 24 is provided below the mounting plate 4; the base 2 is provided with a third slot on the side near the first slot, which serves as a placement port 19 for the liquid level sensor;

[0052] The environmental detection module includes a temperature and humidity sensor 15, a light sensor 16, a liquid level sensor 28, an LCD screen 18, and a temperature sensor. The temperature and humidity sensor 15 is installed above the base 2 and covered by the truncated cone-shaped planting rack 5. It is used to detect the temperature and humidity inside the truncated cone-shaped planting rack 5, using the internal temperature as the first environmental information and the humidity as the second environmental information. The temperature sensor is used to detect the outdoor temperature information, using the outdoor temperature as the first environmental information. It should be noted that the temperature sensor can use the temperature acquisition function of the temperature and humidity sensor, or it can be configured with a sensor that only has the temperature acquisition function. In this embodiment of the invention, the same model of sensor as the temperature and humidity sensor 15 is used.

[0053] The light sensor 16 is installed on the side of the horizontal bar 24 away from the center and is used to detect the light intensity, that is, the light intensity received by the planted vegetables, and use the light intensity as the first environmental information.

[0054] The liquid level sensor 28 is installed inside the liquid level sensor placement port 19 and is used to detect the liquid level of the nutrient solution in the nutrient tank 8.

[0055] The LCD screen 18 is mounted on the outside of one side support rod 3 and is used to display the detected first and second environmental information, to provide a prompt when the nutrient solution level is insufficient, and further to provide a prompt when the first and second environmental information are within the appropriate range.

[0056] Furthermore, the mounting plate 4 is connected to the transverse rod 24 via a fixing plate 25. The fixing plate 25 has a concave surface and a convex surface. The concave surface of the fixing plate 25 is a circular concave post located in the middle of the mounting plate 4, with a groove inside. The convex surface of the fixing plate 25 is an annular protrusion located in the middle of the transverse rod 24, with slots at both ends. Referring to the mortise and tenon fixing structure, the annular protrusion next to the slot in the annular protrusion on the convex surface of the fixing plate 25 is inserted into the groove in the circular concave post on the concave surface of the fixing plate 25. By rotating, the concave and convex surfaces of the fixing plate 25 are closed to connect the mounting plate 4 to the transverse rod 24. A light sensor 16 is installed on the outermost side of the transverse rod 24 to detect the light intensity of the intelligent vegetable growing machine.

[0057] For example, the base is cylindrical, and the first, second, and third slots are all fan-shaped groove areas of the base 2; the nutrient tank cover 26 and the central control tank cover 27 are parts cut out of the base 2, which can be removed separately and fixed to the base 2 with fixing screws 12. The nutrient tank cover 26 is used to cover the nutrient tank 8 and the liquid level sensor placement port 19, and the central control tank cover 27 is used to cover the central control tank 21. The nutrient tank cover 26 and the central control tank cover 27 separate the nutrient tank 8, the liquid level sensor placement port 19, the central control tank 21, and the truncated cone-shaped planting rack 5; the recovery port 9 is located in the middle of the base 2 and above the nutrient tank 8. It is a small fan-shaped groove that allows excess nutrient solution in the truncated cone-shaped planting rack 5 to flow into the nutrient tank 8 through the recovery port 9. Furthermore, the first and second slots are arranged adjacent to each other, and the first and third slots are arranged opposite to each other; the partition between the adjacent first and second slots is a thin plate, so that the liquid level sensor 28 installed in the liquid level sensor placement port 19 can detect the liquid level of the nutrient solution in the nutrient tank 8.

[0058] Furthermore, wheels 22 are installed at the bottom of the base 2, with a quarter-sector spacing between each pair of wheels 22, for a total of four wheels 22. The installation of wheels 22 facilitates the movement of the vegetable planting machine.

[0059] Furthermore, the mounting plate 4 is 50cm away from the truncated cone-shaped planting rack 5 to facilitate the installation of the light sensor 16, camera 17, circular track 23, etc.

[0060] Furthermore, the bottom sides of the truncated cone-shaped planting rack 5 are provided with planting rack buckles 20 for fitting the truncated cone-shaped planting rack 5 into the base 2 to fix the truncated cone-shaped planting rack 5. Each step surface of the truncated cone-shaped planting rack 5 is provided with planting grooves 6 at intervals. The planting grooves 6 are provided with many root openings 13, allowing the roots of vegetable plants to pass through the root openings 13 to reach the interior of the truncated cone-shaped planting rack 5 and receive water inside. The middle layer of the truncated cone-shaped planting rack 5 is provided with a ventilation device 14, which connects the interior and exterior of the truncated cone-shaped planting rack 5, allowing for air circulation between the inside and outside of the truncated cone-shaped planting rack 5 to ensure air renewal inside the truncated cone-shaped planting rack 5 and provide a certain amount of oxygen for the vegetable plant roots.

[0061] Further, the first execution component includes:

[0062] The light strip 7 is installed on the inner side of the two support rods 3, the inner side of the support rod 3 being the side closest to the truncated cone-shaped planting frame 5, so that the intelligent control module can drive the light strip 7 to open / close according to the received first environmental information, and in the open state, supplement the light intensity to achieve the light intensity required by the vegetables.

[0063] The arc-shaped track 23 is installed on the side of the mounting plate 4 near the truncated cone-shaped planting rack 5. The arc-shaped track 23 is provided with a sliding plate to connect the heat preservation film. The intelligent control module drives the motor module according to the received first environmental information, and the motor module drives the sliding plate to move, and the movement of the sliding plate drives the heat preservation film to open / close.

[0064] For example, such as Figure 1As shown, the base 2 is connected to the mounting plate 4 via two support rods 3 arranged axially along the frustum-shaped planting frame 5. One end of the support rod 3 is connected to the mounting plate 4 by a fastening screw, and the other end of the support rod 3 is connected to the base 2 by a screw. Two arc-shaped tracks 23 are installed on the side of the mounting plate 4 near the frustum-shaped planting frame 5. One arc-shaped track 23 is located on one side of the two support rods, and the other arc-shaped track 23 is located on the other side of the two support rods, so that the two circular tracks 23 are installed in a ring along the inner edge of the mounting plate 4. A rotating shaft is installed at one end of the two arc-shaped tracks 23, and a pressure sensor is installed at the other end of the two arc-shaped tracks 23, that is, there are two rotating shafts and two... Pressure sensors are used in the system. When powered on, the intelligent control module drives the motor module, which rotates a shaft with a thin thread wound around it and arranged along a track. A slider is mounted on the thread. Rotating the shaft clockwise releases the thread, causing the slider to follow the thread. An insulation film is mounted on the slider, which moves along the track to close the insulation film. Two pressure sensors, installed at the connection between the mounting plate 4 and the support rod 3, detect the pressure on the sliders on both sides of the track to determine if the sliders have reached their designated positions. If they do, the pressure sensors receive a pressure signal and command the motor module to stop rotating the shaft. Similarly, if the motor module drives the shaft counter-clockwise, the insulation film opens, controlling its opening and closing.

[0065] The intelligent control module drives the motor module based on the received first environmental information. Specifically, when the external temperature sensor detects that the ambient temperature is outside the set range, the insulation film is closed. Then, the temperature and humidity sensor 15 detects the temperature inside the vegetable growing machine. If the detected temperature is lower than the suitable temperature range, the drive motor module controls the insulation film to raise the temperature until it reaches the suitable range, at which point the temperature increase stops. If the temperature is higher than the suitable temperature range, the insulation film is controlled to lower the temperature until it reaches the suitable range, at which point the temperature decreases. When the temperature detected by the external temperature sensor remains within the set range for a preset duration, the motor is controlled to open the insulation film.

[0066] The intelligent control module drives the light strip 7 to turn on / off according to the received first environmental information. Specifically, if the light sensor 16 detects and determines that the light intensity is lower (higher) than the appropriate range of light intensity, it automatically controls the light strip 7 to increase (decrease) the brightness until the detected light intensity reaches the appropriate range, then stops the adjustment of the brightness of the light strip 7.

[0067] Furthermore, the seed germination rate identification module includes a camera 17 and a host computer. The camera 17 is installed on the mounting plate 4 near the truncated cone-shaped planting rack 5. The camera 17 captures images from the truncated cone-shaped planting rack 5 and transmits them to the host computer. The host computer uses a small target detection algorithm based on an improved YOLOv5 to monitor the germination rate. The camera 17 is installed at the center of the horizontal bar 24, and the planting status of seeds or seedlings in the planting trough 6 is observed through the camera 17.

[0068] Further, the second actuating component includes a liquid guide tube 10, a drip tube 11, a first water pump, and a second water pump; the drip tube 11 is installed inside the truncated cone-shaped planting rack 5, arranged in a circular pattern around the interior of each truncated cone; the liquid guide tube 10 is installed inside the truncated cone-shaped planting rack 5, connecting the upper and lower drip tubes 11; the drip tube 11 has small holes corresponding to the planting trough 6; the intelligent control module is used to drive the first water pump / second water pump to operate based on the received second environmental information and the germination rate information identified by the seed germination rate identification module: when the first water pump operates, water is drawn from the nutrient tank through one end of the first water pump. Nutrient solution is extracted from the nutrient tank 8. The other end of the first water pump is connected to the liquid guide pipe 10, which transmits the extracted nutrient solution to the drip pipe 11. After the drip pipe 11 is filled with nutrient solution, the nutrient solution is squeezed out from the small hole of the drip pipe 11 by pressure and dripped into the planting trough 6. A water-absorbing sponge is placed in the planting trough 6 to absorb the dripped nutrient solution and maintain a certain humidity. When the second water pump is activated, one end of the second water pump is responsible for extracting nutrient solution from the nutrient tank 8, and the other end of the second water pump is connected to a nozzle. The nozzle is installed on the top of the inside of the cone-shaped planting rack 5 to spray water the seedling roots in the planting trough 6.

[0069] Specifically, the camera 17 can be installed in the left groove of the horizontal bar 24 below the mounting plate 4 to monitor the growth status of vegetables and the germination rate of seeds in the planting trough 6 in real time. The planting trough 6 is set on each layer of the truncated cone-shaped planting rack 5 and is staggered according to the size of each layer of truncated cone. Vegetable seeds or seedlings can be placed in the planting trough 6 for planting. The root openings 13 are densely arranged on each planting trough 6 to ensure that the vegetable roots pass through the root openings 6 to reach the interior of the truncated cone-shaped planting rack 5 to absorb the sprayed nutrient solution. When the temperature and humidity sensor 15 detects that the current humidity is not within a suitable range, and the camera 17 identifies that the current seed germination rate is greater than 90%, the second water pump is used to extract nutrient solution from the nutrient tank 8. One end of the water pump is responsible for extracting nutrient solution from the nutrient tank 8, and the other end is connected to a nozzle. The nozzle is installed on the top inside the cone-shaped planting rack 5 to spray water onto the seedling roots in the planting trough 6. When the temperature and humidity sensor 15 detects that the current humidity is not within a suitable range, and the camera 17 identifies that the current seed germination rate is less than 90%, the first water pump is activated to extract nutrient solution from the nutrient tank 8 to achieve drip irrigation of the seeds.

[0070] According to a second aspect of the present invention, a control method for an intelligent integrated vegetable growing machine is provided, the method comprising the following steps:

[0071] Humidity control: The camera 17 in the seed germination rate identification module takes pictures, and the host computer calls the improved YOLOv5 small target detection algorithm to identify the target based on the obtained image information, and calculates the seed germination rate based on the target identification result; when the temperature and humidity sensor 15 detects that the current humidity is not within the preset humidity range and identifies that the current seed germination rate is greater than or equal to the preset ratio value, the intelligent control module drives the second water pump in the second execution component to operate; when the temperature and humidity sensor 15 detects that the current humidity is not within the preset humidity range and identifies that the current seed germination rate is less than the preset ratio value, the intelligent control module drives the first water pump in the second execution component to operate.

[0072] Temperature control: When the external temperature sensor detects that the outside temperature is not within the preset range, the insulation film in the first actuator is closed. Then, the temperature and humidity sensor 15 detects the temperature inside the vegetable growing machine. If the detected temperature is lower than the preset range, the intelligent control module drives the motor module in the first actuator to control the insulation film to raise the temperature until it reaches the preset range, at which point the temperature increase stops. If the temperature detected by the temperature and humidity sensor 15 is higher than the preset range, the insulation film is controlled to lower the temperature until it reaches the preset range, at which point the temperature decreases. When the temperature detected by the external temperature sensor remains within the preset range for a preset duration, the motor is controlled to open the insulation film.

[0073] Light control: If the light sensor 16 detects and determines that the light intensity is lower than the preset range, the intelligent control module drives the light strip 7 to increase its brightness until the detected light intensity is within the preset range, at which point the brightness adjustment of the light strip 7 stops; if the light sensor 16 detects and determines that the light intensity is higher than the preset range, the intelligent control module drives the light strip 7 to decrease its brightness until the detected light intensity is within the preset range, at which point the brightness adjustment of the light strip 7 stops.

[0074] Liquid level control: If the liquid level sensor 28 detects that the nutrient solution level is lower than the preset position, the liquid level sensor 28 will display a prompt message indicating that the nutrient solution level is insufficient.

[0075] The following is an example of how to use the intelligent integrated vegetable growing machine provided by the present invention:

[0076] The specific hardware specifications are as follows: The first and second water pumps in this invention are purchased USB watering pumps, with a 365 enlarged motor; the motor module driving the insulation film is an L298N motor drive board module; the pressure sensor is a KEYES resistive thin-film pressure sensor module; the liquid level sensor 28 is an XKC-Y25-PNP / V non-contact liquid level sensor; the light sensor is a TCRT5000 infrared reflection sensor; the temperature and humidity sensor is a DHT11 temperature and humidity sensor module; the intelligent control module uses three ESP-32-WROOM development boards; and the camera is an esp32_CAM remote monitoring camera. The usage steps are as follows:

[0077] S1. The user uses the four wheels 22 at the bottom of the vegetable growing machine to place the vegetable growing machine in a suitable location at home;

[0078] S2. Users sow vegetable seeds or place vegetable seedlings into the planting troughs 6 of each layer of the conical planting rack 5;

[0079] S3. The user needs to pour nutrient solution into the nutrient tank 8 so that the nutrient solution level reaches two-thirds of the nutrient tank 8.

[0080] S4. When the user turns on the home WIFI, after powering on the ESP32 chip in the central control compartment 21, it can connect to the WIFI on its own.

[0081] S5. Power on the ESP32 chip in the central control compartment 21 to power the light strip 7, temperature and humidity sensor 15, light sensor 16, camera 17, LCD screen 18, liquid level sensor 28, and temperature sensor on the device.

[0082] S6. The camera 17 monitors the plants planted in the planting trough 6 and determines the seed germination rate. When the temperature and humidity sensor 15 detects that the current humidity is not within the preset humidity range and identifies that the current seed germination rate is greater than or equal to 90%, it can be determined that the planted plants or most of the seeds have germinated. Therefore, the irrigation method is set to spray irrigation, i.e., the second water pump is activated. Conversely, if the current seed germination rate is identified as less than 90%, it is determined that the user has planted seeds and the germination rate is low. The irrigation method is set to drip irrigation, i.e., the first water pump is activated.

[0083] S7. When the temperature and humidity sensor 15, light sensor 16, liquid level sensor 28, and temperature sensor are powered, they begin to detect the temperature, humidity, nutrient solution level, external temperature, and light intensity near the vegetable growing machine.

[0084] S8. The detected data is transmitted to the ESP32 chip, which analyzes the data and determines whether it is within the appropriate range.

[0085] If the external temperature sensor detects that the outside temperature is outside the preset range, the insulation film in the first actuator (i.e., the insulation film surrounds the vegetable growing machine) is closed. Then, the temperature and humidity sensor 15 detects the temperature inside the vegetable growing machine. If the detected temperature is lower than the preset range, the intelligent control module drives the motor module in the first actuator to control the insulation film to raise the temperature until it reaches the preset range, at which point the temperature increase stops. If the temperature detected by the temperature and humidity sensor 15 is higher than the preset range, the insulation film is controlled to lower the temperature until it reaches the preset range, at which point the temperature decreases. When the temperature detected by the external temperature sensor remains within the preset range for more than 10 seconds, the motor is controlled to open the insulation film.

[0086] If the light sensor 16 detects and determines that the light intensity is lower than the preset range, the intelligent control module drives the light strip 7 to increase its brightness until the detected light intensity is within the preset range, at which point the brightness adjustment of the light strip 7 stops. If the light sensor 16 detects and determines that the light intensity is higher than the preset range, the intelligent control module drives the light strip 7 to decrease its brightness until the detected light intensity is within the preset range, at which point the brightness adjustment of the light strip 7 stops.

[0087] If the liquid level sensor 28 detects that the nutrient solution level is less than one-quarter of the container's capacity, it will display a message on the LCD screen 18 that reads "Insufficient nutrient solution, please replenish nutrient solution in time".

[0088] As described above, the LCD screen 18 displays the current temperature, humidity, light intensity, and nutrient solution level. Regarding the nutrient solution level, when the sensor detects a level less than one-quarter full, the LCD screen 18 will display a message: "Nutrient solution insufficient, please replenish nutrient solution promptly." For temperature, humidity, and light intensity, if the values ​​detected by the temperature and humidity sensor 15 and the light sensor 16 are below the appropriate range, "Too low, adjusting..." will be displayed after the corresponding value; if the values ​​detected by the temperature and humidity sensor 15 and the light sensor 16 are above the appropriate range, "Too high, adjusting..." will be displayed after the corresponding value; and if the values ​​detected by the temperature and humidity sensor 15 and the light sensor 16 are within the appropriate range, "Normal" will be displayed after the corresponding value.

[0089] According to a third aspect of the present invention, a method for identifying seed germination status is provided, which uses a small object detection algorithm based on an improved YOLOv5 to identify seed germination status; wherein, the small object detection algorithm based on the improved YOLOv5 uses a simplified version of the YOLOv5s model, which uses the standard YOLOv5s model as a framework, and modifies the backbone network of the standard YOLOv5s model as follows:

[0090] Layer 2: Reduce the number of repetitions of the C3 module in Layer 2 of the backbone network in the standard model from 3 to 2.

[0091] Layer 4: The C3 module of Layer 4 in the backbone network of the standard model is repeated 6 times, but only 3 times.

[0092] Layer 6: The C3 module of layer 6 in the backbone network of the standard model is repeated 4 times instead of 9 times.

[0093] Layer 8: The C3 module of layer 7 in the backbone network of the standard model is repeated twice instead of three times.

[0094] The small object detection algorithm based on the improved YOLOv5 needs to be trained before it can be used. The specific training process is given below:

[0095] Step 1: Download the public dataset of seeds and seedlings from the Kaggle platform. Use the labeling software LabelImg to annotate the dataset in YOLO format. The seedling images in the dataset are labeled "alive", and the ungerminated holes are labeled "dormant".

[0096] Step 2: For each image labeled in Step 1, create a corresponding .txt file with the same name as the image. Each line in the .txt file represents a labeled instance, with 5 columns in total. From left to right, these columns represent the label category, the ratio of the horizontal coordinate of the label box center to the image width, the ratio of the vertical coordinate of the label box center to the image height, the ratio of the label box width to the image width, and the ratio of the label box height to the image height. The final result is a training set of 800 images and a test set of 300 images.

[0097] Step 3: Apply Mosaic-9 data augmentation to the collected vegetable seedling dataset. This involves randomly cropping, arranging, and scaling nine images, then combining them into a single image, while also introducing some random noise. The dataset suffers from overly simplistic image formats and backgrounds, with the target objects occupying too large a portion of the image space. Furthermore, the images captured by the camera on the vegetable growing machine are mostly top-down views, resulting in smaller targets. Mosaic-9 data augmentation increases data diversity by stitching multiple images together into a single image.

[0098] Furthermore, the specific steps of Mosaic-9 data augmentation in Step 3 are as follows:

[0099] 1) Assume the model input size parameter is s, and generate a grayscale image with a size of 3s*3s;

[0100] 2) Randomly select a point C as the splicing center within the rectangle defined by points A(s / 2, s / 2) and B(5s / 2, 5s / 2). c y c );

[0101] x c =s / 2+(5s / 2-s / 2)·rand()

[0102] y c =s / 2+(5s / 2-s / 2)·rand()

[0103] 3) Randomly select 9 images and stitch parts of them into this image. The 9 colors represent the 9 sample images. Any parts that are not included will be discarded.

[0104] 4) After transforming the images, the final output is obtained, which is the dataset augmented with Mosaic-9. The final augmented training set consists of 800 images, and the augmented test set consists of 300 images.

[0105] Step 4: To improve detection speed while maintaining high recognition accuracy, the model was streamlined and optimized. This optimization aims to reduce computational complexity and inference time while ensuring the model's feature extraction capabilities, achieving the optimal balance between detection accuracy and speed. The specific optimization scheme is as follows:

[0106] Option 1: MobileNet_yolov5s:

[0107] The original YOLOv5 backbone is replaced with MobileNetV3. This module is a lightweight neural network architecture that achieves high feature extraction capabilities with lower computational cost. Replacing it with the YOLOv5 backbone effectively reduces the model's computational complexity and inference time.

[0108] The complex convolutional layers and C3 module of the original YOLOv5 backbone network have been replaced by MobileNetV3 modules, forming a lightweight architecture. This not only accelerates computation but also reduces memory footprint and resource consumption, making the model more suitable for the embedded camera devices in this project and significantly improving inference speed.

[0109] The backbone network has been modified as follows:

[0110] The first layer is MobileNetV3 with parameters [24, 1]. As the initial feature extraction layer of the network, it is responsible for capturing basic feature information in the image. '24' represents the number of output channels, and '1' indicates that one convolution operation is used. The output feature map size is 80×80, and the stride is 3 / 8 to capture more detailed information in the image.

[0111] The second layer is MobileNetV3 with parameters [48, 2]. The number of channels in the second layer is increased to '48', and two convolutional operations are performed to extract higher-level features. The output feature map size is 20×20, with a stride of 4 / 16. Compared to the first layer, the feature map size is significantly smaller, but it has stronger abstraction capabilities, capturing more important object information in the image.

[0112] The third layer is MobileNetV3 with parameters [576, 3]. The number of channels in the third layer is increased to '576', and three convolutional operations are performed to further enhance deep feature extraction. The output feature map size remains at 20×20, with a stride of 5 / 32. This layer is responsible for integrating high-level image features so that subsequent layers can effectively identify complex objects and background information.

[0113] Option 2: Simplify_yolov5s:

[0114] In the original YOLOv5 architecture, the C3 module is used for deep feature extraction and achieves efficient feature learning through multiple repetitions, thereby enhancing the model's recognition capabilities. However, this repetitive layer design also leads to high computational costs, affecting the model's inference speed.

[0115] In this approach, the number of repetitions of the C3 module is reduced to decrease computational cost, and the parameter settings of the convolutional layers are adjusted to further simplify the network. This retains the core feature extraction capabilities of the YOLOv5 model while reducing its complexity, thus achieving a more reasonable balance between detection speed and accuracy. The detailed simplification is as follows:

[0116] Layer 2: The original network used the C3 module and repeated it 3 times. The simplified version reduces the number of repetitions to 2 to reduce computational cost.

[0117] Layer 4: In the original network, the C3 module is repeated 6 times. The simplified version reduces it to 3 times, also in order to reduce computational cost.

[0118] Layer 6: In the original network, the C3 module was repeated 9 times. The simplified version reduces it to 4 times, which significantly reduces the computational cost.

[0119] Layer 8: The C3 module of the original network is repeated 3 times, and the simplified version further reduces it to 2 times.

[0120] After adjustments, the model retains its core feature extraction capabilities while significantly simplifying its structure to achieve a good balance between accuracy and detection speed.

[0121] Step 5: Feed the enhanced training set into different networks for iterative training, repeat the loop 100 times to obtain the optimal training model, and use the test set to test the training model.

[0122] Step 6: To verify the effectiveness of the algorithm of this invention, the improved algorithms MobileNet_yolov5s and Simplify_yolov5s were tested on the same test set as the original yolov5s. The results of the performance comparison are shown in Table 1. As can be seen from Table 1, yolov5s, as the standard model, performs well in recall and accuracy, and also has the lowest bounding box loss and confidence loss, but its processing speed is relatively slow. MobileNet_yolov5s significantly improves processing speed by adopting the MobileNetV3 architecture, achieving more than three times the FPS of yolov5s, but its accuracy decreases. Simplify_yolov5s, on the other hand, achieves a good balance between accuracy and speed, approaching the accuracy of yolov5s while maintaining high FPS, making it suitable for vegetable germination rate monitoring in this project, which balances accuracy and real-time performance. Therefore, this model was selected as the final detection model.

[0123] Table 1 Comparison of Model Parameters

[0124] Model cycle FPS Recall rate accuracy Border loss Confidence loss yolov5s 100 20.034 0.92654 0.94243 0.022031 0.0096117 MobileNet_yolov5s 100 69.262 0.90558 0.90063 0.024510 0.0100680 Simplify_yolov5s 100 68.897 0.90873 0.94224 0.022664 0.0094546

[0125] Step 7: Validate the model by detecting germinating and non-germinating targets under different lighting, background, and angle conditions; from Figure 17 As can be seen, both sprouted (0) and non-sprouted (1) categories were effectively detected in the image. All sprouted and non-sprouted targets were bounded, and the labels of different categories were correctly distinguished. The ratio of the number of sprouted targets to the total number of targets was taken as the sprouting rate, and the total number of targets was "sprouted targets + non-sprouted targets". During the training process, the loss functions and performance indicators of the model were continuously optimized, and the performance on the test set also showed that the model has good generalization ability and robustness. This indicates that the model has good detection performance in various detection tasks and meets the actual requirements of target detection in this project.

[0126] The specific embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited to the above embodiments. Within the scope of knowledge possessed by those skilled in the art, various changes can be made without departing from the spirit of the present invention.

Claims

1. An intelligent integrated vegetable growing machine, characterized in that, The system includes a casing (1), an environmental detection module, an intelligent control module, a first execution component, a seed germination rate identification module, and a second execution component. The casing (1) includes a base (2), a support rod (3), a mounting plate (4), a frustum-shaped planting rack (5), and a nutrient chamber (8). The base (2) has a first slot for mounting the nutrient chamber (8) which moves relative to the base (2). The base (2) also has a second slot for mounting the intelligent control module. The frustum-shaped planting rack (5) is mounted on the base (2), and the frustum-shaped planting rack (5) and the nutrient chamber (8) are connected by the base. (2) The recycling ports (9) on the top are arranged in a connected manner; the base (2) is connected to the mounting plate (4) by a support rod (3) arranged along the axial direction of the truncated cone planting frame (5), and there is a gap between the mounting plate (4) and the truncated cone planting frame (5); the environmental detection module is used to transmit the detected first environmental information and second environmental information to the intelligent control module, the intelligent control module is used to drive the first execution component to act according to the received first environmental information, and the intelligent control module is used to drive the second execution component to act according to the received second environmental information and the germination rate information identified by the seed germination rate identification module.

2. The intelligent integrated vegetable growing machine according to claim 1, characterized in that, The mounting plate (4) is provided with a detachable horizontal bar (24) below it; the base (2) is provided with a third slot on the side near the first slot as a liquid level sensor placement port (19).

3. The intelligent integrated vegetable growing machine according to claim 2, characterized in that, The environmental detection module includes a temperature and humidity sensor (15), a light sensor (16), a liquid level sensor (28), a liquid crystal display (18), and a temperature sensor. The temperature and humidity sensor (15) is installed above the base (2) and covered by the truncated cone-shaped planting rack (5). It is used to detect the temperature and humidity inside the truncated cone-shaped planting rack (5), and uses the internal temperature as the first environmental information and the humidity as the second environmental information. The temperature sensor is used to detect the outdoor temperature information and uses the outdoor temperature as the first environmental information. The light sensor (16) is installed on the side of the horizontal bar (24) away from the center, and is used to detect the light intensity and use the light intensity as the first environmental information; The liquid level sensor (28) is installed in the liquid level sensor placement port (19) and is used to detect the liquid level of the nutrient solution in the nutrient tank (8); The liquid crystal display screen (18) is installed on the outside of one side support rod (3) to display the first and second environmental information detected, and to provide a prompt when the nutrient solution level is insufficient.

4. The intelligent integrated vegetable growing machine according to claim 1, characterized in that, The first execution component includes: The light strip (7) is installed on the inner side of the two side support rods (3), the inner side of the support rods (3) being the side close to the truncated cone-shaped planting frame (5), so that the intelligent control module can drive the light strip (7) to open / close according to the received first environmental information; The arc-shaped track (23) is installed on the side of the mounting plate (4) near the truncated cone-shaped planting rack (5). The arc-shaped track (23) is provided with a sliding plate to connect the heat preservation film, so that the intelligent control module can drive the motor module according to the first environmental information received, drive the sliding plate to move through the motor module, and drive the heat preservation film to open / close through the movement of the sliding plate.

5. The intelligent integrated vegetable growing machine according to claim 1, characterized in that, The second actuator includes a liquid guide tube (10), a drip tube (11), a first water pump, and a second water pump; the drip tube (11) is installed inside the truncated cone-shaped planting rack (5), and surrounds the inside of each layer of the truncated cone in a circular pattern; the liquid guide tube (10) is installed inside the truncated cone-shaped planting rack (5) and connects the drip tubes (11) of the upper and lower layers; the drip tube (11) is provided with small holes, which correspond to the planting trough (6); the intelligent control module is used to drive the first water pump / second water pump to operate according to the received second environmental information and the germination rate information identified by the seed germination rate identification module: when the first water pump... When the pump is activated, the first pump draws nutrient solution from the nutrient tank (8) at one end and connects the other end of the first pump to the liquid guide pipe (10). The drawn nutrient solution is then transferred to the drip pipe (11) through the liquid guide pipe (10). After the drip pipe (11) is filled with nutrient solution, the nutrient solution is squeezed out from the small hole of the drip pipe (11) by pressure and dripped into the planting trough (6) in a drip irrigation manner. When the second pump is activated, one end of the second pump is responsible for drawing nutrient solution from the nutrient tank (8) and the other end of the second pump is connected to the nozzle. The nozzle is installed on the top of the inside of the truncated cone planting rack (5) to spray water into the planting trough (6).

6. The intelligent integrated vegetable growing machine according to claim 1, characterized in that, The seed germination rate identification module includes a camera (17) and a host computer. The camera (17) is installed on the mounting plate (4) near the side of the truncated cone-shaped planting rack (5). The camera (17) captures images from the truncated cone-shaped planting rack (5) and transmits them to the host computer. The host computer uses a small target detection algorithm based on the improved YOLOv5 to monitor the germination rate.

7. A control method for an intelligent integrated vegetable growing machine, characterized in that, The method steps are as follows: Humidity control: The camera (17) in the seed germination rate identification module takes pictures, and the host computer calls the small target detection algorithm based on the improved YOLOv5 to identify the target based on the obtained image information. The seed germination rate is calculated based on the target identification result. When the temperature and humidity sensor (15) detects that the current humidity is not within the preset humidity range and identifies that the current seed germination rate is greater than or equal to the preset ratio value, the second water pump in the second execution component is driven by the intelligent control module. When the temperature and humidity sensor (15) detects that the current humidity is not within the preset humidity range and identifies that the current seed germination rate is less than the preset ratio value, the first water pump in the second execution component is driven to operate through the intelligent control module. Temperature control: When the external temperature sensor detects that the outside temperature is not within the preset range of the outside temperature, the insulation film in the first execution component is closed. Then, the temperature inside the vegetable growing machine is detected by the temperature and humidity sensor (15). If the detected temperature is lower than the preset range of the internal temperature, the motor module in the first execution component is driven by the intelligent control module to control the insulation film to raise the temperature until it reaches the preset range of the internal temperature and then stops raising the temperature. If the temperature detected by the temperature and humidity sensor (15) is higher than the preset range of the internal temperature, the insulation film is controlled to lower the temperature until it reaches the preset range of the internal temperature and then stops lowering the temperature. When the temperature detected by the external temperature sensor is continuously within the preset range of the outside temperature and reaches the preset time, the motor is controlled to open the insulation film. Light control: If the light sensor (16) detects and determines that the light intensity is lower than the preset range of light intensity, the intelligent control module drives the light strip (7) to increase the brightness until the detected light intensity is within the preset range of light intensity, then the adjustment of the brightness of the light strip (7) is stopped; if the light sensor (16) detects and determines that the light intensity is higher than the preset range of light intensity, the intelligent control module drives the light strip (7) to decrease the brightness until the detected light intensity is within the preset range of light intensity, then the adjustment of the brightness of the light strip (7) is stopped. Liquid level control: If the liquid level sensor (28) detects that the nutrient solution level is lower than the preset position, the liquid level sensor (28) will display a prompt message that the nutrient solution level is insufficient.

8. The control method for the intelligent integrated vegetable growing machine according to claim 7, characterized in that, The improved YOLOv5-based small object detection algorithm uses a simplified version of the YOLOv5s model. The simplified YOLOv5s model uses the standard YOLOv5s model as its framework, and the backbone network of the standard YOLOv5s model is modified as follows: Layer 2: Reduce the number of repetitions of the C3 module in Layer 2 of the backbone network in the standard model from 3 to 2. Layer 4: The C3 module of Layer 4 in the backbone network of the standard model is repeated 6 times, but only 3 times. Layer 6: The C3 module of layer 6 in the backbone network of the standard model is repeated 4 times instead of 9 times. Layer 8: The C3 module of layer 7 in the backbone network of the standard model is repeated twice instead of three times.

9. A method for identifying seed germination status, characterized in that, A small object detection algorithm based on an improved YOLOv5 is used to identify seed germination status. This improved YOLOv5 algorithm employs a simplified version of the YOLOv5s model. The simplified YOLOv5s model uses the standard YOLOv5s model as its framework, with the following modifications to the backbone network: Layer 2: Reduce the number of repetitions of the C3 module in Layer 2 of the backbone network in the standard model from 3 to 2. Layer 4: The C3 module of Layer 4 in the backbone network of the standard model is repeated 6 times, but only 3 times. Layer 6: The C3 module of layer 6 in the backbone network of the standard model is repeated 4 times instead of 9 times. Layer 8: The C3 module of layer 7 in the backbone network of the standard model is repeated twice instead of three times.

Images