Plant adaptive spectral regulation system based on image recognition
The plant adaptive spectral adjustment system, which combines image recognition and environmental spectral sensors, can identify plant status in real time and dynamically adjust the spectrum, solving the problem of insufficient spectral control precision and achieving efficient light energy utilization and optimized plant growth.
Patent Information
- Application Number
- CN202522300228.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Utility models(China)
- Current Assignee / Owner
- Filing Date
- 2025-10-30
- Publication Date
- 2026-11-10
- Estimated Expiration
- 2035-10-30
AI Technical Summary
Existing technologies cannot achieve high-resolution, full-band spectral power distribution data acquisition, resulting in limited spectral control precision, failing to meet the dynamic spectral requirements of plants, leading to energy waste and plant growth problems.
A plant adaptive spectral adjustment system based on image recognition is adopted. It uses a high-definition RGB camera and a Raspberry Pi microcomputer combined with an environmental spectral sensor to identify plant varieties and growth stages in real time. Dynamic spectral adjustment is performed through a multi-channel LED light source to form a closed-loop control.
It achieves high-precision spectral matching, eliminates the effects of LED light source attenuation and ambient light, ensures that plants receive the most suitable light, and improves plant growth efficiency and light energy utilization.
Smart Images

Figure CN224959791U_ABST
Abstract
Description
Technical Field
[0001] This utility model relates to the field of plant growth control technology, and in particular to a plant adaptive spectral adjustment system based on image recognition. Background Technology
[0002] Plant factories, through precise control of environmental factors such as temperature, light, water, air, and fertilizer, break the dependence of traditional agriculture on natural conditions and achieve year-round, high-efficiency crop production. Among these, light is a core environmental factor affecting plant photosynthesis, morphogenesis, synthesis of secondary metabolites, and ultimately, yield and quality. Different plant varieties, and even the same variety at different growth stages (such as seedling stage, vegetative growth stage, flowering stage, and fruit ripening stage), exhibit significant differences in their photosynthetic system's requirements for spectral energy distribution (i.e., the spectrum). For example, blue light (400-500nm) typically helps control plant shape and promote leaf thickening; while red light (600-700nm) is crucial for promoting stem elongation, flowering, and fruit ripening.
[0003] Traditional and widespread fixed-spectrum lighting completely ignores the dynamic spectral requirements of plants. Growing under suboptimal spectra leads to energy waste (e.g., providing excessive blue light during periods when red light is needed) and can cause problems such as excessive vegetative growth, insufficient biomass accumulation, disordered flowering, and decreased fruit quality, failing to achieve the "on-demand lighting" and maximum efficiency sought by precision agriculture. In recent years, some studies have attempted to introduce sensor technologies, such as using colorimetric sensors or simple light intensity sensors to monitor the light environment. These sensors typically only acquire light intensity information or integrated colorimetric values within a limited wavelength band, and cannot provide high-resolution, full-band spectral power distribution data. Therefore, the control precision is limited, making it difficult to achieve truly precise spectral matching. With the development of artificial intelligence technology, especially the maturity of deep learning in image classification and object detection, the academic community has begun to explore the use of computer vision for plant type information extraction (such as disease identification and maturity assessment). However, most existing technologies remain at the identification and monitoring stage, failing to form a deep closed-loop integration with light environment control systems. Specifically, there is a lack of a complete hardware system solution that can combine visually perceived plant physiological states (variety, growth stage, health) with high-precision spectral sensing feedback, and thereby drive multi-channel LED light sources for real-time, dynamic, and closed-loop control. Utility Model Content
[0004] The purpose of this invention is to solve the problems in the prior art.
[0005] The technical solution adopted by this utility model to solve its technical problem is: to provide a plant adaptive spectral adjustment system based on image recognition, comprising:
[0006] The image recognition module is used to acquire and recognize plant images, and the recognition results include plant variety and growth stage.
[0007] The main controller module, connected to the image recognition module, is used to receive the recognition results and output the spectral adjustment signal;
[0008] The spectral adjustment module includes a driving circuit and an LED light board. The driving circuit is connected to the main controller module to receive the spectral adjustment signal and outputs a driving signal to the LED light board based on the spectral adjustment signal. The LED light board includes a first LED chip, a second LED chip, and a third LED chip, which emit three different types of light respectively. The three different types of light include red light, blue light, and broadband white light.
[0009] Preferably, the image recognition module includes an acquisition device and a recognition device;
[0010] The acquisition device uses a high-definition RGB camera, which is fixedly installed above the plant canopy, directly facing the center of the plant, to ensure that the complete morphological features of the plant are captured; the output of the high-definition RGB camera is connected to the input of the recognition device.
[0011] The identification device uses a trained plant identification model, and the output of the plant identification model is connected to the main controller module.
[0012] Preferably, the driving circuit of the spectrum adjustment module is a PWM constant current driving circuit, which includes three circuit channels, namely:
[0013] The first channel has its input end connected to the main controller and its output end connected to the first LED chip of the LED light board.
[0014] The second channel has its input end connected to the main controller and its output end connected to the second LED chip of the LED light board.
[0015] The third channel has its input end connected to the main controller and its output end connected to the third LED chip on the LED light board.
[0016] Preferably, the three types of LED chips on the LED light board are arranged in a uniform and staggered manner to ensure that the spectrum irradiated onto the plant canopy is uniformly mixed.
[0017] Preferred options also include:
[0018] An environmental spectral sensing module is used to collect real-time spectral data of the environment surrounding the plant; the output of the environmental spectral sensing module is connected to the main controller module.
[0019] Preferably, the environmental spectral sensing module uses an integrated digital spectral sensor AS7341, which is installed at the middle height of the plant canopy with the probe facing upwards, to accurately measure the actual light spectrum and intensity received by the plant leaves.
[0020] Preferably, the main controller module uses a Raspberry Pi 4B microcomputer, and its interface connection method is as follows:
[0021] Connect the power interface to a 5V power source.
[0022] The USB-A interface connects to the output interface of the identification device;
[0023] The 3.3V VCC interface connects to the power supply pin of the integrated digital spectral sensor AS7341;
[0024] The GPIO2 interface connects to the I / O pin of the AS7341 integrated digital spectral sensor. 2 C data cable pin;
[0025] The GPIO3 interface connects to the I / O pin of the AS7341 integrated digital spectral sensor. 2 C clock line pin;
[0026] The GPIO12 interface is connected to the first channel input of the driver circuit;
[0027] The GPIO13 interface is connected to the second channel input of the driver circuit;
[0028] The GPIO18 interface is connected to the third channel input of the driver circuit.
[0029] This utility model has the following beneficial effects:
[0030] (1) High integration and simple structure: The Raspberry Pi is used as a single control core to uniformly process image recognition, spectral sensing and light source driving. The system structure is compact and the hardware cost is low.
[0031] (2) Closed-loop control and precise regulation: The environmental spectrum is fed back in real time by the spectral sensor to form a closed-loop control, which eliminates the effects of LED light source decay and environmental light interference, and ensures that the plants always receive the most suitable light.
[0032] (3) Active identification and intelligent adaptation: The system actively identifies plant varieties and growth stages using cameras without the need for manual preset. It can adapt to different crops and growth cycles, and has a high degree of intelligence and strong practicality.
[0033] (4) Easy to expand: Raspberry Pi has a rich software ecosystem and interfaces, making it easy to add temperature and humidity sensors, communication modules, etc. to achieve functional expansion.
[0034] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments, but the present invention is not limited to the embodiments. Attached Figure Description
[0035] Figure 1 This is a system structure diagram of an embodiment of the present invention;
[0036] Figure 2 This is a system structure diagram of an embodiment of the present utility model;
[0037] Figure 3 This is a connection diagram of the main control module in an embodiment of the present utility model;
[0038] Figure 4 This is a schematic diagram showing the connection between the main control module and the spectral adjustment module in an embodiment of the present invention. Detailed Implementation
[0039] See Figure 1 The diagram shown is a system structure diagram of this utility model, including a main controller module 101, an image recognition module 102, an environmental spectrum sensing module 103, and a spectrum adjustment module 104; wherein, the image recognition module and the environmental spectrum sensor module are respectively connected to the main controller module, and the main controller is also connected to the spectrum adjustment module.
[0040] Specifically, the system structure diagram of this utility model embodiment is as follows: Figure 2 As shown, the main controller module 101 uses a Raspberry Pi 4B microcomputer as the core control unit of the system. The image recognition module 102 includes a high-definition RGB camera with a USB interface for capturing plant images. The environmental spectrum sensor 103 uses an AS7341 model spectrum sensor, which is connected via I... 2 The Type-C interface connects to the Raspberry Pi for measuring spectral intensity in the 400nm-700nm range. The Raspberry Pi's GPIO pins connect to a custom PWM LED driver board (for the spectral adjustment module's driver circuitry), which drives a three-channel LED board containing red (660nm), blue (450nm), and cool white (5000K) light. The main controller module 101's specific functions include:
[0041] (1) Call the camera driver, capture images, and load and execute the recognition model.
[0042] (2) Through I 2 C communicates with the spectral sensor to read and parse spectral data.
[0043] (3) Receive the recognition result, query the spectral strategy library, compare the target spectrum with the actual spectrum fed back by the sensor, and run the control algorithm (PID controller) to calculate the required PWM signal adjustment value (i.e., the spectral adjustment signal). The spectral strategy library is stored in the Raspberry Pi SD card in the form of a database. It is a key data structure that stores mapping relationships such as {plant variety: lettuce, growth stage: seedling stage} -> {target spectral formula: red light accounts for 30%, blue light accounts for 50%, white light accounts for 20%}.
[0044] (4) Generate a PWM signal and control the LED driver circuit through the GPIO pin.
[0045] Specifically, in the absence of an environmental spectrum sensor, the main controller module 101 directly generates the corresponding PWM signal after querying the target spectrum, without needing to compare it with the actual spectrum.
[0046] Specifically, the image recognition module 102 of this embodiment includes an acquisition device and a recognition device. The acquisition device uses a high-definition RGB camera (1080p resolution) with a USB interface. The camera is fixed above the plant canopy, directly facing the center of the plant, to ensure that the complete morphological features of the plant are captured. The recognition device uses a pre-trained plant recognition model, which can be selected from existing plant recognition apps on the market; alternatively, a recognition model can be trained according to the actual situation. For example, a recognition model can be trained using a large dataset of labeled images containing the target plant (such as lettuce, strawberry, tomato) at different growth stages (seedling, vegetative growth, flowering, fruiting). Based on the acquired images, the model outputs the plant variety (such as "strawberry") and growth stage (such as "fruit ripening stage").
[0047] Specifically, the environmental spectral sensing module 103 in this embodiment of the invention employs an integrated digital spectral sensor AS7341. This type of sensor integrates a photosensitive element, an optical filter (for separating different wavelength bands), and a signal processing circuit, directly communicating via a digital interface (I...). 2 C) Outputting light intensity values for different wavelength channels simplifies system design. The AS7341 has 8 visible light channels and 1 near-infrared channel. Its spectral response range should cover the main region of photosynthetically active radiation (PAR, 400-700nm). The AS7341 sensor is mounted at mid-height in the plant canopy with the probe facing upwards to accurately measure the actual light spectrum and intensity received by the plant leaves. This complements the camera's overhead imaging. The AS7341 sensor provides closed-loop feedback, directly measuring the combined spectrum actually emitted by the LED panel, which may be mixed with ambient light (such as sunlight coming through a window). The main controller uses this to determine the difference between the current light intensity and the target formula, thereby performing precise compensation to eliminate the effects of LED light decay, temperature drift, or changes in ambient light.
[0048] Specifically, the spectral adjustment module 104 in this embodiment of the invention employs a PWM constant current drive circuit and a multi-channel LED light board. The constant current source ensures stable current flowing through each LED channel, avoiding brightness fluctuations caused by changes in the LED's forward voltage, thereby guaranteeing spectral stability and LED lifespan. The Raspberry Pi's GPIO pins output PWM signals with different duty cycles to the drive circuit. The drive circuit converts the weak PWM signals into switching signals capable of driving high-current LEDs. The PWM duty cycle is linearly adjusted from 0% to 100%, corresponding to stepless adjustment of LED brightness from off to maximum brightness. The multi-channel LED light board includes three independent channels: red light (peak wavelength 660nm), blue light (peak wavelength 450nm), and broadband white light (color temperature 5000K). This design is because red / blue light is the most efficient wavelength for plant photosynthesis and is the core of spectral regulation. White light, on the one hand, provides a continuous broadband spectrum, supplementing the missing wavelengths of red and blue light (such as green light), which may be beneficial to the morphogenesis of certain plants; on the other hand, it makes the light emitted by the light source more natural to the human eye, facilitating artificial management and observation.
[0049] Specifically, the connection relationship of the main controller 101 in this embodiment of the present invention is as follows: Figure 3 and Figure 4 As shown, the Raspberry Pi's interface connections are as follows: the power connector connects to a 5V power supply; the USB-A connector connects to the output interface of the identification device; the 3.3V VCC connector connects to the power pin of the integrated digital spectral sensor AS7341; and the GPIO2 connector connects to the I / O pin of the integrated digital spectral sensor AS7341. 2 C data line pin; GPIO3 interface connects to the I / O pin of the integrated digital spectral sensor AS7341. 2 The C pin is the clock line pin; GPIO12 is connected to the first channel input of the driver circuit; GPIO13 is connected to the second channel input of the driver circuit; and GPIO18 is connected to the third channel input of the driver circuit. The three channels of the driver circuit control the switching signals of three different LEDs respectively.
[0050] Specifically, in this embodiment of the invention, the LED chips should be evenly and alternately distributed on the lamp board to ensure that the spectrum irradiated onto the plant canopy is uniformly mixed and without obvious color spots.
[0051] In operation, the Raspberry Pi periodically controls the camera to capture images of the plant and calls the locally running recognition model or recognition APP for real-time recognition to determine the plant variety (e.g., strawberry "Yuexiu") and growth stage (e.g., the ripening stage). Simultaneously, the Raspberry Pi reads data from the spectral sensor. The Raspberry Pi has a pre-stored spectral strategy library. Based on the identified "Yuexiu - ripening stage", it queries the target spectral formula (e.g., R:B:W = 60%:30%:10%), then calculates the difference between the current spectral sensor reading and the target formula, and dynamically adjusts the brightness of each channel of the LED board by adjusting the duty cycle of the PWM signal output to the LED driver board until the actual spectrum fed back by the spectral sensor matches the target spectrum.
[0052] Therefore, this invention provides a plant adaptive spectral adjustment system that is compact, low-cost, and fast-responding. This system uses a camera to identify the plant's physical state and combines this with environmental spectral sensing to achieve precise closed-loop control of the LED light source.
[0053] The above are merely preferred embodiments of the present utility model and are not intended to limit the present utility model. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present utility model shall be included within the protection scope of the present utility model.
Claims
1. A plant adaptive spectral regulation system based on image recognition, characterized in that, include: The image recognition module is used to acquire and recognize plant images, and the recognition results include plant variety and growth stage. The main controller module, connected to the image recognition module, is used to receive the recognition results and output the spectral adjustment signal; The spectral adjustment module includes a driving circuit and an LED light board. The driving circuit is connected to the main controller module to receive the spectral adjustment signal and outputs a driving signal to the LED light board based on the spectral adjustment signal. The LED light board includes a first LED chip, a second LED chip, and a third LED chip, which emit three different types of light respectively. The three different types of light include red light, blue light, and broadband white light.
2. The plant adaptive spectral regulation system based on image recognition according to claim 1, characterized in that, The image recognition module includes an acquisition device and a recognition device; The acquisition device uses a high-definition RGB camera, which is fixedly installed above the plant canopy and directly facing the center of the plant to ensure that the complete morphological features of the plant are captured. The output of the high-definition RGB camera is connected to the input of the recognition device. The identification device uses a trained plant identification model, and the output of the plant identification model is connected to the main controller module.
3. The plant adaptive spectral regulation system based on image recognition according to claim 2, characterized in that, The drive circuit of the spectrum adjustment module is a PWM constant current drive circuit, which includes three circuit channels, namely: The first channel has its input end connected to the main controller and its output end connected to the first LED chip of the LED light board. The second channel has its input end connected to the main controller and its output end connected to the second LED chip of the LED light board. The third channel has its input end connected to the main controller and its output end connected to the third LED chip on the LED light board.
4. The plant adaptive spectral regulation system based on image recognition according to claim 3, characterized in that, The three types of LED chips on the LED light board are evenly and alternately arranged to ensure that the spectrum irradiated onto the plant canopy is uniformly mixed.
5. The plant adaptive spectral regulation system based on image recognition according to claim 4, characterized in that, Also includes: An environmental spectral sensing module is used to collect real-time spectral data of the environment surrounding the plant. The output of the environmental spectral sensing module is connected to the main controller module.
6. The plant adaptive spectral regulation system based on image recognition according to claim 5, characterized in that, The environmental spectral sensing module uses an integrated digital spectral sensor AS7341, which is installed at the middle height of the plant canopy with the probe facing upwards to accurately measure the actual light spectrum and intensity received by the plant leaves.
7. The plant adaptive spectral regulation system based on image recognition according to claim 6, characterized in that, The main controller module uses a Raspberry Pi 4B microcomputer, and its interface connection method is as follows: Connect the power interface to a 5V power source. The USB-A interface connects to the output interface of the identification device; The 3.3V VCC interface connects to the power supply pin of the integrated digital spectral sensor AS7341; The GPIO2 interface connects to the I / O pin of the AS7341 integrated digital spectral sensor. 2 C data cable pin; The GPIO3 interface connects to the I / O pin of the AS7341 integrated digital spectral sensor. 2 C clock line pin; The GPIO12 interface is connected to the first channel input of the driver circuit; The GPIO13 interface is connected to the second channel input of the driver circuit; The GPIO18 interface is connected to the third channel input of the driver circuit.
