Agricultural greenhouse automatic dimming image acquisition method based on visual task feedback

Abstract

The invention discloses an agricultural greenhouse automatic dimming image acquisition method based on visual task feedback, and relates to the field of agricultural data acquisition, and the method comprises the steps: collecting a crop image through a depth camera, carrying out the target detection, extracting an ROI region of a crop, and carrying out the target detection based on an RGB image and a depth image in a detection frame of the crop; the method comprises the steps of calculating a comprehensive image quality index used for representing the overall texture richness and the overall detection reliability of each crop, and then dynamically adjusting the driving current of an LED lamp panel based on the comprehensive image quality index so as to dynamically adjust light. Even if the illumination conditions of different areas of the agricultural greenhouse change and the crops have the phenomena of strong reflection, local shielding and the like, closed-loop dimming can be carried out based on the local image quality of the crops, so that the quality of the collected crop images is improved, and the accuracy and reliability of image analysis are improved.

Description

Technical Field

[0001] This application relates to the field of agricultural data acquisition, and in particular to an automatic dimming image acquisition method for agricultural greenhouses based on visual task feedback. Background Technology

[0002] In facility agriculture production, the greenhouse environment provides controlled growth conditions for crops, effectively improving crop yield and quality. In recent years, with the rapid development of technologies such as computer vision and deep learning, image analysis-based technologies for crop growth monitoring, maturity recognition, yield prediction, and disease detection have become research hotspots in the field of smart agriculture. The application of these technologies largely depends on the quality of images collected of crops inside agricultural greenhouses. Currently, the common method is to use agricultural robots to move along the crop cultivation rows in agricultural greenhouses to collect images. However, in the actual agricultural greenhouse production environment, the lighting inside the greenhouse is affected by factors such as weather, time of day, sun angle, light transmittance of greenhouse covering materials, and aging degree, resulting in uneven light intensity and obvious directionality. In addition, complex background environments and other interference factors lead to the current low quality of crop image acquisition in agricultural greenhouses. Summary of the Invention

[0003] This application addresses the aforementioned problems and technical needs by proposing an automatic dimming image acquisition method for agricultural greenhouses based on visual task feedback. The technical solution of this application is as follows: An automatic dimming image acquisition method for agricultural greenhouses based on visual task feedback is disclosed. This method is used in an automatic dimming image acquisition system for agricultural greenhouses. The system includes a mobile platform, a depth camera, an LED light panel, a constant current drive circuit, and an edge computing device mounted on the mobile platform. The edge computing device is connected to the depth camera and the constant current drive circuit, which in turn connects to the LED light panel to provide drive current. The field of view of the depth camera covers the area where the crop cultivation rows are located in the agricultural greenhouse, and the illumination range of the LED light panel covers the field of view of the depth camera. As the mobile platform moves along the crop cultivation rows in an agricultural greenhouse, the automatic dimming image acquisition method for the agricultural greenhouse executed by the edge computing device includes: Using a depth camera to collect the first Frame image, number Frame images include RGB images and depth images , The parameter is an integer. Using a pre-trained crop detection network to analyze RGB images The target detection in the middle is obtained as the first In the frame image A detection box for each crop, with integer parameters. ; Based on the RGB images within the detection boxes of each crop and depth images Calculate the first Comprehensive image quality index of frame images , No. The higher the overall texture richness of each crop in the frame image, the higher the overall detection reliability. The higher the image quality of a frame, the better the overall image quality index. The larger the value; Based on comprehensive image quality indicators Convergence to the target quality index To control the target, the constant current drive circuit is controlled to adjust the drive current supplied to the LED light board and continue to acquire the next frame of image.

[0004] A further technical solution involves controlling the constant current drive circuit to adjust the drive current supplied to the LED light board, including: when or At that time, the constant current drive circuit maintains a constant drive current supplied to the LED light board. This is the error threshold; when and At that time, from Begin, calculate the first... Comprehensive image quality index of frame images Overall image quality index compared with the previous frame Changes in quality indicators between And based on the change in quality indicators The constant current drive circuit is controlled to adjust the drive current supplied to the LED light board.

[0005] Its further technical solution is to base it on the change in quality indicators. The control of the constant current drive circuit regulates the drive current supplied to the LED board, including: when At that time, the constant current drive circuit controls the drive current supplied to the LED board to adjust in the same current adjustment direction as the previous drive current adjustment; when At this time, the constant current drive circuit controls the drive current supplied to the LED board to adjust in the opposite direction of the previous drive current adjustment. The direction of current adjustment is either the direction of increasing drive current or the direction of decreasing drive current.

[0006] Its further technical solution is to base it on the change in quality indicators. The control of the constant current drive circuit to regulate the drive current supplied to the LED light board also includes: Calculate the reference drive current based on the current adjustment direction. ;in, This is the driving current currently supplied to the LED board by the constant current driving circuit. It refers to the current adjustment step size; when the current adjustment direction is in the direction of increasing drive current... When the current adjustment direction is in the direction of decreasing drive current, ; Based on the reference drive current Obtain the target driving current The constant current drive circuit is controlled to adjust to the target drive current. Provided to LED light panels.

[0007] A further technical solution is to use a reference drive current. Obtain the target driving current include: For reference drive current Smoothing filtering is performed to obtain the filtered reference drive current. , Indicates the smoothing parameter; For the filtered reference drive current The target driving current is obtained by performing a limiting process. ;in, This is the minimum operating current of the LED light board. This is the maximum operating current of the LED light panel.

[0008] Its further technical solution is to calculate the first... Comprehensive image quality index of frame images include: According to the Any frame in the image Depth image within the detection box of each crop Calculation yields the first Spatial correlation weight of each crop Integer parameters ; According to the RGB image within the detection box of each crop Calculation yields the first Image texture index of a crop The image texture index of each crop is weighted using the spatial correlation weights of each crop to obtain the first... Overall texture richness index of frame image ; The confidence scores of the detection boxes for each crop are weighted using the spatial correlation weights of each crop to obtain the first crop. Overall detection reliability index of frame images , It is the first The first frame of the image Confidence of the detection box for each crop; Get the first Comprehensive image quality index of frame images Among them, the weighting coefficient .

[0009] Its further technical solution is, according to the first Any frame in the image Depth image within the detection box of each crop Calculation yields the first Spatial correlation weight of each crop include: Calculate the first The average of the effective depth data of the detection box of each crop is used to obtain the first crop. Average depth of crops ; Comprehensive Average depth of crops and the The attribute parameters of the detection box for the crop are obtained. Spatial correlation weight of each crop Among them, the first Average depth of crops The smaller, the first The larger the detection frame for each crop, the better. The higher the confidence level of the detection box for each crop, the better. Spatial correlation weight of each crop The larger.

[0010] A further technical solution involves assigning attribute parameters to each detection box, including its width, height, and confidence level, and calculating the confidence level using the following formula. Spatial correlation weight of each crop :

[0011] in, It is the first The area of ​​the detection frame for each crop and , It is the first The width of the detection box for each crop. It is the first The height of the detection frame for each crop. It is to prevent the parameter from being removed from zero.

[0012] Its further technical solution is, according to the first RGB image within the detection box of each crop Calculation yields the first Image texture index of a crop include: For the first RGB image within the detection box of each crop Perform grayscale conversion and statistically analyze any grayscale level within the grayscale range. Number of pixels Obtain the histogram and normalize it to get the gray levels. probability distribution , Represents grayscale level The number of pixels, This represents summing the number of pixels across all gray levels; Calculate the first Information entropy of a crop ; Information entropy Normalization is performed to obtain the first Image texture index of a crop ,in, It is the first All in the frame image The minimum information entropy of a crop. It is the first All in the frame image The maximum value of the information entropy of a crop. It is to prevent the parameter from being removed from zero.

[0013] A further technical solution is that the automatic dimming image acquisition method for agricultural greenhouses also includes: During the self-calibration phase, as the mobile platform moves along the crop cultivation rows in the agricultural greenhouse, the edge computing device sequentially controls the constant current drive circuit to switch to... Multiple different drive currents within a given range are supplied to the LED light board. Under each drive current, a depth camera is used to acquire self-calibrated images, and the overall image quality index of the self-calibrated images is calculated. This is the minimum operating current of the LED light board. This is the maximum operating current of the LED light board; Determine the maximum value of the overall image quality index for all self-calibrated images. And determine the target quality indicators that match the current environment. ;in, .

[0014] The beneficial technical effects of this application are: This application discloses an automatic dimming image acquisition method for agricultural greenhouses based on visual task feedback. The method uses a depth camera to acquire crop images and performs target detection to extract the ROI region of the crop. Based on the RGB image and depth image within the crop's detection box, a comprehensive image quality index is calculated to characterize the overall texture richness and overall detection reliability of each crop. Then, the driving current of the LED light panel is dynamically adjusted based on the comprehensive image quality index to achieve dynamic dimming. As the automatic dimming image acquisition system moves along the crop cultivation rows in the agricultural greenhouse, even if the lighting conditions change in different areas of the agricultural greenhouse, or if the crops exhibit strong reflections or partial shading, closed-loop dimming can be performed based on the local image quality of the crops. This helps to improve the quality of the acquired crop images, thereby improving the reliability of subsequent intelligent algorithms such as fruit counting, grading, and disease identification.

[0015] This method proposes an image quality evaluation system based on fruit ROI. It calculates spatial correlation weights for each crop's ROI and uses these weights to fuse information entropy and confidence. This makes dimming control directly related to crop detection results, without being misled by background highlights and overall image brightness, which better meets the actual needs of agricultural image acquisition.

[0016] This method employs a trend-based adaptive dimming approach, determining whether to increase or decrease light by analyzing the trend of the overall image quality index as a function of supplementary lighting. This allows the system to automatically approach the locally optimal supplementary lighting level under complex lighting conditions. Furthermore, this method introduces an online self-calibration approach. As the edge computing device moves along the crop cultivation rows in an agricultural greenhouse, it adjusts the drive current to automatically obtain the target quality index under the current environment. It requires no manual calibration and can adapt to different crops, greenhouses and light conditions through automated calibration. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the structure of an automatic dimming image acquisition system for an agricultural greenhouse according to an embodiment of this application.

[0018] Figure 2 This is a flowchart of an embodiment of an agricultural greenhouse automatic dimming image acquisition method according to this application.

[0019] Figure 3 This is the first embodiment of the present application. Comprehensive image quality index of frame images The method flowchart. Detailed Implementation

[0020] The specific embodiments of this application will be further described below with reference to the accompanying drawings.

[0021] This application discloses a method for automatic dimming image acquisition in agricultural greenhouses based on visual task feedback. This method is applied to an automatic dimming image acquisition system for agricultural greenhouses. Please refer to [reference needed]. Figure 1 The system structure diagram shown indicates that the automatic dimming image acquisition system for agricultural greenhouses includes a mobile platform 1, a depth camera 2, an LED light panel 3, a constant current drive circuit 4, and an edge computing device 5 mounted on the mobile platform 1.

[0022] Depth camera 2 is an integrated RGB-D camera used to simultaneously acquire RGB images and depth maps. Edge computing device 5 uses a high-performance edge computing module with a GPU or NPU to execute the automatic dimming image acquisition method for this agricultural greenhouse. Edge computing device 5 connects to depth camera 2; in one example, the two communicate via Ethernet or USB 3.0.

[0023] The edge computing device 5 is also connected to a constant current drive circuit, which provides drive current to the LED light panel 3. The constant current drive circuit supports PWM or analog current regulation. In one embodiment, the agricultural greenhouse automatic dimming image acquisition system can also be equipped with multiple LED light panels and constant current drive circuits, with the edge computing device 5 driving multiple LED light panels 3 through multiple sets of constant current drive circuits.

[0024] The automatic dimming image acquisition system for agricultural greenhouses moves as a whole, driven by a mobile platform 1. The mobile platform 1 can also be directly driven by an edge computing device 5, or controlled by a separate motion control module; details will not be elaborated here. The positions and orientations of the depth camera 2 and the LED light panel 3 are pre-calibrated so that, as the mobile platform 1 drives the automatic dimming image acquisition system along the crop rows in the agricultural greenhouse, the field of view of the depth camera 2 covers the area where the crop rows are located, and the illumination range of the LED light panel 3 covers the field of view of the depth camera 2. This automatic dimming image acquisition system for agricultural greenhouses also includes other modules, such as power supply circuits, which will not be detailed here.

[0025] As the mobile platform moves along the crop cultivation rows in the agricultural greenhouse, the edge computing device 5 executes the following automatic dimming image acquisition method for the agricultural greenhouse, including the following steps. Please refer to [link / reference needed]. Figure 2 : Step 210, use depth camera 2 to collect the first... Frame image.

[0026] The obtained number Frame images include RGB images and depth images , It is an integer parameter with an initial value of 1.

[0027] Step 220: Use a pre-trained crop detection network to process the RGB image. The target detection in the middle is obtained as the first In the frame image The detection box for each crop.

[0028] Where, integer parameters The crop detection network is a target detection network pre-trained on a neural network and deployed in an edge computing device. The type of crop to be identified and detected is determined according to the actual detection needs, such as various types of fruit.

[0029] The bounding box for each crop output by the crop detection network is the ROI region for that crop, and the resulting... Any frame in the image The attribute parameters of the detection box for each crop include: the center coordinates of the detection box, and the confidence level of the detection box. Detection frame width Detection frame height Integer parameters .

[0030] Step 230: Based on the RGB images within the detection boxes of each crop. and depth images Calculate the first Comprehensive image quality index of frame images .

[0031] No. Comprehensive image quality index of frame images Used to characterize the Image quality of frame image, the first The higher the overall texture richness of each crop in the frame image, the higher the overall detection reliability, and the more beneficial it is for subsequent use of the first-order image. If various image analysis operations are performed on the crop using the frame image, the resulting image will be the first... Comprehensive image quality index of frame images The larger the value, the more it represents the first... The higher the image quality of the frame, the better.

[0032] In order to make the first Comprehensive image quality index of frame images Capable of accurately characterizing the first In one embodiment, the overall texture richness and overall detection reliability of each crop in a frame image are calculated using the following method: Comprehensive image quality index of frame images Please refer to Figure 3 The flowchart shown: 1. First, according to the first Any frame in the image Depth image within the detection box of each crop Calculation yields the first Spatial correlation weight of each crop ,include: Calculate the first The average of the effective depth data of the detection box of each crop is used to obtain the first crop. Average depth of crops In one embodiment, the first The effective depth data for the detection box of a crop is the depth image within the detection box. After data filtering, the depth data will be... Depth image within the detection box of each crop Depth values ​​of 0 and those exceeding the working range are filtered out; the remaining depth data is the first... The effective depth data of the detection frame for each crop can be used to eliminate noisy depth data and improve accuracy.

[0033] Then, combining the first Average depth of crops and the The attribute parameters of the detection box for the crop are obtained. Spatial correlation weight of each crop . No. Average depth of crops Used to characterize the The distance between each crop and depth camera 2 is considered important. Crops closer to depth camera 2 are more important for image resolution and detail. Additionally, crops with higher confidence levels and larger areas in the image are also more important for image resolution and detail. Based on this consideration, the first... Average depth of crops The smaller, the first The larger the detection frame for each crop, the better. The higher the confidence level of the detection box for each crop, the better the result. Spatial correlation weight of each crop The larger. In one embodiment, the first Spatial correlation weight of each crop The calculation formula is:

[0034] in, It is the first The area of ​​the detection frame for each crop and , It is the first The width of the detection box for each crop. It is the first The height of the detection frame for each crop. It is to prevent the parameter from being removed from zero.

[0035] 2. Then according to the first RGB image within the detection box of each crop Calculation yields the first Image texture index of a crop .

[0036] No. Image texture index of a crop Used to characterize the Texture richness of individual crops, image texture index The larger the value, the higher the texture richness represented. In one embodiment, the calculation of the... Image texture index of a crop The methods include: For the first RGB image within the detection box of each crop Perform grayscale conversion and statistically analyze any grayscale level within the grayscale range. Number of pixels Obtain the histogram and normalize it to get the gray levels. probability distribution Then the first one can be calculated. Information entropy of a crop for:

[0037] in, Represents grayscale level The number of pixels, This represents summing the number of pixels across all gray levels.

[0038] Finally, regarding information entropy... Normalization is performed to obtain the first Image texture index of a crop :

[0039] in, It is the first All in the frame image The minimum information entropy of a crop. It is the first All in the frame image The maximum value of the information entropy of a crop. It is to prevent the parameter from being removed from zero.

[0040] 3. The image texture indices of each crop are weighted using spatial correlation weights to obtain the first... Overall texture richness index of frame image The formula is written as:

[0041] 4. Similarly, the confidence scores of the detection boxes for each crop are weighted using the spatial correlation weights of each crop to obtain the first... Overall detection reliability index of frame images The formula is written as:

[0042] 5. Finally, a comprehensive summary of the first... Overall texture richness index of frame image and overall testing reliability indicators , obtained the Comprehensive image quality index of frame images for:

[0043] Among them, the weighting coefficient The specific value can be predetermined.

[0044] Step 240, using comprehensive image quality indicators Convergence to the target quality index To control the target, the constant current drive circuit 4 adjusts the drive current supplied to the LED light board 3 and continues to acquire the next frame of image.

[0045] The target quality indicator This is a pre-calibrated value, representing the theoretical maximum value achievable by the comprehensive image quality index under current agricultural greenhouse conditions. Therefore, during adjustment: when or When the convergence target is reached, the constant current drive circuit maintains a constant drive current supplied to the LED board. It is an error threshold and can be customized.

[0046] when and At this time, it indicates that the convergence target has not yet been reached. In the context of agricultural greenhouses, the target quality index... The relationship between the brightness of the LED light panel and the brightness of the LED light panel is not monotonically changing. Therefore, increasing the driving current to improve brightness will not necessarily improve image quality. Based on this, the adjustment method provided in this embodiment is as follows: when At time 1, the current adjustment direction is preset to either the direction of increasing or decreasing drive current, and then the constant current drive circuit is controlled to adjust the drive current according to the preset current adjustment direction.

[0047] from Begin, calculate the first... Comprehensive image quality index of frame images Overall image quality index compared with the previous frame Changes in quality indicators between And based on the change in quality indicators The constant current drive circuit is controlled to adjust the drive current supplied to the LED board. Specifically: when This indicates that after the last drive current adjustment, it helps to improve the overall image quality index towards the target quality index. When the LED approaches, the constant current drive circuit continues to adjust the drive current supplied to the LED board in the same direction as the previous drive current adjustment.

[0048] when If the previous drive current adjustment resulted in a decrease in overall image quality, the constant current drive circuit will adjust the drive current supplied to the LED board in the opposite direction to the previous adjustment. In other words, if the previous drive current adjustment was in the direction of increasing drive current, the current adjustment will be in the direction of decreasing drive current. Conversely, if the previous drive current adjustment was in the direction of decreasing drive current, the current adjustment will be in the direction of increasing drive current.

[0049] Through iterative adjustments, the overall image quality index can be improved. Convergence to the target quality index Furthermore, because a method of dynamically adjusting the drive current is used, therefore... The current adjustment direction set at time 1 can be random. Even if the set current adjustment direction is inaccurate, The current regulation direction can be quickly adjusted in 2 hours.

[0050] Once the direction of current adjustment is determined, the reference drive current can be calculated based on that direction. :

[0051] in, It is the driving current currently supplied to the LED board by the constant current driving circuit. This is a preset current adjustment step size. To avoid sudden changes in image brightness caused by excessive variations in supplementary light intensity within a single frame, this current adjustment step size... It is usually a small value. It is a sign function indicating the direction of current adjustment. When the direction of current adjustment is the direction of increasing drive current, When the current adjustment direction is in the direction of decreasing drive current, .

[0052] One approach is to directly control the constant current drive circuit 4 to adjust it according to the reference drive current. Provided to LED light board 3. However, to avoid sudden changes in image brightness caused by excessive variations in supplementary light intensity within a single frame, in another embodiment, the reference drive current is further adjusted. The target driving current is obtained after processing. Then, the constant current drive circuit 4 is controlled to adjust to the final determined target drive current. Provided to LED light board 3.

[0053] To avoid sudden brightness changes, the reference drive current is adjusted. The processes performed include: First, the reference drive current is... Smoothing filtering is performed to obtain the filtered reference drive current. :

[0054] in, Indicates the smoothing parameter. The value can be customized, and The smaller the value, the slower the change in driving current and the more stable the brightness.

[0055] Then the filtered reference drive current... The target driving current is obtained by performing a limiting process. :

[0056] in, This is the minimum operating current of the LED light board. This is the maximum operating current of the LED light board. This ensures the final target driving current. Within the operating current range of LED light board 3, the safe operation of LED light board 3 is ensured.

[0057] When the agricultural greenhouse automatic dimming image acquisition system is equipped with multiple LED light panels 3, each LED light panel 3 is adjusted according to the method provided in the above embodiment.

[0058] The above dynamic dimming operation is based on a pre-set target quality index. The target quality indicator While empirical values ​​can be used, considering the complexity of the agricultural greenhouse environment—where ambient light intensity varies between different greenhouses or at different times—and the reflective properties differ depending on the crop type, all these factors contribute to variations in the overall image quality index. To avoid the subjectivity of manually set empirical values, in one embodiment, the edge computing device in this agricultural greenhouse automatic dimming image acquisition system pre-determines the target quality index through a self-calibration phase before moving to acquire the image. This includes the following processes: As the mobile platform moves along the crop cultivation rows in the agricultural greenhouse, the edge computing devices sequentially control the constant current drive circuit to switch to... Multiple different drive currents within a given range are supplied to the LED light board 3. Under each drive current, a self-calibration image is acquired using the depth camera 2, and the comprehensive image quality index of the self-calibration image is calculated according to steps 220 and 230 described above. During self-calibration... Within the range, select several typical drive currents with larger step sizes; there is no need to adjust the step size according to the smaller current during actual adjustment. Make fine adjustments.

[0059] Then determine the maximum value of the overall image quality index for all self-calibrated images. This maximum value reflects the best image quality ever achieved under the combined effects of natural and supplemental lighting. Then, a target quality index matching the current environment is finally determined. .in, Thus, the target quality indicators Slightly below the observed maximum value This allows for a buffer and avoids frequent oscillations near extreme values.

[0060] The above descriptions are merely preferred embodiments of this application, and this application is not limited to the above embodiments. It is understood that other improvements and variations that can be directly derived or conceived by those skilled in the art without departing from the spirit and concept of this application should be considered to be included within the protection scope of this application.

Claims

1. A method for automatic dimming image acquisition in agricultural greenhouses based on visual task feedback, characterized in that, The automatic dimming image acquisition method for agricultural greenhouses is used in an automatic dimming image acquisition system for agricultural greenhouses. The automatic dimming image acquisition system for agricultural greenhouses includes a mobile platform, and a depth camera, an LED light panel, a constant current driving circuit, and an edge computing device mounted on the mobile platform. The edge computing device is connected to the depth camera and the constant current driving circuit, and the constant current driving circuit is connected to the LED light panel to provide driving current. The field of view of the depth camera covers the area where the crop cultivation rows are located in the agricultural greenhouse, and the illumination range of the LED light panel covers the field of view of the depth camera. As the mobile platform moves along the crop cultivation rows in an agricultural greenhouse, the automatic dimming image acquisition method for the agricultural greenhouse executed by the edge computing device includes: Using a depth camera to collect the first Frame image, number Frame images include RGB images and depth images , Integer parameter; Using a pre-trained crop detection network to analyze RGB images The target detection in the middle is obtained as the first In the frame image A detection box for each crop, with integer parameters. ; Based on the RGB images within the detection boxes of each crop and depth images Calculate the first Comprehensive image quality index of frame images , No. The higher the overall texture richness of each crop in the frame image, the higher the overall detection reliability. The higher the image quality of a frame, the better the overall image quality index. The larger the value; Based on comprehensive image quality indicators Convergence to the target quality index To control the target, the constant current drive circuit is controlled to adjust the drive current supplied to the LED light board and continue to acquire the next frame of image.

2. The automatic dimming image acquisition method for agricultural greenhouses according to claim 1, characterized in that, The control of the constant current drive circuit regulates the drive current supplied to the LED light board, including: when or At that time, the constant current drive circuit maintains a constant drive current supplied to the LED board. This is the error threshold; when and At that time, from Begin, calculate the first... Comprehensive image quality index of frame images Overall image quality index compared with the previous frame Changes in quality indicators between And based on the change in quality indicators The constant current drive circuit is controlled to adjust the drive current supplied to the LED light board.

3. The automatic dimming image acquisition method for agricultural greenhouses according to claim 2, characterized in that, Based on the change in quality indicators The control of the constant current drive circuit regulates the drive current supplied to the LED light board, including: when At that time, the constant current drive circuit controls the drive current supplied to the LED board to adjust in the same current adjustment direction as the previous drive current adjustment; when At this time, the constant current drive circuit controls the drive current supplied to the LED board to adjust in the opposite direction of the previous drive current adjustment. The direction of current adjustment is either the direction of increasing drive current or the direction of decreasing drive current.

4. The automatic dimming image acquisition method for agricultural greenhouses according to claim 3, characterized in that, Based on the change in quality indicators The control of the constant current drive circuit to regulate the drive current supplied to the LED light board also includes: Calculate the reference drive current based on the current adjustment direction. ;in, This is the driving current currently supplied to the LED board by the constant current driving circuit. It refers to the current adjustment step size; when the current adjustment direction is in the direction of increasing drive current... When the current adjustment direction is in the direction of decreasing drive current, ; Based on the reference drive current Obtain the target driving current The constant current drive circuit is controlled to adjust to the target drive current. Provided to LED light panels.

5. The automatic dimming image acquisition method for agricultural greenhouses according to claim 4, characterized in that, Based on the reference drive current Obtain the target driving current include: For reference drive current Smoothing filtering is performed to obtain the filtered reference drive current. , Indicates the smoothing parameter; For the filtered reference drive current The target driving current is obtained by limiting the amplitude. ;in, This is the minimum operating current of the LED light board. This is the maximum operating current of the LED light panel.

6. The automatic dimming image acquisition method for agricultural greenhouses according to claim 1, characterized in that, Calculate the first Comprehensive image quality index of frame images include: According to the Any frame in the image Depth image within the detection box of each crop Calculate the first Spatial correlation weights of individual crops Integer parameters ; According to the RGB image within the detection box of each crop Calculate the first Image texture index of a crop The image texture index of each crop is weighted using the spatial correlation weights of each crop to obtain the first... Overall texture richness index of frame image ; The confidence scores of the detection boxes for each crop are weighted using the spatial correlation weights of each crop to obtain the first crop. Overall detection reliability index of frame images , It is the first The first frame of the image Confidence of the detection box for each crop; Get the first Comprehensive image quality index of frame images Among them, the weighting coefficient .

7. The automatic dimming image acquisition method for agricultural greenhouses according to claim 6, characterized in that, According to the Any frame in the image Depth image within the detection box of each crop Calculate the first Spatial correlation weights of individual crops include: Calculate the first The average of the effective depth data of the detection box of each crop is used to obtain the first crop. Average depth of crops ; Comprehensive Average depth of crops and the The attribute parameters of the detection box for the crop are obtained. Spatial correlation weights of individual crops Among them, the first Average depth of crops The smaller, the first The larger the detection frame for each crop, the better. The higher the confidence level of the detection box for each crop, the better. Spatial correlation weights of individual crops The larger.

8. The automatic dimming image acquisition method for agricultural greenhouses according to claim 7, characterized in that, Each detection box's attribute parameters include its width, height, and confidence level. The confidence level is calculated using the following formula: Spatial correlation weights of individual crops : in, It is the first The area of ​​the detection frame for each crop and , It is the first The width of the detection box for each crop. It is the first The height of the detection frame for each crop. It is to prevent the parameter from being removed from zero.

9. The automatic dimming image acquisition method for agricultural greenhouses according to claim 6, characterized in that, According to the RGB image within the detection box of each crop Calculate the first Image texture index of a crop include: For the first RGB image within the detection box of each crop Perform grayscale conversion and statistically analyze any grayscale level within the grayscale range. Number of pixels Obtain the histogram and normalize it to get the gray levels. probability distribution , Represents grayscale level The number of pixels, This represents summing the number of pixels across all gray levels; Calculate the first Information entropy of a crop ; Information entropy Normalization is performed to obtain the first Image texture index of a crop ,in, It is the first All in the frame image The minimum information entropy of a crop. It is the first All in the frame image The maximum value of the information entropy of a crop. It is to prevent the parameter from being removed from zero.

10. The automatic dimming image acquisition method for agricultural greenhouses according to claim 1, characterized in that, The automatic dimming image acquisition method for agricultural greenhouses also includes: During the self-calibration phase, as the mobile platform moves along the crop cultivation rows in the agricultural greenhouse, the edge computing device sequentially controls the constant current drive circuit to switch to... Multiple different drive currents within a given range are supplied to the LED light panel. Under each drive current, a self-calibrated image is acquired using a depth camera, and a comprehensive image quality index of the self-calibrated image is calculated. This is the minimum operating current of the LED light board. This is the maximum operating current of the LED light board; Determine the maximum value of the overall image quality index for all self-calibrated images. And determine the target quality indicators that match the current environment. ;in, .

Images