Vegetation canopy photography steady-state exposure control method based on feature-constrained dynamic range

Abstract

This invention discloses a steady-state exposure control method for vegetation canopy photography based on feature-constrained dynamic range. A standardized reference time window is selected to acquire a baseline image set. The exposure energy system constant is calculated based on the baseline exposure time and irradiance. The target dynamic range is extracted from the baseline image set. In each exposure control round, the target dynamic range of the current image set is extracted, and the low-light and high-light deviations from the target dynamic range of the baseline image set are calculated and fused to obtain a comprehensive error. This error, along with the exposure energy system constant, current irradiance, and current exposure time, is used to predict the exposure time for the next round. This invention predicts the exposure time by calculating the deviation between the target dynamic range of the current image set and the target dynamic range of the baseline image set, thus solving the technical problem of overexposure and underexposure truncation in vegetation canopy photography under drastic fluctuations in dynamic lighting conditions in the field, and ensuring the edge geometric fidelity of the underlying image.

Description

Technical Field

[0001] This invention belongs to the field of ecological environment remote sensing monitoring technology, and more specifically, it relates to a method for steady-state exposure control of vegetation canopy photography based on feature-constrained dynamic range. Background Technology

[0002] The Leaf Area Index (LAI) is defined as half the total area of ​​green leaves per unit area of ​​the Earth's surface. It is a key biophysical parameter characterizing vegetation canopy structure, photosynthetic capacity, and the exchange of matter and energy within an ecosystem. In unattended IoT-based continuous observation systems in the field, ground-based photogrammetry based on digital images periodically captures images of the canopy hemisphere from bottom to top, and uses image segmentation techniques to extract the gap ratio between vegetation and the sky, thereby achieving high-frequency dynamic inversion of LAI. In this type of measurement framework, the radiative distribution characteristics of the front-end physical imaging fundamentally determine the separability of the subsequent vegetation and sky regions.

[0003] However, traditional optical measurement methods implicitly assume stringent conditions regarding lighting conditions, requiring either isotropic and uniform distribution of diffuse skylight or the avoidance of direct sunlight entering the sensor. In actual continuous all-weather observations, the system inevitably encounters severe challenges such as bright spots from direct sunlight, variable cloud cover, and high dynamic range (HDR) scenes. Existing automatic exposure control (AEC) algorithms for industrial cameras primarily rely on the grayscale mean feedback across the entire image. Under high dynamic range and intense lighting, this traditional control logic is prone to blindly pursuing overall brightness balance, leading to overexposure and overflow of large areas of sky gaps, forced grayscale truncation to 255, or severe underexposure and truncation of effective leaf edges hidden in shadows. This directly results in the inability of subsequent traditional pixel mean feedback to effectively preserve the deep radiation distribution details of the canopy image, causing a significant decrease in the accuracy of the retrieved vegetation parameters.

[0004] Meanwhile, relying solely on post-processing algorithms is insufficient to compensate for the loss of underlying information caused by front-end imaging control failures. Therefore, designing a steady-state exposure control method at the imaging source that can adapt to highly dynamic changes in the radiation field and ensure that the geometric edge features of the canopy are not distorted is a key bottleneck in achieving the standardization and long-term stable inversion of continuous observation data. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of the prior art and provide a steady-state exposure control method for vegetation canopy photography based on Feature-constrained Dynamic Range Control (F-CDRC). By calculating the deviation between the target dynamic range of the image set obtained in the current round and the target dynamic range of the reference image set, the exposure time is predicted. This solves the technical problem of overexposure and underexposure truncation in vegetation canopy photography under drastic fluctuations in high dynamic light in the field, and ensures the edge geometric fidelity of the underlying image.

[0006] To achieve the above-mentioned objective, the vegetation canopy photography steady-state exposure control method based on feature-constrained dynamic range of the present invention includes the following steps:

[0007] S1: Deploy an image acquisition module and an ambient light sensing module in the target area. The image acquisition module acquires images of the vegetation canopy at a set sampling rate. The ambient light sensing module synchronously collects the light intensity irradiance acting on the vegetation canopy. , Indicates the sampling time;

[0008] S2: Select a standardized reference time window and adjust the exposure time of the image acquisition module to ensure that the captured vegetation canopy images have optimal canopy structure visibility, within the current baseline exposure time. The baseline image set was acquired by the following method. , This indicates the number of reference images; the ambient light sensing module synchronously samples the corresponding... Each irradiance is measured individually, and then averaged to obtain the reference irradiance. ;

[0009] S3: Extract the baseline image set Target dynamic range of the middle vegetation canopy ,in , These represent the ideal dark area shadow detail target reference value and the ideal highlight unsaturated target reference value, respectively; and the exposure energy system constant is calculated. ;

[0010] S4: Record the current exposure control cycle as... An image set is created by collecting vegetation canopy images within the most recent time period. The current irradiance is obtained by averaging the light intensity irradiance sampled synchronously by the ambient light sensing module. ;

[0011] Using the same method as step S2, based on the image set Extracting the target dynamic range yields the target value for shadow detail in dark areas. and the target value of unsaturated highlights Low light deviation was calculated separately. and highlight deviation :

[0012] ,

[0013] ,

[0014] Then calculate the comprehensive error. :

[0015] ,

[0016] in, Indicates weight;

[0017] Then, based on the exposure energy system constant... Current irradiance Current exposure time Comprehensive error The timing of the next round of exposure can be predicted.

[0018] This invention relates to a steady-state exposure control method for vegetation canopy photography based on feature-constrained dynamic range. A standardized reference time window is selected to acquire a baseline image set. The exposure energy system constant is calculated based on the baseline exposure time and baseline irradiance. The target dynamic range is extracted from the baseline image set. In each exposure control round, the target dynamic range of the current image set is extracted, and the low-light deviation and high-light deviation between the target dynamic range and the baseline image set are calculated and fused to obtain a comprehensive error. This error, along with the exposure energy system constant, current irradiance, and current exposure time, is used to predict the exposure time for the next round.

[0019] The present invention has the following beneficial effects:

[0020] 1) Improving Radiation Distortion at the Imaging Source: This invention overcomes the limitations of traditional automatic exposure control that blindly pursues global average brightness. It introduces a dynamic histogram dual-constraint feature extraction method specifically for high dynamic range scenes to obtain the target dynamic range, anchoring the feedback target within the separable boundary between vegetation and sky gaps. Through asymmetric weighting of physical weight coefficients, higher sensitivity is given to low-light dark areas, effectively solving the technical defects of traditional AEC algorithms, such as "pore erosion" at forest window edges caused by large-area sky overexposure and underexposure truncation of leaves in dark areas, thus improving the quality of vegetation canopy images.

[0021] 2) Enhancing the steady-state robustness of all-weather time-series remote sensing inversion: This invention designs a control law that combines the rapid response of the feedforward channel with the precise adjustment of the multiplicative PID feedback channel. Even when encountering non-steady-state severe radiation shocks such as cloud dissipation and sudden changes in direct strong light, the imaging boundary can be rapidly converged. This fundamentally suppresses the diurnal anomalous curve oscillation of the leaf area index (LAI) time series generated by continuous high-frequency inversion throughout the day, ensuring the temporal smoothness and robustness of ecological IoT observation data. Attached Figure Description

[0022] Figure 1 This is a flowchart illustrating a specific implementation of the steady-state exposure control method for vegetation canopy photography based on feature-constrained dynamic range according to the present invention.

[0023] Figure 2 This is a flowchart of the exposure time prediction in this embodiment;

[0024] Figure 3 This is a structural diagram of the steady-state exposure control IoT experimental platform in this embodiment;

[0025] Figure 4 This is a control architecture diagram of the central processing and control module in this embodiment;

[0026] Figure 5 This is a comparison image of the vegetation canopy after exposure control using the present invention and the comparison method in this embodiment. Detailed Implementation

[0027] The specific embodiments of the present invention will now be described with reference to the accompanying drawings to enable those skilled in the art to better understand the invention. It should be particularly noted that in the following description, detailed descriptions of known functions and designs that might obscure the main content of the invention will be omitted here.

[0028] Example

[0029] Figure 1 This is a flowchart illustrating a specific implementation of the steady-state exposure control method for vegetation canopy photography based on feature-constrained dynamic range according to the present invention. Figure 1 As shown, the steady-state exposure control method for vegetation canopy photography based on feature-constrained dynamic range of the present invention includes the following steps:

[0030] S101: Deploy data acquisition device:

[0031] An image acquisition module and an ambient light sensing module are deployed in the target area. The image acquisition module acquires images of the vegetation canopy at a set sampling rate. The ambient light sensing module synchronously collects the light intensity irradiance acting on the vegetation canopy. , Indicates the sampling time.

[0032] S102: Calibrate initial exposure parameters:

[0033] To establish the boundaries of absolute features, initial exposure parameters need to be calibrated first. The specific method is as follows:

[0034] A standardized reference time window was selected, and the exposure time of the image acquisition module was adjusted to ensure optimal canopy structure visibility in the captured vegetation canopy image, meaning that the sky area in the vegetation canopy image was not overexposed and the edges of leaves in the dark areas were clear. Then, at the current reference exposure time... The baseline image set was acquired by the following method. , This indicates the number of reference images. The ambient light sensing module synchronously samples the corresponding... The average of the individual irradiance values ​​is used to obtain the reference irradiance. Generally, the normalized reference time window is set within the normalized reference time window of pure diffuse light, which is naturally governed by solar geometry and atmospheric diffuse scattering, namely the golden diffuse light window about 10 minutes before sunrise or after sunset.

[0035] In practical applications, the specific method for determining whether a vegetation canopy image has optimal canopy structure visibility is as follows:

[0036] Extract the target dynamic range of the vegetation canopy in the baseline image obtained at the current exposure time. If the lower limit of the target's dynamic range , If the preset threshold is met, the vegetation canopy image is determined to have optimal canopy structure visibility.

[0037] S103: Extract the baseline calibration parameters:

[0038] Extracting the benchmark image set Target dynamic range of the middle vegetation canopy ,in , These represent the ideal dark area shadow detail target reference value and the ideal highlight unsaturated target reference value, respectively.

[0039] The method for extracting the target dynamic range in this embodiment is as follows:

[0040] Using Otsu's method, each reference image was segmented into a vegetation foreground region and a sky background region, and statistical analysis was performed on each reference image. The 10th percentile of the vegetation foreground region in the vegetation foreground grayscale histogram , Seeking The 10th percentile The average value is used as a reference value for ideal shadow detail targets in dark areas. At this point, it is insensitive to extremely dark noise, thus effectively characterizing the boundaries of shadow details. Simultaneously, statistical analysis of each reference image is obtained. The 90th percentile of the grayscale histogram of the sky background Seeking The 90th percentile The average value is used as a reference value for an ideal unsaturated specular target. This effectively identifies the critical unsaturation level of the highlight region, thereby obtaining the target's dynamic range.

[0041] .

[0042] Based on the fundamental principle of photoelectric sensors within their linear response region, the exposure time of the reference stage is adopted. and reference irradiance To calibrate a system constant of exposure energy required to achieve the target image features under optimal diffuse illumination conditions. :

[0043] .

[0044] S104: Dynamically set exposure time:

[0045] The current exposure control round is: An image set is created by collecting vegetation canopy images within the most recent time period. The current irradiance is obtained by averaging the light intensity irradiance sampled synchronously by the ambient light sensing module. .

[0046] Using the same method as step S102, based on the image set Extracting the target dynamic range yields the target value for shadow detail in dark areas. and the target value of unsaturated highlights To eliminate the dimensional differences between different feature dimensions and to directly anchor the control target to the grayscale constraint boundary of the dynamic target range, the real-time feedback value and its corresponding target value are normalized, and the low-light deviation is calculated separately. and highlight deviation The specific formula is as follows:

[0047] ,

[0048] .

[0049] Considering that in vegetation canopy parameter estimation, the preservation of details in the shaded area is generally more sensitive to the accuracy of the final estimation than the need to suppress high light saturation, a weighting coefficient is introduced. The two error signals are fused to construct a comprehensive error. :

[0050] ,

[0051] in, Indicates the weight.

[0052] Then, based on the exposure energy system constant... Current irradiance Current exposure time Comprehensive error Predict the exposure time for the next round. .

[0053] In this embodiment, a multiplicative incremental PID feedback control law and a feedforward composite control law are constructed to collect the current irradiance. With comprehensive error Deep cascading and cross-coupling are performed to enable the prediction of exposure time. Figure 2 This is a flowchart illustrating the exposure time prediction in this embodiment. For example... Figure 2 As shown, the specific steps for exposure time prediction in this embodiment include:

[0054] S201: Feedback exposure time calculated based on multiplicative incremental PID control law:

[0055] This embodiment addresses the approximately logarithmic linear (non-linear) mapping relationship between exposure time and image brightness output in industrial cameras by designing a multiplicative incremental PID control law, based on the current exposure time. The multiplicative base is used to calculate the predicted exposure time for the next moment predicted by the feedback channel. :

[0056] ,

[0057] In the formula, This is the proportional gain coefficient. This is the integral gain coefficient. The differential gain coefficient, This represents the overall error of the previous exposure control time.

[0058] S202: Predicting transient exposure based on light intensity feedforward channel:

[0059] To improve the response to sudden changes in ambient light, the feedforward open-loop prediction channel uses the exposure energy system constant obtained from calibration. Calculate the current irradiance Feedforward Fast Response Expected Shutter Time under Disturbance :

[0060] .

[0061] S203: Constructing the overall control law for the feedforward-feedback composite control:

[0062] Then, the feedback gain coefficient is introduced. This is used to adjust the strength of the feedback correction, construct a feedforward-feedback composite control law, decouple and solve for the final ideal exposure total control law target value. :

[0063] .

[0064] To better illustrate the technical effects of the present invention, specific examples are used for experimental verification. This embodiment constructs an IoT experimental platform for steady-state exposure control. Figure 3 This is a structural diagram of the steady-state exposure control IoT experimental platform in this embodiment. Figure 3 As shown, the core hardware system configuration of the steady-state exposure control IoT experimental platform in this embodiment is as follows: The ambient light sensing module uses a digital light intensity meter (model: KM9560) with a spectral response range covering 400-700nm (highly consistent with the effective photosynthetic radiation band of vegetation) to continuously acquire real-time ambient irradiance data at a sampling frequency of 10 Hz; the image acquisition and exposure control module uses an industrial camera (model: HT-SUF1200C / M) with global shutter and external trigger functions. The CMOS sensor with a resolution of 4080×3072 pixels supports microsecond-level precise programming control of exposure time based on the GenICam protocol. The camera lens is a fisheye lens (model: KM12Y11-12MP) with a fixed focal length of 1.1 mm, fixed on a tripod with a bottom-up shooting angle; the central processing and control module acts as the intelligent hub, communicating with the digital light intensity meter through an RS-232 serial port and connecting to the industrial camera through a USB 3.0 interface.

[0065] The central processing and control module is used to implement the control logic in this invention. Figure 4 This is the control architecture diagram of the central processing and control module in this embodiment. (For example...) Figure 4 As shown, in this embodiment, the central processing and control module is implemented using a microprocessor. Through the general-purpose low-level programmable interface of the industrial camera software development kit (SDK), the calculated exposure time is processed... As the target for controlling the dynamic range mapping boundary of the camera's imaging, the parameter is sent to the firmware of the industrial camera in real time with high priority. The internal hardware exposure timer register value is directly rewritten through the high-speed hardware communication interface bus to complete the hardware timing reconstruction of the dynamic range of the dot matrix in the underlying imaging stage. After the parameter reconstruction is completed, the control core puts the system into a suspend waiting state until the next hardware control clock signal is triggered by the timer, so that the system automatically enters the next round of k+1 dynamic iterative control loop.

[0066] To comprehensively evaluate the overall technical performance of this invention, this embodiment conducted a multi-scene exposure control comparison experiment on four vegetation observation plots: woodland, shrubland, grassland, and farmland. The comparison methods included the camera's built-in traditional AEC mode, as well as seven existing exposure control methods: ZCM method, Blue Channel method, AAS method, improved information entropy method, HDR multiple exposure fusion method, and ROI optimization method.

[0067] This embodiment uses a standard window image of pure diffuse light as the reference image and conducts an exposure control comparison experiment during extreme periods of drastic fluctuations in the daytime high dynamic radiation field (such as the period when the direct sunlight intensity is highest at noon). Figure 5 These are comparison images of vegetation canopy after exposure control using the present invention and the comparative method in this embodiment. Table 1 is a comparison table of quality parameters between the present invention and the comparative method in this embodiment.

[0068]

[0069] Table 1

[0070] like Figure 5 As shown in Table 1, this invention uses the Structural Similarity Index (SSIM) to quantitatively quantify the pixel-level feature alignment consistency between the image to be evaluated acquired by each exposure control method and the pure diffuse light reference image captured by the standardized reference time window described in step S102 under extreme periods of drastic fluctuations in the daytime high dynamic radiation field. It comprehensively considers the similarity between the image to be evaluated and the reference image in three components: brightness, contrast, and structural information, to objectively reflect the ability of the acquired image to reproduce optimal imaging conditions. Statistical results show that the mean SSIM of the method of this invention (…) The SSIM standard deviation was reduced to 0.074, reaching 0.82. These two indicators represent the steady-state fidelity of imaging quality and the degree of fluctuation over time, respectively. Both radiometric consistency fidelity indicators are superior to the comparison methods.

[0071] The feedforward-feedback composite control architecture employed in this invention actively predicts and compensates for changes in illumination, and performs precise closed-loop adjustment constrained by the target dynamic range, thereby achieving optimal performance in both average quality and temporal stability. Visual comparisons and statistical data consistently demonstrate that, among eight exposure control methods, this invention exhibits more significant effectiveness in improving image quality and ensuring its steady-state stability.

[0072] Although the illustrative specific embodiments of the present invention have been described above to enable those skilled in the art to understand the invention, it should be understood that the invention is not limited to the scope of the specific embodiments. For those skilled in the art, various changes are obvious as long as they are within the spirit and scope of the invention as defined and determined by the appended claims, and all inventions utilizing the concept of the present invention are protected.

Claims

1. A method for steady-state exposure control of vegetation canopy photography based on feature-constrained dynamic range, characterized in that, Includes the following steps: S1: Deploy an image acquisition module and an ambient light sensing module in the target area. The image acquisition module acquires images of the vegetation canopy at a set sampling rate. The ambient light sensing module synchronously collects the light intensity irradiance acting on the vegetation canopy. , Indicates the sampling time; S2: Select a standardized reference time window and adjust the exposure time of the image acquisition module to ensure that the captured vegetation canopy images have optimal canopy structure visibility, within the current baseline exposure time. The baseline image set was acquired by the following method. , This indicates the number of reference images; the ambient light sensing module synchronously samples the corresponding... Each irradiance is measured individually, and then averaged to obtain the reference irradiance. ; S3: Extract the baseline image set Target dynamic range of the middle vegetation canopy ,in , These represent the ideal dark area shadow detail target reference value and the ideal highlight unsaturated target reference value, respectively. The exposure energy system constant was calculated. ; S4: Record the current exposure control cycle as... An image set is created by collecting vegetation canopy images within the most recent time period. The current irradiance is obtained by averaging the light intensity irradiance sampled synchronously by the ambient light sensing module. ; Using the same method as step S2, based on the image set Extracting the target dynamic range yields the target value for shadow detail in dark areas. and the target value of unsaturated highlights Low light deviation was calculated separately. and highlight deviation : ， ， Then calculate the comprehensive error. : ， in, Indicates weight; Then, based on the exposure energy system constant... Current irradiance Current exposure time Comprehensive error The timing of the next round of exposure can be predicted.

2. The steady-state exposure control method for vegetation canopy photography according to claim 1, characterized in that, The method for determining the optimal canopy structure visibility is as follows: Extract the target dynamic range of the vegetation canopy in the baseline image obtained at the current exposure time. If the lower limit of the target's dynamic range , If the preset threshold is met, the vegetation canopy image is determined to have optimal canopy structure visibility.

3. The steady-state exposure control method for vegetation canopy photography according to claim 1, characterized in that, The method for determining the target dynamic range is as follows: Using Otsu's method, each reference image was segmented into a vegetation foreground region and a sky background region, and statistical analysis was performed on each reference image. The 10th percentile of the vegetation foreground region in the vegetation foreground grayscale histogram and various reference images The 90th percentile of the grayscale histogram of the sky background , Then, the following target dynamic range is obtained by averaging: 。 4. The steady-state exposure control method for vegetation canopy photography according to claim 1, characterized in that, The specific method for predicting the exposure time is as follows: 1) Calculation based on multiplicative incremental PID control law according to the current exposure time Calculate the predicted exposure time for the next moment based on the feedback channel prediction. : ， In the formula, This is the proportional gain coefficient. This is the integral gain coefficient. The differential gain coefficient, This represents the overall error of the previous exposure control time. 2) Based on the exposure energy system constant Calculate the current irradiance Feedforward Fast Response Expected Exposure Time under Disturbance : ； 3) The feedforward-feedback composite control law is used to decouple and solve for the final exposure time of the next round. : ， in, This represents the feedback gain coefficient.

Images