The spectral reflectance of the vegetation canopy is defined as the ratio of the light energy reflected by the canopy in a specific band to the incident light energy in that band. The up-facing optical sensing unit of the VCR sensor captures incident light, while the down-facing optical sensing unit detects light reflected from the vegetation canopy. Given that the vegetation canopy receives light from the entire sky, the up-facing optical sensing unit has a 180°FOV. On sunny days, the light from the sky predominantly consists of direct sunlight (Liu and Jordan, 1960; Sen, 2004). The solar altitude influences the transmittance of the direct sunlight entering the sensor, thereby affecting the accuracy of the reflectance measurements. The optical window of the up-facing optical sensing unit is equipped with a cosine corrector made of polytetrafluoroethylene (PTFE), which could reduce the impact of the solar altitude on the measurements. Moreover, the cosine corrector ensures that the up-facing optical sensing unit maintains a 180°FOV. Additionally, an attenuator, a filter, and a photodetector are installed below the cosine corrector ( Figure 1B ). The down-facing optical sensing unit is equipped with filters and photodetectors from bottom to top. The FOV significantly influences the area of vegetation measured by the sensor when capturing spectral information, thereby affecting the measurement accuracy. A larger FOV covers a wider vegetation area, enabling the spectral information to better represent the overall vegetation conditions. However, in areas with incomplete vegetation coverage, particularly for row-planted crops, an excessively large FOV may capture spectral information about both the vegetation and soil, leading to inaccurate measurements. Conversely, a smaller FOV can effectively reduce or eliminate soil interference, but it may result in measurements that are not representative of the vegetation. Therefore, the FOV must be carefully balanced. Currently, an FOV of approximately 30° is commonly used. The VCR sensor is designed with an FOV of 28°. A schematic diagram of the FOV is shown in Figure 1C .

The filters used are FB710-10 and FB870-10 (Thorlabs Inc., Newton, NJ, USA), which have wavelengths of 710 nm and 870 nm, respectively, and a full width at half-maximum (FWHM) of 10 nm. The photodetector converts the optical signal into an electrical signal, and its spectral responsivity determines the performance of the sensor. The VCR sensor uses OPT101 photodiodes (Texas Instruments, Dallas, TX, USA), which have a spectral responsivity of 0.5–0.6 A/W at 710 nm and 870 nm. The photosensitive surface of the OPT101 measures 2.29 mm × 2.29 mm. The OPT101 photodiode operates in photoconductive mode, resutling in excellent linearity and a low dark current.

The sensor’s circuitry primarily consists of a power module, optical signal acquisition module, calibration module for the amplifier circuit resistance, and wireless communication module. The power module, wireless communication module, and up-facing optical signal acquisition module each has its own printed circuit board (PCB), while the down-facing optical signal acquisition module and microcontroller unit (MCU) are on the same PCB. This split design facilitates sensor installation and maintenance.

The power module utilizes two 3.7 V 14500 batteries connected in parallel to supply 3.3 V and −3.3 V power to the system. The OPT101 photodetector has an integrated operational amplifier that enables current-to-voltage (I–V) conversion, with outputs from pin 5. Different gains can be achieved by connecting an external feedback resistor between pins 2 and 5. The VCR sensor is intended for year-round operation. Given that the ambient light intensity varies significantly across the seasons, it is necessary to dynamically adjust the gain so as to obtain an appropriate voltage. The VCR sensor employs a T-type feedback network to achieve a high amplification factor using three small resistors ( Figure 2A ). R1 and R2 are precision resistors with a resistance of 68 kΩ and an accuracy of 0.01%. R3 is replaced by a digital potentiometer TPL0102 (Texas Instruments, Dallas, TX, USA), which has 256 taps and a resistance range of 0–100 kΩ. A larger resistor number indicates a higher resistance value. A smaller resistance value of R3 means a greater gain of the OPT101 (Equation 1). The accuracy of R3 directly affects the gain of the OPT101 and consequently the accuracy of the sensor’s reflectance measurement. Therefore, it is essential to calibrate the different resistance values of the TPL0102. Specifically, the TPL0102 is connected to the feedback network and a constant current source through an analog multiplexer, namely, ADG774 (Analog Devices, MA, USA), to achieve resistance calibration ( Figure 2A ). Due to the limited FOV and weak light reflected from the canopy, the optical signal received by the down-facing optical sensing unit is very weak and requires a higher gain. Based on preliminary testing under both strong and low-light conditions, the required gain ranges approximately from 1 M to 6 M. Hence, only resistors 1–25 of the TPL0102 are calibrated. The up-facing optical sensing unit receives strong light from the sky, so it requires only a small amount of amplification. The minimum required gain is approximately 0.2 M, as determined through initial testing. Hence, resistors 1–250 are calibrated. In the down-facing optical signal acquisition circuit, the TPL0102 is connected to a 100 µA constant current source, while the up-facing TPL0102 is connected to a 10 µA constant current source. The 100 µA current is provided by REF200 (Texas Instruments, Dallas, TX, USA), and the 10 µA current is provided by REF200 and OPA128 (Texas Instruments, Dallas, TX, USA) ( Figure 2B ). The on-resistance of ADG774 is 2.2 Ω, which is minimal and has a negligible impact on the other circuits during operation. The MCU uses the 16-bit MSP430I2041 (Texas Instruments, Dallas, TX, USA) for signal acquisition control, data calculation, and communication. The MSP430I2041 features a low supply voltage (2.2–3.6 V), ultra-low power consumption, five power-saving modes, and four 24-bit analog-to-digital converters (ADCs), making it suitable for portable devices and equipment used in field operations. The VCR sensor utilizes the MSP430I2041’s four built-in ADCs to convert analog electrical signals, which represent the light intensity, into digital signals. The sensor uses the ESP-12S (Ai-thinker Technology Co., Shenzhen, China) for wireless communication.

where G is the amplification (K); and R is the resistance value after calibration of the TPL0102 (kΩ). The calibration method is described in Section 2.2.1.

The embedded software system of the VCR sensor handles the measurement and communication tasks. The software was developed using the Contiki embedded operating system, which is event-driven and supports multithreading. The Contiki requires minimal resources during the run period, making it highly suitable for sensor applications (Dunkels et al., 2004). Upon powering on, the VCR sensor initializes, and then, it enters a sleep state and waits for events to occur. The events and processes of this system are shown in Figure 3 . The events include event_received_information, event_collecting_reflectance, event_collecting_reflectance_by_group, event_collecting_voltages, event_collecting_voltages_by_group, event_calibrating_resistance, event_calibrating_sensor, event_collecting_dark_current, and event_sending_data. The system processes include parsing_received_information, collecting_reflectance, collecting_reflectance_by_group, collecting_voltages, collecting_voltages_by_group, calibrating_resistors, calibrating_sensor, collecting_dark_current, and sending_data_to_WiFi.

The user group for this sensor includes staff who measure the spectral reflectance of the vegetation canopies and scientific researchers. To facilitate their work, a command set was designed, which includes the following commands: “COLLECT_Ref” for collecting the reflectance, “COLLECT_Ref Repeat[]” for collecting the reflectance by group, “COLLECT_Voltages” for collecting the voltages of the optical channels, “COLLECT_Voltages Repeat[]” for collecting the voltages of optical channels by group, “Calibrate_Resistors” for calibrating the resistors of TPL0102, “Calibrate_Sensor” for calibrating the reflectance of the sensor, and “Dark_Current” for collecting each channel’s dark current. The sensor’s temperature is also recorded during each measurement. The commands “COLLECT_Ref Repeat[]” and “COLLECT_Voltages Repeat[]” enable multiple measurements to be obtained using a single command. The brackets can be filled with multiple numbers separated by commas, each of which represents the number of sampling times during the measurement. For example, [1,5,10] means three measurements are required: the first is a single measurement, the second measurement is taken five times consecutively and the average is taken as the result, and the third measurement is taken ten times consecutively with the average as the result. Similarly, [5,5,5,5] indicates four measurements, each taken five times consecutively, and the averages are taken as the results. This type of command enables repeated measurements within a short time frame, which is particularly useful for studying the impact of the wind speed on the vegetation canopy reflectance.

Devices such as mobile phones and computers can connect to the sensor via Wi-Fi and control it by sending commands, and the sensor will send back the measured data. Upon receiving a command, the parsing_received_information process parses the content and triggers the corresponding event based on the command. Once the event is triggered, the corresponding process will run. For measurement commands, the event_sending_data event is triggered after the data are measured, and the sending_data_to_WiFi process sends the data back.

Figure 4 illustrates the processes for measuring the reflectance or voltages, as well as for measuring the reflectance or voltages by group. When measuring the reflectance or voltages ( Figure 4A ), the sensor first measures the voltage (V_710U) of the 710 nm up-facing optical sensing unit. If the voltage is less than 300 mV, the resistance of the TPL0102 is reduced to increase the gain of the OPT101. If the voltage exceeds 900 mV, the gain of the OPT101 is reduced. After multiple measurements, the final voltage (V _710U) is 300 m V ≤ V 710 U ≤ 900 m V . Subsequently, the gains of the 870 nm up-facing optical sensing unit and the 710 nm and 870 nm down-facing optical units are adjusted. The voltages of all four optical units are then re-measured. If reflectance measurements are required, the reflectance is calculated. Finally the event_sending_data event is triggered.

When measuring by group ( Figure 4B ), the gains of all of the OPT101 optical units are first adjusted using the method illustrated in Figure 4A . After adjustment, the first number (n ₁) from the sequence [n ₁, n ₂, …, n _i, …] is selected and V _710U, V _870U, V _710D, and V _870D are measured successively n ₁ times. The average voltages or reflectance values are calculated and recorded as the result. Next, the second number (n ₂) is taken from the sequence, and V _710U, V _870U, V _710D, and V _870D are measured n ₂ times. The average voltages or reflectance values are calculated and recorded again. This process is repeated until all of the numbers in the sequence are utilized. Finally, the event_sending_data event is triggered.

As can be seen from Equation 1, the accuracy of the TPL0102 resistance directly influences the gain of the OPT101. During calibration, the “Calibrate_Resistors” command was sent to the VCR sensor. The MCU then controlled the ADG774 ( Figure 2A ) to connect the TPL0102 chips within the 710 nm and 870 nm down-facing optical unit to a 100 µA current for calibration of resistors 1–25. The calibrated resistance value is denoted as R _D. Similarly, the ADG774 was controlled to connected the TPL0102 chips within the 710 nm and 870 nm up-facing optical unit to a 10 µA current for calibration of resistors 1–250. The calibrated resistance value is denoted as R _U. The calculation method for the resistance values is shown in Equations 2, 3.

where V is the output voltage of the sensor (0.1 mV).

The VCR sensor was placed in a black, light-proof box. The command “Dark_Current” was sent to the sensor. The MCU controlled the ADG774 to connect the TPL0102 to the OPT101 circuit, and the resistor number of TPL0102 switched to adjust the gain of the OPT101. The dark current, expressed as the measured voltage, of each optical unit of the sensor at different gains was then measured.

The sensor’s down-facing optical unit detects the light reflected from vegetation and converts it into voltage signal U _D, while the up-facing optical unit detects incident light from the sky and converts it into voltage signal U _U. The reflectance is calculated as the ratio ρ = U D U U . Calibration can be performed on the raw reflectance values, as ρ = f ( U D U U ) , where f() is calibration function. Alternatively, the voltages U _D and U _U can be calibrated separately, yielding a reflectance relationship of ρ = f 1 ( U D ) f 2 ( U U ) , where f ₁() and f ₂() are calibration functions. It is also possible to calibrate only one of the voltage signals, resulting in reflectance relationships such as ρ = f 1 ( U D ) U U or ρ = U D f 2 ( U U ) .

An integrating sphere was used to calibrate the VCR sensor’s output voltages, thereby calibrating the reflectance values. By adjusting the voltage of the integrating sphere’s light source, different light intensities were achieved. Under consistent light intensity conditions, the up-facing and down-facing optical windows of the sensor were aligned with the light outlet of the integrating sphere. The output voltage of the sensor was then measured, and the dark current was subtracted. Since the light intensity is uniform, the voltage of the down-facing optical unit in the same band should match that of the up-facing optical unit. Based on this principle, a voltage calibration equation f ( ) was constructed to calibrate the reflectance. The reflectance was calibrated using Equation 4.

where ρ λ is the reflectance at wavelength λ, U_D and G_D are respectively the measured voltage and gain of the down-facing optical unit at wavelength λ, U_U and G_U are the measured voltage and gain, respectively, of the up-facing optical unit at the same wavelength.

The accuracy of the reflectance measurement of the sensor was verified using a gray scale board with a reflectance of 30%. The experiment was conducted on the roof of the School of Life Sciences building at Huaiyin Normal University on August 4, 2024. The sensor was positioned approximately 17 cm above the 20 cm × 20 cm gray scale board. If the distance was greater than 17 cm, the sensor’s FOV would have extended beyond the gray scale board.

To verify the accuracy of the VCR sensor’s reflectance measurements, a total of seven gray scale boards were used as the measurement targets. The nominal reflectance values of the gray calibration panels were 10%, 20%, 25%, 40%, 60%, 80%, and 99%. However, the reflectance of the same gray panel varies across different spectral bands. According to the datasheets provided by the manufacturer, the reflectance values of the seven gray scale boards at the 710 nm wavelength are 11.5%, 18.8%, 24.3%, 40.2%, 61.5%, 79.2%, and 98.9%. At the 870 nm wavelength, the reflectance values are 11.8%, 18.0%, 23.0%, 38.3%, 61.2%, 79.1%, and 98.7%. The experiment was conducted from 14:57 to 15:06 on August 5, 2024, under clear sky conditions, during which the solar altitude decreased from 47.9° to 46.3°. The instrument was located on the roof of the School of Life Sciences building at Huaiyin Normal University. Each measurement was repeated ten times.

On a clear day, the VCR sensor and the ASD Fieldspec 4 spectroradiometer (ASD Inc., Alexandria, VA, USA) were used to simultaneously measure different vegetation canopies to further verify the accuracy of the VCR sensor’s vegetation reflectance measurements. A total of 14 vegetation sites were selected. Sites 1, 2, 4, 8, 10, 11, 12, and 14 contained Imperata cylindrica in varying growth stages or densities. Sites 3 and 9 consisted of Goosegrass, site 5 contained Setaria viridis, site 6 contained Tall Fescue, and sites 7 and 13 contained Clover. The FOV of the VCR sensor and the ASD spectroradiometer were 28° and 25°, respectively. To ensure consistent canopy coverage in both measurements, the VCR sensor and the ASD spectroradiometer’s fiber optic probe were located 40 cm and 48 cm above the measured vegetation, respectively. Given that the FWHM of the VCR sensor filter is 10 nm, the reflectance of the 705–715 nm band measured by the ASD spectroradiometer was averaged to represent the reflectance at 710 nm, and the average of the reflectance in the 865–875 nm band was taken as the reflectance at 870 nm.

To verify the accuracy of the sensor for full-day measurements, the spectral reflectance of Bermuda grass was measured using the VCR sensor from 8:00 to 17:00 under clear sky conditions (with occasional clouds after 15:00) ( Figure 5 ). The solar altitude ranged from 23.0° to 74.5°. The VCR sensor was positioned 50 cm above the vegetation, and measurements were taken once per minute while recording the solar altitude.

The calibration results of the resistance values of the TPL0102s are shown in Figure 6 . As can be seen, the resistance values of the TPL0102 in all of the optical units have a strong linear relationship with the resistor number, with R ² values reaching 0.99999.

The calibration results of the dark currents of the OPT101 in each optical unit are shown in Figure 7 . As illustrated, there is a strong linear relationship between the dark current and the gain of the circuit, with an R ² value of 0.9999.

We used an integrating sphere to calibrate the sensor’s output voltages, thereby achieving reflectance calibration for the VCR sensor. The results of the voltage calibration are shown in Figure 8 . As can be seen, the down-facing output voltage and the up-facing output voltage at the same wavelength exhibit a strong linear relationship, with R ² values of 0.9998 at 710 nm and 0.9999 at 870 nm.

A 30% reflectance gray panel was used as the monitoring target to examine the variations in the reflectance measured by the calibrated VCR sensor throughout the day. Due to limitations imposed by the distance between the sensor and the gray panel, the sensor’s shadow entered the FOV of the 710 nm optical unit between 11:00 and 13:40, while its shadow entered the FOV of the 870 nm optical unit between 10:30 and 13:00. Consequently, these data points were removed during the preprocessing. The results are shown in Figure 9 . As shown, the reflectance measurements show a distinct trend throughout the day, with lower values occurring closer to noon ( Figures 9A, B ). Specifically, the reflectance decreases nonlinearly as the solar elevation angle increases ( Figures 9C, D ). The experimental results indicate that the cosine corrector alone cannot fully eliminate the nonlinear influence of the solar elevation on reflectance measurements made by spectral sensors. Therefore, to improve measurement accuracy, it is necessary to develop a solar elevation correction model.

Sky light can be categorized into two main types: sky-scattered light and direct sunlight. Sky-scattered light is diffuse and not affected by the solar elevation angle. As the solar elevation angle increases, the angle of the direct sunlight entering the sensor’s optical window decreases, allowing more light to pass into the sensor. Under clear weather conditions, the proportion of the direct sunlight in the total sky light can reach 80–90%. However, when the solar elevation angle is zero (the Sun at the horizon), all of the light in the sky is diffuse, with no direct sunlight, making the proportion of direct light 0%. To model the variations in the proportions of direct and diffuse light with changes in the solar elevation, a sine function can be used. To ensure that the proportion of direct sunlight is 90% at a solar elevation angle of 90° and 0% at an elevation of 0°, the proportion of direct sunlight (PD) can be expressed as P D = sin ( 0.7128 θ ) , where θ is the solar elevation angle. The corresponding proportion of diffuse light (PS) is expressed as P S = 1 − sin ( 0.7128 θ ) . This model effectively captures the gradual increase in direct sunlight as the sun rises and the predominance of diffuse light when the sun is low. The voltage value measured by the sensor’s up-facing optical unit (U_U ), which represents the light intensity, can be separated into two components: the voltage converted from the diffuse light, expressed as U U · P S and the voltage converted from the direct sunlight, expressed as U U · P D . The direct sunlight can be decomposed into two vectors. One portion follows a vertically downward path, perpendicular to the optical window of the sensor. This portion of the light enters the sensor, and the measured voltage U U · P D is generated from this light. The other part of the sunlight travels horizontally, parallel to the sensor’s optical window, and does not enter the sensor. Based on the vector decomposition, the intensity of the direct sunlight before decomposition can be determined using the formula U U · P D sin θ . Thus, by taking Equation 4 into consideration, the following solar elevation correction model for reflectance can be derived (Equation 5).

where ρ λ , θ is the reflectance in wavelength λ when the solar elevation is θ (°). The meanings of the other symbols are the same as in Equation 4.

The reflectance measured using the VCR sensor can be corrected using Equation 5 to reduce the impact of the solar altitude on the measurements, enabling all-day operation of the sensor. The corrected results for the data presented in Figure 9 are shown in Figure 10 . It is evident that after correction, the diurnal variations in the reflectance at 710 nm and 870 nm are mostly stable, indicating an excellent correction effect.

The accuracy of the VCR sensor measurements was tested using seven gray boards with known reflectance values. The root mean square error (RMSE), mean absolute error (MAE), and 1:1 plot of the measured versus standard reflectance were used to assess the accuracy of the measurements. The results shown in Figure 11 indicate that the measured reflectance values closely matched those of the gray scale boards. The MAE and RMSE for the 710 nm measurements were 1.07% and 0.63%, respectively, and those for the 870 nm measurements were 0.94% and 0.50%, respectively, demonstrating a high measurement accuracy.

The VCR sensor and ASD spectrometer were used to simultaneously measure the spectral reflectance of vegetation canopies at 14 sites. The results are shown in Figure 12 . The reflectance values at 710 nm and 870 nm measured by the VCR sensor closely matched those measured by the ASD spectroradiometer. The slight deviations observed in some measurements may be due to inconsistency in the vegetation coverage within the fields of view of the VCR sensor and the ASD spectroradiometer.

The VCR sensor was used to measure the reflectance of Bermuda grass throughout the day. The measured reflectance values were corrected for the solar elevation using Equation 5. The coefficients of variation (CVs) of the reflectances at 710 nm and 870 nm were calculated both before and after the solar altitude correction. The results are shown in Figure 13 . As can be seen, after correcting for the solar altitude the reflectance measurements remained relatively stable, effectively eliminating the influence of the solar altitude on the results. The all-day reflectance at 710 nm changed from 12.3–19.2% to 11.1–13.4%, and the CV decreased from 10.86% before correction to 2.93% after correction. For the 870 nm all-day reflectance, the values changed from 41.6–60.3% to 39.0–42.0%, and the CV decreased from 9.69% before correction to 1.53% after correction. It can also be seen from the graph that after 16:00, when the solar elevation angle dropped below 23°, the effectiveness of the solar elevation correction model in mitigating the impact of the solar elevation on the measurements began to decline.