Agriculture has an important role both in providing food for society and in making an important contribution to the country's economy. At the same time, agriculture is an important source of livelihood in most parts of the world. Grains have been used as the main food source of humans for thousands of years and have contributed greatly to the development of our civilization. Almost all over the world, corn, rice, sorghum, wheat, rye, and millet are important food sources in people's daily lives. In addition to its vital importance, grains are also used in baby foods, confectionery products, processed meats, batters, sweeteners, thickeners, coatings, and beverages such as juice. The three most important food crops in the world are rice, wheat, and maize (Awika, 2011). The main stages of cereal production are tillage, sowing, maintenance, and harvesting.

There are various techniques for seeding cover crops, including universal seed drills, precision seed drills, and no-till drilling. Each method has its own set of advantages and limitations. The seed drill is a key component in developing successful conservation plantings. It is a complex machine that delivers seeds at a metered rate, places them at a consistent depth in the soil, and produces light compaction to provide good seed to soil contact. One of the most important parts is the seed metering mechanism. The ultimate objective of the seed metering mechanism is to achieve precise seed distribution within the row (Boydas & Turgut, 2007). There are many different seed metering mechanisms. Generally, volumetric seed metering units such as parallel fluted feed rollers, studded feed rollers, and helical fluted feed rollers are used for sowing cereals (Collins et al., 1976). Volumetric flow metering mechanisms send the seeds in a continuous flow in furrows at a uniform rate. The seed rate is varied by adjusting the gear ratio between the seed metering unit and the seed drill ground wheels for a studded feed roller, and adjusting the gate or shifting the fluted wheel sideways for fluted feed type mechanisms. The seed rate does not change with travel speed. However, the mass of the flowing seed can be increased or decreased depending on the ground of the seed drill machine. Seed flowing through the metering mechanism must be monitored for maintaining the recommended plant spacing and population (Nandede et al., 2012). It is easier to determine the spacing of large grain seeds such as corn, bean, and peas and to evaluate the seed metering mechanism according to small seeds such as wheat, rice, and rye in the laboratory and in the field. Because the seed metering systems used in grain seed drills are delivered the seeds into furrow lines in a continuous flow. The difficulty in designing a measuring of the flow of small seeds has been that the seeds flow in clumps (Al-Mallahi & Kataoka, 2016). The seed metering mechanism is evaluated either in the field or in the laboratory. Evaluation done manually in the field is a very tedious process (Nandede et al., 2012). Therefore, researchers are using easier and faster methods to evaluate seed metering mechanisms in the laboratory. Some of the researchers used a sticky belt test stand (Collins et al., 1976; Kelly & Gould, 1995; Liu et al., 2018; Yin et al., 2018), optoelectronic sensor (Griepentrorong, 1994; Kocher et al., 1998; Lan et al., 1999; Panning et al., 2000; Al-Mallahi & Kataoka, 2013), acoustic technique (Karimi et al., 2015), high-speed camera (Karayel et al., 2006; Cakir et al., 2016), digital scale (Boydas & Turgut, 2007; Al-Mallahi & Kataoka, 2013). Each of these measurement methods has advantages and disadvantages. The sticky belt is the most common method used to evaluate seed metering mechanisms. However, it is a labor-intensive procedure. Karayel et al. (2006) used a high-speed camera system to evaluate the spacing uniformity of the wheat and soybean seeds. They measured the distribution uniformity of seeds with 96% accuracy. They stated that the high-speed camera system did not miss any wheat and soybean seeds. But, the high cost of high-speed cameras is a disadvantage. Both seed flow rate and flow evenness of volumetric flow metering mechanisms are effectively measured with a digital scale. Nevertheless, none of these measurement methods are suitable for using in the field. It is very important for tractor operators and researchers to monitor the workability of the metering unit while working in the field. Therefore, studies are carried out on the development of methods to be used both in the field and in the laboratory. Researchers such as Grift et al. (2001), Nandede et al. (2012), and Al-Mallahi & Kataoka (2016) developed measurement systems by using an optoelectronic sensor to evaluate seed metering units both in the field and in the laboratory. Al-Mallahi & Kataoka (2016) stated that the overall error in estimating flow rates for rye seeds at all speeds was 5.3%, and the sensor can be used on the grain drill for estimating seeds at variable sowing rates within certain speed limits. Li et al. (2015) used electrostatic and capacitance sensors to measure the mass flow rate of glass balls. They determined that the measurement error of the mass flow rate of glass beads within the range of 0.13 kg s^-1-0.9 kg s^-1 is -3%-8%.

The aim of this research was to develop a low cost measurement system for a low volumetric flow that could be used both in the field and in the laboratory. The microwave Doppler sensor was used for this. The microwave Doppler sensor appears to have some advantages compared to the other sensors described above. The microwave Doppler sensors have small dimensions, simplicity, low cost, and easy installation (Yan, 1996). So, it will be easier to place and adjust in a planter. In addition, signals reflected from stationary particles do not create the Doppler Effect, and thus measurement errors can be prevented (Pang et al., 2018). Yang et al. (2020) developed a granular fertilizer mass flow measurement system based on the microwave Doppler method. They built a linear regression model of sensor output value and fertilizer mass flow by using the least squares method. They stated that the relative errors of the microwave Doppler method were not more than 5%, and the measurement system can be installed on the fertilizing machine. Pang et al. (2018) used the microwave Doppler measurement method to determine the mass flow rate of quartz sand particles of different diameters. They investigated the relationship between the solid mass flow rate and the output signal of the microwave Doppler measurement system. They stated that the microwave Doppler measurement method can be applied to other solid particles with different sizes, and unsteady solid flow as well as steady solid flow.

In this study, the cumulative seed weights flowing from the studded feed roller on the scale were subtracted from each other in unit time intervals and the number of seeds in the unit time interval was determined. Change in the number of seeds in the unit time interval at a flow rate of 3.20 g s^-1 is shown in Figure 8 as an example. The voltage values measured from the microwave sensor were simultaneously recorded on the computer. The voltage variation obtained from the microwave sensor at a flow rate of 3.20 g s^-1 is shown in Figure 9 as an example. Then, the mean, standard deviation, and coefficients of variation of these obtained values were calculated. These values were determined for each flow rate considered in the study and are given in Table 3. Correlation analysis was performed to determine the relationship between the mean values obtained from the balance and the mean values obtained from the microwave sensor. Two correlation coefficients were used; the Pearson and Spearman correlation coefficients, to measure linear and nonlinear relationships, respectively. The results of the correlation analysis of the mean values obtained from the scales and the microwave sensor in the study are given in Table 4.

The Pearson correlation coefficient was 0.961. Also, the Pearson correlation coefficient is significant, and the p-value is less than 0.001. There was a strong positive correlation between scales mean values and microwave sensor mean values. Thus, the Pearson correlation coefficients confirm that the scales mean value and microwave sensor mean value were correlated strongly and positively. Regression analysis is widely used to estimate the relationship between two variables. When the independent variable is changed, approximate values of the dependent variables can be found. Thus, with the regression analysis, it can be derived a mathematical equation for the values obtained from the microwave sensor with the help of the mean values obtained from the scales. It was found that the coefficient of determination (R²) was 0.974 (Table 5). So, mean 97.4% variation in the dependent variable (scales mean value) could be explained by the independent variable (microwave sensor mean value). The Durbin-Watson value was found as 1.761. Since this value was between 1.5 and 2.5, it shows that there was no autocorrelation in the residual.

The ANOVA test was performed to determine the importance of the relationship between variables (Table 6). The ANOVA test results showed that there was a linear relationship between the variables, and this relationship was highly significant (F1,14=171.188, p<0.001). To determine the regression equation, the regression coefficients were given in Table 7. Thus, the following equation 1 was obtained with the help of the regression coefficients.

where, y is scales mean value (g s^-1), x is microwave sensor mean value (mV)

The regression coefficient was significant (p<0.001). The regression slope was 0.066. It was found that both the intersection and the variable were highly significant with the t-test (p<0.05). The scatterplot, the regression line, and the regression equation were showed in Figure 10.

The difference between the actual values in the data and the predicted values from the model is called the residual value. If the model does not fit well, the residual values will be large. Large residual values indicate that we have an outlier. Residual statistical values were given in Table 8. The Cook’s distance and Mahalanobis distance give us information about outlier. The Cook’s distance must be less than 1, and the Mahalanobis distance must be less than 16.27. Looking at Table 8, the maximum Cook’s distance value was 0.684, and the maximum Mahalanobis distance value was 3.069. Both values were found to be appropriate. In addition, the errors must have a normal distribution. Therefore, a histogram of the residuals was plotted to check normality (Figure 11). Looking at Figure 11, It was observed that the histogram exhibited a normal distribution. Homoscedasticity is important in linear regression models. This was shown in Figure 12. The scatter plot showed residual versus predicted value. There should be no pattern in the scatter. The residual value against the predicted value was fit to the linear regression model. The width of the scatter remained roughly the same as predicted values increased, so the assumption was met. In particular, it was observed that these sensors should be well calibrated. Similarly, Väyrynen et al. (2013) tested four different noncontact measurement methods, and found that all measurement systems provided good results. However, they emphasized that calibration equations, parameters, and measurement noise levels needed to be considered. Same way, Swisher et al. (2002) measured the granular fertilizer flow rate by using an optical sensor. They detected that for good accuracy, the measurement system should be calibrated separately for each fertilizer.

Determining the flow rate of agricultural products such as grain is of great importance for sustainable precision agriculture. In laboratory conditions, the sensitivity of seed metering units can be determined by measuring methods such as precision scales or adhesive tape conveyors. However, these measurement systems are not capable of measuring the instantaneous amount of seed released from a seed drill operating in the field. Instead, it is necessary to use measurement systems that will measure the seed flow amount without contacting the seed. Thus, with the instant intervention to be made in the field, both the amount of seeds and energy usage can be reduced. In this study, the seed flow rate was tried to be estimated with a low-cost microwave Doppler sensor. For this, a seed drill simulator prepared in the laboratory was used. A studded feed roller metering unit was used in the seed drill simulator. In the experiment, 16 different seed rates of wheat were taken into consideration. these mass flow values ranged between 1.50 and 11,91 g s^-1. The seed flow rate was measured cumulatively with a precision balance placed under the seed tube. Meanwhile, the measurement was made with the microwave Doppler sensor placed in the seed tube at an angle of 45 degrees. The velocity of the seeds passing through the detection area of the microwave Doppler sensor and the seed density caused a change in the microwave Doppler sensor output frequency. These obtained frequencies were converted into voltage with the help of an electronic card using the LM2917 integrated circuit. Thus, a relationship was established between the mass flow rate obtained from the precision balance and the voltage values obtained from the microwave Doppler sensor. Some statistical methods were established to describe this relationship. Correlation and regression analysis were applied to the obtained measurements with the help of the SPSS statistics program. The Pearson correlation coefficient was significant (p<0.001). Results show that there was a strong positive correlation between scales mean values and microwave sensor mean values. According to ANOVA regression analysis, there was a linear relationship between the scales mean values and the microwave Doppler sensor mean values. This relationship was highly significant (p<0.001). R squared was 0.974. Thus, 97.4% of the variability in microwave sensor values was explained by the linear model. The other 2.6% could be explained by other variables. The regression equation y=0.0662x-1.388 was found to estimate the mass flow rate from the voltage values obtained from the microwave Doppler sensor. The results of the research showed that the microwave sensor can be used to measure the seed flow rate in seed drills. This can potentially offer benefits such as non-contact measurement and real-time monitoring. However, it is crucial to ensure proper calibration of the sensors for accurate measurements. In order to obtain reliable results in similar studies, it is of great importance to consider calibration equations, parameters, measurement noise levels, and different seeds.