Low-Cost and Convenient Experimental Methods for Research on the Physical Characteristics of Green Manure Seeds

Abstract

To improve the performance of green manure sowing and processing equipment, it is necessary to investigate the physical properties of seeds, including their dimensions, bulk density, and frictional characteristics. Focusing on the main cultivated varieties of green manure in China—milk vetch, hairy vetch, and sesbania—this study proposes three low-cost, convenient, and high-precision experimental methods to measure the geometric size of small seeds. The measured results show that the milk vetch seeds were 2.60–3.40 mm in length, 1.90–2.40 mm in width, and 0.71–0.97 mm in height. The hairy vetch pods were 23.03–32.83 mm in length, 7.39–9.74 mm in width, and 4.06–6.15 mm in height, and the seeds were 3.04–4.10 mm in diameter. The sesbania seeds were 3.81–4.29 mm in length, 1.98–2.37 mm in width, and 1.85–2.08 mm in height. The thousand-seed weights of milk vetch, hairy vetch, and sesbania seeds were 3.40 g, 26.50 g, and 15.58 g, with moisture contents of 7.18%, 9.81%, and 8.73%, respectively. The bulk densities of milk vetch, hairy vetch, and sesbania seeds were 732 g/L, 761 g/L, and 845 g/L. The angles of repose of milk vetch, hairy vetch, and sesbania seeds were 31.66°, 28.15°, and 29.82°, measured using an angle-of-repose tester. Sliding friction angles of 25.85°, 23.55°, and 24.03° for milk vetch, hairy vetch, and sesbania seeds were obtained using a dividing head. These methods and results provide valuable references for the measurement of the physical properties of small-seeded crops (e.g., green manures) as well as other small or irregularly shaped particles.

1. Introduction

Green manure (also known as cover crop) is of great significance in global agricultural development. With the rapid recovery of green manure cultivation in China, the demand for green manure seeds has also increased sharply. Therefore, it is urgent to develop sowing, seed harvesting, and processing equipment that is urgently needed for green manure production. The physical parameters of seeds, including size, bulk density, and frictional characteristics, directly affect the key structural design of processing equipment such as seeders, harvesters, and seed threshing machines, while also providing essential parameters for simulation analysis and a basis for the optimal design of green manure mechanized production equipment [1,2,3,4,5,6,7].

The following research indicates that seed geometric dimensions, thousand-seed weight, angle of friction, and angle of repose are key physical characteristic parameters affecting the design of agricultural equipment. To optimize the design of seed metering devices for millet and wheat seeders, as well as processing equipment for broomcorn millet, existing studies have measured the geometric dimensions, angle of friction, angle of repose, and thousand-seed weight [8,9,10]. To obtain the key structural parameters of precision seeding equipment for vegetable plug trays, existing studies have measured the geometric dimensions, thousand-seed weight, bulk density, porosity, angle of repose, and friction angle of six vegetable seeds. Scholars have conducted specialized research on the angle of repose of granular materials such as rice, soybeans, and transparent sand. Different measurement methods have different impacts on the results, and the particle size also affects the angle of repose. To determine the potential applications of rosehip seeds, existing studies have measured the physicochemical properties and thousand-grain weight of rosehip seeds; however, the testing method used for thousand-grain weight was not reported in their research. To provide a basis for lilac seed sorting, existing studies have measured the geometric dimensions and sliding friction angles of lilac seeds. However, the testing instruments and procedures were complex, leading to low efficiency [11,12,13,14,15,16]. Based on the above analysis, research on the physical properties of agricultural granular materials mainly focuses on grain and vegetables, and the physical properties of green manure seeds have not been systematically studied. In these studies, the geometric dimensions of materials are primarily measured using vernier calipers. Although the measurement method is simple and easy to operate, it is time-consuming and labor-intensive for small seeds. The sliding friction angle measurement of materials is mainly measured using a self-designed experimental apparatus or inclined plane friction tester. The two testing approaches for the sliding friction angle either provide inaccurate measurements or increase costs.

For small-sized green manure seeds, when measuring the thousand-seed weight, inexpensive counting instruments provide insufficient precision and large errors. High-precision equipment is more expensive. For these reasons, this study developed three simple and low-cost or zero-cost testing methods to solve the above problems. Based on the low-cost experimental methods, we systematically measured the seed geometry, thousand-seed weight, angle of repose, and angle of sliding friction of the main plant varieties of green manure, including milk vetch, hairy vetch and sesbania, providing a reference for the measurement of the physical properties of green manure seeds and other small seeds, as well as small particles and irregular particles.

2. Materials and Methods

2.1. Determination of the Moisture Content of Green Manure Seeds

The moisture content of the milk vetch, hairy vetch, and sesbania seeds was determined using the drying method with a DGF30/7-IA electrical heating blast drying oven (Nanjing testing instrument factory, Nanjing, China). The moisture contents of these three seed types were 7.18%, 9.81%, and 8.73%, respectively.

2.2. Geometric Dimension Measurement

2.2.1. Measurement Method

Milk vetch, hairy vetch, and sesbania are the main green manure varieties cultivated in China. Due to their small geometric dimensions, it is time-consuming and laborious to use a vernier caliper for measurements. Therefore, a method for measuring the external dimensions of green manure and other crop seeds based on images has been developed. A flowchart of the method is shown in Figure 1. The system mainly consists of four primary modules: an acquisition module, import module, scaling module, and measurement module. The acquisition module is used to obtain a picture that simultaneously contains the scale image and the seed image. The seeds were directly fixed using glue or other adhesives on the paper with a scale. Alternatively, a steel ruler can be horizontally fixed to the paper with the seeds, and this setup is subsequently scanned to obtain the image information. The scanning effect is more accurate than the photo effect. The import module involves importing the picture obtained using the acquired module into the measurement software. In this study, the scanned image was imported into ZWCAD Mechanical 2024 for direct dimensional measurements. The scaling module is used to scale the acquired image to establish a 1:1 proportional relationship between the measurement scale in ZWCAD Mechanical 2024 and the reference scale (either from the steel ruler or graduated paper) captured in the acquired images. In the measurement module, the measurement function of the software is used to directly measure the seed dimensions in the scaled image, obtaining the length and width data of the seeds. Milk vetch is used as an example, as shown in Figure 1. This method can measure multiple seeds at once, with convenient operation, high accuracy, and no fixture cost investment. The measurement process does not require calculations. It can be repeatedly verified for measurement and is inexpensive.

Seed height can be measured using an electronic height gauge. When measuring the height of seeds, the seeds are directly placed on the platform of the height tester for measurement without using glue, eliminating projection errors, as shown in the high-definition details of the measurement and data reading presented in Figure 2a. The overall measurement is shown in Figure 2.

2.2.2. Statistical Method

Briefly, 100 seeds were randomly selected to measure their length, width, and height and perform frequency analysis. The sampling was repeated three times. Then, the measured dimensions of milk vetch and sesbania seeds were analyzed using SPSS 19.0 statistical analysis software [17,18,19].

The dimensions of hairy vetch pods directly affected the loss rate and impurity rate during seed harvesting, which vary by cultivar and harvest timing. To analyze these effects, we conducted dimensional measurements on 50 mature pods and 150 seeds selected at random during the harvest season. The sampling was repeated five times, and the average value of the sampling results was taken as the final result. All geometric measurements were subject to frequency and statistical analyses using SPSS 19.0 software.

2.3. Measurement of Thousand-Seed Weight and Bulk Density

2.3.1. Measurement of Thousand-Seed Weight

Thousand-seed weight directly affects the design of the seed metering device of the seeder, as well as the quality indicators of the sorting of the harvester and seed sorter. Therefore, measuring the thousand-seed weight of green manure seeds is of great significance for the development of efficient and applicable green manure equipment. When measuring the thousand-seed weight of green manure seeds, such as milk vetch, hairy vetch, and sesbania, the seeds are first counted using a seed counter and then weighed. On the market, microcomputer-based automatic seed counters with weighing function are highly accurate but expensive, whereas vibration-type seed counters are cheaper but less accurate, especially for small green manure seeds. To address this, without increasing production costs, an image recognition method for counting green manure seeds and other granular materials based on artificial intelligence algorithms was developed. After counting the seeds, they are weighed to determine the thousand-seed weight. This can greatly improve labor efficiency and measurement accuracy.

The method for counting green manure seeds and other granular materials is based on artificial intelligence (AI) algorithms and the following methodology. First, photos of green manure seeds and other granular materials are taken with a mobile phone. Then, the photos are imported into a computer, and Python 3.12.5 programming language is used to call the open-source computer vision libraries Open CV (Open Source Computer Vision Library) 4.5.5.64 and NumPy (Numerical Python) 1.21.6 to process the original image using the following pipeline: grayscale conversion → Gaussian blur → binarization → morphological denoising → output results. The processing flow, principles, and precautions are shown in Figure 3.

The key aspects of this method involve scene-specific optimization. During the binarization process, adaptive thresholding is prioritized over complex algorithms. In terms of balancing accuracy and efficiency, the watershed algorithm is bypassed in favor of area-based filtering, thereby reducing computational overhead while maintaining accuracy. Regarding parameter adjustability, critical parameters, such as blockSize and minArea, are designed to be adjustable for diverse imaging conditions. Sesbania seeds are used as an example for counting statistics, as shown in Figure 4.

Because this method only involves counting, a smaller sample size was deliberately used to more clearly illustrate the principle and operation process of the method. In actual experiments, depending on the seed size and the pixel difference of the image-capturing tool, the sample size is unlimited as long as there is no overlap between seeds.

2.3.2. Bulk Density Measurement

The bulk densities of milk vetch, hairy vetch, and sesbania seeds were measured using the GHCS-1000AP electronic grain densitometer (Taizhou Nong’ao grain instrument factory, Taizhou, China). The sampling was repeated five times, and the average value of the sampling results was taken as the final result.

The GHCS-1000AP electronic grain densitometer mainly consists of two cylinders, including a bulk density cylinder and an electronic cylinder; an exhaust block; and a plug. When in operation, the weighing electronic cylinder of the electronic grain densitometer is first calibrated. Then, the plug is inserted into the bulk density cylinder. The exhaust block is placed on top of the bulk density cylinder. The bulk density cylinder is placed with the plug and the exhaust block onto the electronic cylinder used for weighing, setting it to zero. The upper cylinder is filled with green manure seeds. The middle cylinder is placed on the bulk density cylinder. Then, the upper hopper filled with seeds is placed onto the middle cylinder. The switch between the upper cylinder and the middle cylinder is opened, and the seeds in the upper cylinder will naturally fall into the middle cylinder. After all the seeds have fallen, the plug between the middle cylinder and the bulk density cylinder is removed, and the exhaust block along with the seeds will quickly fall into the bulk density cylinder. Then, the plug is inserted back into the bulk density cylinder, and the excess seeds outside the bulk density cylinder are discarded. The bulk density cylinder is reweighed, and the data displayed on the electronic cylinder represent the bulk density of the seeds. The sampling was repeated five times, and the average value of the sampling results was taken as the final result. This instrument can directly display the measurement results, as shown in Figure 5.

2.4. Angle of Repose Measurement

The angle of repose is an important and conveniently measurable indicator of material flowability. Material flowability directly affects seeding stability, distribution uniformity during sowing, the efficiency of seed rubbing, and the quality of seed cleaning and grading. Using this easily measurable physical characteristic of seeds to adjust and optimize key equipment parameters, equipment performance can be improved quickly and accurately, while also providing reference data for simulation model establishment. For this purpose, the angles of repose of milk vetch, hairy vetch, and sesbania seeds were measured using the angle-of-repose tester, as shown in Figure 6. During the experiment, the glass disk was first aligned with the scale on the bottom plane of the measuring instrument. Subsequently, the seeds were poured into the funnel above the instrument. The seeds naturally accumulated within the glass disk of standard dimensions below, forming a cone. The height of the cone was measured. Specifically, it was the directly read value minus the height of the glass disk. By integrating the radius R of the glass disk, the angle of repose φ was computed using trigonometric functions, as follows:

φ = arctan (h/r)

Here, φ, h, and r are shown in the figure. Of note, h is not the directly read value; rather, it is the directly read value minus the height of the glass disc.

2.5. Measurement of the Sliding Friction Angle

The measurement of sliding friction characteristics is essential in agricultural processes, including sowing, harvesting, and seed cleaning, as well as in simulation analyses. It is also necessary to study the friction characteristics of granular materials in the chemical industry. The sliding friction properties are primarily determined by measuring the sliding friction angle of seeds or materials. At present, the custom-built test bench shown in Figure 7a is used more often for measurement. During the experiment, the seeds or materials are placed on a horizontal test board. By shaking the handle, the test board gradually tilts. When the materials slide down, the lifting height of the board is recorded at that moment. Then, the sliding friction angle of the tested seeds or materials is calculated using the tangent value of this height and the length of the board. These test devices are mostly self-assembled frames, which may introduce experimental errors.

To reduce experimental errors and simplify the calculation process, a method for measuring the sliding friction angle of green manure seeds and other granular materials using the indexing head (Figure 7b) is developed. The testing method shares the same principle as the existing inclined plane testing method. The main difference is that it used a new tool, namely, the indexing head, which can precisely measure angles. When this equipment is available in the laboratory, it can be directly used. A test plate and bracket suitable for green manure seeds and other granular materials are manufactured and fixed on the indexing head. The initial data are recorded. Then, the indexing head is rotated. When the seeds begin to slide, the current value of the indexing head is directly read. The difference between this value and the initial value of the dividing head represents the sliding friction angle. The sampling was repeated five times, and the average value of the sampling results was taken as the final result. Different materials such as Q235 steel plates (Nanjing Suyuan Metal Materials Co., Ltd., Nanjing, China) and wood boards can also be installed on the test plate to measure the sliding friction angle of green manure seeds or granular materials on different material surfaces.

3. Results

3.1. Overall Measurements of the Physical Properties of Three Green Manure Seeds

To facilitate the comparison of the geometric dimension differences among different varieties of green manure seeds, the geometric dimensions of the seeds of milk vetch, hairy vetch, and sesbania are statistically summarized in

Table 1. Statistical data on the geometric dimensions of three types of green manure seeds.

Variety	Dimension	N	Mean Value (mm)	Median (mm)	Coefficient of Variation	Minimum Value (mm)	Maximum Value (mm)
Milk vetch seed	Length	300	2.99	3.03	8.36%	2.23	3.52
	Width	300	2.16	2.16	8.80%	1.73	2.68
	Height	300	0.85	0.85	9.41%	0.66	1.08
Hairy vetch pod	Length	150	27.22	26.91	11.68%	22.67	33.6
	Width	150	8.47	8.33	9.67%	6.6	10.32
	Height	150	5.1096	5.12	14.91%	3.81	6.43
Hairy vetch seed	Diameter	360	3.5837	3.6	11.65%	2.52	4.67
Sesbania seed	Length	300	4.04	4.04	4.84%	3.57	4.64
	Width	300	2.17	2.17	6.88%	1.75	2.55
	Height	300	1.96	1.96	4.53%	1.76	2.30

. Their physical properties, including thousand-seed weight and bulk density, are shown in

Table 2. Measurement results of the physical properties of three green manure seeds.

Variety	Physical Characteristic Parameters
	Moisture Content/(%)	Thousand-Seed Weight/(g)	Bulk Density/(g/L)	Angle of Repose/(°)	Angle of Sliding Friction/(°)
Milk vetch	7.18	3.40	732	31.66	25.85
Hairy vetch	9.81	26.50	761	28.15	23.55
Sesbania	8.73	15.58	845	29.82	24.03

.

3.2. Measurement Results of Geometric Dimensions

3.2.1. Measurement Results of the Geometric Dimensions of Milk Vetch

In total, 300 milk vetch seeds were randomly selected to measure their length, width, and height. The measured dimensions of the milk vetch seeds were subject to frequency and statistical analyses using SPSS statistical software. Milk vetch pods and seeds as well as a histogram of geometric dimensions are shown in Figure 8 [20].

Figure 8 shows the dimensional distribution of milk vetch seeds. The frequency distribution characteristics of the milk vetch seed geometric dimensions are relatively concentrated. Table 1 shows that the mean and median of the geometric dimensions of the milk vetch seeds are approximately the same, indicating a symmetrical data distribution. The coefficient of variation for each dimension is less than 10%, suggesting insignificant variation in the geometric dimensions of milk vetch seeds, which is convenient for classification and screening. This provides a reference for the design of the seed metering device for milk vetch sowing and seed harvesting as well as the design of the sieve aperture size for seed classification, screening, and seed rubbing.

3.2.2. Measurement Results of the Geometric Dimensions of Hairy Vetch

Due to the larger size of the pods, which leads to smaller errors, damaged pods were removed, and only intact pods were retained. Therefore, sampling was conducted in three batches, with 50 pods each. The results show that the measured data accurately reflect the geometric morphology of the sampled material. The mature hairy vetch pods were randomly selected during the harvest period for geometric dimension measurements. Additionally, 360 hairy vetch seeds were also selected for length, width, and height measurements. Frequency and statistical analyses of the hairy vetch pod and seed dimensions were conducted using SPSS statistical software. The histograms of the geometric dimensions of the pod and seeds for hairy vetch are shown in Figure 9.

Figure 9 clearly shows the dimensional distribution of hairy vetch pods and seeds. According to the frequency distribution characteristics, the length and height of hairy vetch pods are relatively dispersed, whereas the width distribution is more concentrated. The geometric dimension distribution of the seed diameter is also relatively concentrated. As shown in Table 2, the mean and median of hairy vetch pod and seed diameters are similar, indicating a symmetrical data distribution. Only the coefficient of variation of the pod width is less than 10%, whereas the coefficients of variation for the other geometric dimensions are all greater than 10%. The coefficients of variation are higher than those of milk vetch seeds. However, the hairy vetch seeds are larger than milk vetch seeds, making the hairy vetch seeds easier to classify and screen.

3.2.3. Measurement Results of the Geometric Dimensions of Sesbania

In total, 300 sesbania seeds were randomly selected for length, width, and height measurements. Frequency and statistical analyses of the measured dimensions of the seeds were conducted using SPSS statistical software. The histograms of the geometric dimensions of the seeds for sesbania are shown in Figure 10.

Figure 10 shows the dimensional distributions of sesbania seeds. The frequency distribution characteristics of the seed dimensions show that the geometric dimensional distribution of sesbania seeds is relatively concentrated, and the height and width of the seed are very similar. In the simulation analysis modeling, the sesbania seed can be regarded as a cylinder. Table 1 shows that the mean and median values of the geometric dimensions of sesbania seeds are identical, indicating that the data distribution is symmetrical. The coefficient of variation of each dimension is less than 10%, indicating minimal geometric dimensional differences in the sesbania seed, which is convenient for classification and screening.

3.2.4. Analysis of the Geometric Dimension Measurement Results Using Two Methods

To verify that the image-based method for measuring the geometric dimensions of crop seeds (referred to as the “new method”) is more convenient, accurate, and rapid than the traditional vernier caliper measurement method (referred to as the “traditional method”), 100 sesbania seeds were randomly selected. Their length, width, and thickness were measured using the traditional method followed by the new method. The comparative test results are shown in

Table 3. Analysis of geometric dimension results of Sesbania seeds obtained using different measurement methods.

Method	Dimension	N	Mean (mm)	Standard Deviation	Time Spent (Minute)
Traditional method	Length	100	2.92	0.28	80–100
	Width	100	2.17	0.31
	Height	100	0.91	0.17
New method	Length	100	2.97	0.19	25–30
	Width	100	2.15	0.15
	Height	100	0.85	0.09

.

Given the small geometric size of the seeds, it is necessary to hold the seeds with tweezers when measuring with a vernier caliper. To measure the length, width, and thickness of seeds, it is also necessary to manually keep changing the direction of the seeds with tweezers. As the number of measurements increases, the simple and repetitive measuring actions are prone to random human errors due to differences in estimation deviations and operating techniques among different operators. Thus, the efficiency will be significantly lower than that at the beginning of the measurement. Table 3 shows that the standard deviations of the geometric dimensions of sesbania seeds measured using the new method are all smaller than those measured using the traditional method with a vernier caliper. These results indicate that the new method has greatly reduced the random errors caused by human operations, and the measurement results are more stable.

When the new method was used to measure the geometric dimensions of green manure seeds, the thickness was measured with a thickness tester (Figure 2a). This method still required manually placing the seeds on the platform of the tester. During the entire measurement process, this step was the only step that was time-consuming. However, the thickness measurement results could be directly displayed, which also reduced the estimation deviation caused by human operations. Overall, compared with the traditional method, the new method for measuring the geometric dimensions of green manure seeds improved the efficiency by 3–4 times and greatly reduced the labor intensity.

3.3. Thousand-Seed Weight and Bulk Density Measurement Results for Three Types of Green Manure Seeds

The thousand-seed weight and bulk density of seeds are closely related to moisture content. When the moisture contents of milk vetch, hairy vetch, and sesbania seeds were 7.18%, 9.81%, and 8.73%, respectively, the thousand-seed weights of milk vetch, hairy vetch, and sesbania were 3.40 g, 26.50 g, and 15.58 g, respectively, using an AI-based image recognition method for counting green manure seeds and other granular materials.

This study developed a new method for counting based on artificial intelligence image recognition that could also measure thousand-seed weight. To verify the advantages of this counting method and using sesbania seeds as an example, 1000 seeds were randomly selected each time, with five repetitions. The counting and weighing results are shown in

Table 4. Comparative analysis of different counting methods using sesbania seeds.

Method	N	Mean (g)	Standard Deviation	Time Spent (Second)
Manual counting	1000	14.97	1.12	50–75
AI counting	1000	15.58	0.49	<8

. The standard deviation of the manual counting results is greater than that of the artificial intelligence counting method, indicating more random errors in manual measurement. In addition, manual measurement is more time-consuming. The efficiency advantage increased further for smaller seeds, and manual counting had a higher error probability.

Using 5 samples with the same moisture content and calculating the average, the bulk densities of milk vetch, hairy vetch, and sesbania seeds were 732 g/L, 761 g/L, and 845 g/L, respectively.

3.4. Repose Angle Measurement Results for Three Green Manure Seeds

The repose angle of seeds is closely related to their moisture content. When the moisture contents of milk vetch, hairy vetch, and sesbania seeds were 7.18%, 9.81%, and 8.73%, respectively, the measured angles of repose were 31.66°, 28.15°, and 29.82°, respectively.

3.5. Sliding Friction Angle Measurement Results for Three Green Manure Seeds

Based on the principle that the indexing head can measure angles conveniently and quickly, a new application of the indexing head has been discovered. Specifically, it can directly measure the sliding friction angle of seeds. To verify the advantages of this method, a comparative experiment between the two methods was conducted using hairy vetch seeds as an example. The results showed that the standard deviation of repeated measurements of the seed sliding friction angle using the indexing head was 0.16, whereas that measured using the traditional inclined plane method was 0.43. This finding indicates that the new method provides more stable measurements and reduces the experimental errors caused by estimation and subsequent calculations with the traditional method.

The sliding friction angle is closely related to seed moisture content. When the moisture contents of the seeds of milk vetch, hairy vetch, and sesbania seeds were 7.18%, 9.81%, and 8.73%, respectively, the measured sliding friction angles of the milk vetch, hairy vetch, and sesbania seeds were 25.85°, 23.55°, and 24.03°, respectively.

4. Discussion

This study systematically investigated critical physical properties of three major green manure varieties (milk vetch, hairy vetch, and sesbania), including seed geometric dimensions, thousand-seed weight, bulk density, angle of repose, and sliding friction angle. These parameters are critically relevant for the design of mechanized equipment used in green manure seeding, seed harvesting, and seed processing operations. This research fills the gap in basic research on the material properties of green manure seeds for mechanical production, providing methods and ideas for future research on the physical properties of seeds of other green manure varieties. In addition, it provides a theoretical basis for the research and development of key components and the parameter design of mechanical production equipment for milk vetch, hairy vetch, and sesbania seeds.

This study innovatively developed a method for measuring the geometric dimensions of green manure seeds, especially small granular materials. It improved the time-consuming and labor-intensive problems associated with measuring geometric dimensions using vernier caliper.

This study innovatively developed an image recognition method for counting green manure seeds based on artificial intelligence. Currently, the market offers automatic seed analysis instruments that can directly recognize the geometric dimensions of granular materials and the thousand-seed weight of seeds using images. However, these instruments are expensive, causing economic waste for individuals or laboratories with low usage frequencies. There are also cheaper vibrating seed counters, but they have low accuracy and large errors and are especially unsuitable for measuring small green manure seeds. Because this method is exclusively for counting, if the contrast between the seeds and the background is insignificant, the accuracy of image recognition under strong light conditions or a non-uniform background may decrease. The method can be further developed to measure seed size simultaneously with counting, but measurement errors may occur due to shooting angle issues. How to eliminate such errors will be the next research topic investigated.

Based on the principle that the indexing head can conveniently and quickly measure angles, the sliding friction angle of seeds and granular materials can be calculated conveniently, accurately, and quickly by making a seeding fixture. This method is applicable for laboratories in universities or research institutes and agricultural machinery equipment workshops of agricultural machinery enterprises that use an indexing head.

The shape of seeds significantly affects their physical properties, including the angle of repose and sliding friction angle. Milk vetch seeds are reniform or obovate in shape, with flattened sides. Hairy vetch seeds are spherical in shape. Sesbania seeds are flat, short, and cylindrical; these seeds can be approximately regarded as cuboids. In the research on the physical properties of the crops mentioned in this paper, eggplant seeds are reniform or round and mostly flattened; these seeds are most similar to milk vetch seeds. Soybean seeds and pakchoi seeds are round or elliptical, but pakchoi seeds are much smaller in diameter than hairy vetch seeds. Therefore, soybeans are more similar to vetch seeds. Sesbania seeds are relatively special, and there are no seeds with similar features in the mentioned research. The angle of repose and sliding friction angle of eggplant seeds are 35.90° and 26.60°, respectively, while those of milk vetch seeds are 31.66° and 25.85°, respectively. The angle of repose of soybeans is 21.14°. This may be because the shape of eggplant seeds is rounder and fuller than that of milk vetch seeds, and soybean seeds are closer to spheres than hairy vetch seeds. This further verifies the scientific validity of this research and its relevance in the field of agricultural engineering.

The size of the seed groove in the seed metering device should be adjusted during sowing based on the sizes of milk vetch, hairy vetch, and sesbania seeds to achieve uniform seeding and reduce the seed usage per unit area. In addition, the size of the cleaning sieve pores during seed harvesting and cleaning can be designed based on the geometric dimensions of the seeds to reduce the mechanical harvesting loss rate.

The milk vetch, hairy vetch, and sesbania seeds were all provided by a green manure seed company and belonged to the same batch. Therefore, all experiments were not repeated under different humidity conditions in this paper. However, it is possible that seed moisture content impacts physical properties of seeds, such as the angle of repose and bulk density. Based on previous studies on other crop seeds, the greater the moisture content is, the larger the geometric dimensions, 1000-seed weight, angle of repose, and sliding friction angle of the seeds are [10,21]. In contrast, the bulk density of seeds decreases as the moisture content increases [21,22,23]. In the future, a systematic study will be conducted specifically on the impact of moisture content changes on the physical properties of seeds.

The most important feature of the proposed experimental methods is that they do not increase equipment costs and are not limited by time or location. These methods can be widely used by individuals or in laboratories and other groups and only require a camera or a smart phone with a camera function. Subsequently, more efficient and accurate testing methods suitable for large laboratories can be systematically explored.

With the development of AI technology, AI will better serve production demands and improve the lives of people. How to apply AI to agricultural production is an important future research topic.

5. Conclusions

In summary, in view of the time-consuming and labor-intensive nature of widely used methods for measuring the geometric dimensions of small seeds, such as green manure seeds, and the poor accuracy of thousand-seed weight measurement, it is important to develop a low-cost, high-precision, and user-friendly system for measuring the geometric dimensions of small seeds based on image recognition. The thousand-seed weight of green manure seeds was measured using a counting method based on an artificial intelligence algorithm for image recognition.

The milk vetch seeds were 2.6–3.4 mm in length, 1.9–2.4 mm in width, and 0.71–0.97 mm in height. The diameter of hairy vetch seeds was 3.0–4.1 mm. When the moisture content of milk vetch, hairy vetch, and sesbania seeds was 7.18%, 9.81%, and 8.73%, respectively, the thousand-seed weights were 3.40 g, 26.50 g, and 15.58 g, respectively. The bulk densities were 732 g/L, 761 g/L, and 845 g/L, respectively. The angles of repose were 31.66°, 28.15°, and 29.82°, respectively, and the sliding friction angles were 25.85°, 23.55°, and 24.03°, respectively.

Thus, the advantages of open-source artificial intelligence algorithms can be fully leveraged to develop more convenient, accurate, and low-cost physical property measurement methods for green manure seeds and other granular materials.