Intelligent control system of agricultural unmanned tractor tillage trajectory

Abstract

As the large error and low efficiency of manual tillage, tractor tillage has become the focus of the current research. However, tractor driving requires a lot of manpower and time. Therefore, the overall efficiency of agricultural tillage is reduced. To address this problem, an intelligent control system of agricultural unmanned tractor tillage trajectory based on machine vision is proposed in this paper. First, the hardware of the automatic control system of agricultural unmanned tractor is designed. Then the intelligent control algorithm is proposed to realize the software design, and the image of tractor tillage environment is processed. Finally, the adaptive path tracking algorithm is applied to realize the adaptive tracking of the agricultural unmanned tractor tillage trajectory, so as to realize the trajectory control. Experimental results show that the designed system can accurately track and control the agricultural unmanned tractor tillage trajectory and improve tractor operation efficiency.

Keywords

Machine vision unmanned agricultural tractor intelligent control algorithm

1 Introduction

With the rapid development of the national economy, more and more rural population has moved to cities [1]. Since 21st Century, the problem of population aging has become more and more serious in China, which means that agricultural labor will be short. At present, there is the trend of traditional farming replaced by agricultural robot operation. Intelligent machines can effectively solve the problem of the reduction of the number of agricultural laboring population and the agricultural and sideline products, which has an indispensable role in today’s society [4, 13]. In addition, in the process of agricultural production, farmers should not only be exposed to the sun, but also to drive agricultural machinery continuously, which is a very boring and monotonous job for the drivers. The driver cannot maintain a high concentration of spirit for a long time, which will reduce the operation efficiency of agricultural machinery and increase the phenomenon of missed broadcasting [2]. The application of intelligent control system for tillage trajectory [3] in agricultural machinery can make drivers out of boring and monotonous working environment, realize the intelligentization of agricultural machinery, and improve the working conditions of farmers. Combined with accurate agricultural technology, the possibility of repetitive operations is reduced [13, 15].

In agricultural machinery, agricultural unmanned tractor is an important tool to reduce the manual labor of farmers [8]. It takes the seedling of agricultural products as the object of operation, and has the ability of partial perception and action of human beings, so that semi-intelligent transplanting can be realized [5, 6]. The agricultural unmanned tractor used in the intelligent control system for tillage trajectory must first determine the position and direction of the tractor. It is necessary to use trajectory intelligent control system to determine walking direction. Trajectory intelligent control system is developed by the technologies of electronic communication, computer, and control [12]. It is mainly according to sensors to detect the pose information of the agricultural unmanned tractor, and then use the acquired pose information to plan and avoid obstacles through the intelligent system, and control the steering mechanism of agricultural unmanned tractor. Intelligent navigation is achieved through the path of planning, thus realizing the intelligent control of agricultural unmanned tractor tillage trajectory [9, 10]. The development of agriculture has become the most important priority in today’s society. Many experts and scholars are researching the design of agricultural machinery intelligent system, especially the intelligent control system of agricultural unmanned tractor tillage trajectory [7].

In the reference [14], a new intelligent navigation and control system for unmanned tractor based on PC embedded technology and Internet of things technology is proposed. With embedded PC processor as the core, combined with Linux embedded system and Internet of things technology, the software and hardware of the intelligent unmanned tractor control system is designed. Based on PC technology, autonomous navigation ability of unmanned tractor was tested. The test results show that by testing the cooperative working ability of multiple unmanned tractors, the autonomous navigation curve of unmanned tractor is in good agreement with the preset navigation curve. The positioning error is small and the speed of error convergence is fast. Based on the Internet of things technology, the cooperative operation ability of multiple tractors was tested. The test results show that the positioning accuracy of 3 unmanned tractors is higher and not more than 0.06 m. However, this method is too complicated to operate, and the cost of technology is high. In the reference [6], a new control method for unmanned ground vehicles based on robust trajectory tracking is proposed. The trajectory tracking error model is used to design the linear model predictive controller, and the feed forward control action and robust control are combined. Experimental results show that the proposed method can reduce the error between the tractor trailer system and the target trajectory according to the straight and curve linear track. However, this error has a certain error and cannot be able to carry out precise guidance. In the reference [11], in order to apply the intelligent vehicle terminal to large tractors, the tractor CAN bus intelligent terminal technology is studied. The vehicle intelligent terminal system is composed of agricultural machinery CAN bus data parser and on-board computer. The CAN bus data parser of agricultural machinery is used to analyze CAN data generated by tractor operation according to ISO11783 agricultural machinery CAN bus protocol. The application program of on-board computer is programmed with VB6.0 software. The program can process, store, transmit, and display the parsed CAN data. Vehicle intelligent terminal has functions of real-time monitoring of tractor operation parameters, operation track display, and virtual instrument. It can assist in driving and precision management. Although this method guarantees accuracy, it is expensive and difficult to popularize.

For the above problems, a design method of intelligent control system of agricultural unmanned tractor tillage trajectory is proposed in this paper. First of all, the hardware structure of the design of agricultural unmanned tractor sub-control system is the wireless transmission of land environment image, the visual navigation of agricultural unmanned tractor upper computer, the construction of wireless transmission link, the signal processing of agricultural unmanned tractor lower computer, the pneumatic execution module of agricultural unmanned tractor and the path tracking module of agricultural unmanned tractor are six modules to complete the hardware design of the whole system. The method of image processing is modified by computer. The low-pass filtering and Sobel edge detection algorithms are transferred to the field programmable gate array, the farming environment images are properly processed, and the control based on the intelligent algorithm is designed System software, complete the intelligent control system design of agricultural unmanned tractor farming track, this method has high feasibility.

2 Hardware design of intelligent control system for agricultural unmanned tractor

The control system of agricultural unmanned tractor based on machine vision mainly includes the hardware design and the software design by intelligent control algorithm.

2.1 Structure of intelligent control system for agricultural unmanned tractor

The hardware of the system can play a great role in the process of intelligentization of the control system of agricultural unmanned tractor. Figure 1 shows the hardware structure of main control system of agricultural unmanned tractor.

Fig. 1

Hardware structure of main control system of agricultural unmanned tractor.

From Fig. 1, it can be seen that, it the hardware structure of the main control system of agricultural unmanned tractor, the intelligent control system of the entire agricultural unmanned tractor is controlled by the main control center. The power-supply module is used to make the whole system run. If it is needed to farm regularly, the alarm clock is set in the clock module. The image information data receiving module converts the exploration results of the land or the surrounding environment into the form of images. The information data transmission module is transmitted to the communication module, and the exploration results are transmitted to the control center by the communication module, so that the control center can consider the overall demand of the cultivation and reduce the control error. In the intelligent control of agricultural unmanned tractor, the path planning and tracking of agricultural unmanned tractor is very important. It can make agricultural unmanned tractor avoid obstacle intelligently and improve the working efficiency of agricultural unmanned tractor. Figure 2 shows the hardware structure of the sub-control system of agricultural unmanned tractor.

Fig. 2

Hardware structure of the sub-control system of agricultural unmanned tractor.

The agricultural unmanned tractor sub-control system is designed as 6 modules, aiming at these 6 modules to realize the specific functions of each module.

2.1.1 Wireless transmission of land environment image

The wireless transmission of the land environment image of the agricultural unmanned tractor intelligent control system is mainly transmitted by the wireless local area network built by the image acquisition end and the image receiver, and realized by the principle of the wireless bridge. The hardware of wireless transmission is mainly composed of AUS2405 front-end bridge and AUS2408 back-end bridge. The two wireless bridges use the IEEE802.11b protocol. The construction of image wireless transmission links is mainly based on the built-in IP of the Middle front end bridge of the lower computer. Since the back-end bridge is directly connected to the monitoring center of the system, the image signal collected by the camera part of the lower computer of the agricultural unmanned tractor is compressed to the system monitoring center. The code is eventually transferred to the system monitoring center.

2.1.2 Visual navigation of upper computer of agricultural unmanned tractor

The upper computer vision navigation system monitors the center host. Through the system monitoring platform, the collected compressed image information is displayed on the platform. According to the final result of the image processing, the monitoring host is to determine the path of agricultural unmanned tractor tillage, so that the subsequent calculation of the deviation of the tractor tillage angle is carried out smoothly. According to fuzzy control method, three parameter values are analyzed. The motion control commands are transmitted to single-chip microcomputer control system. Thus, the driving system and terminal actuator of the agricultural unmanned tractor will complete the tillage operation corresponding to the signal transmission.

2.1.3 Construction of wireless transmission link

In the whole process of wireless signal transmission, the image transmission of the agricultural unmanned tractor and transmission network are independent of each other. The process is accomplished by wireless transmission of the signal by the upper computer and the wireless reception of the image signal by the lower computer. According to the transmission of the image signal of the upper computer connected to the RS232 serial port by the monitoring host, and receiving the commands issued by the monitoring platform, the monitor command is transmitted to the lower computer. Because the single-chip microcomputer control signal module is directly connected to the single receiving module of the lower computer, the lower computer can process the receiving command by the single-chip microcomputer. The transceiver integration process is completed through LSDRF4710M01 wireless transceiver. It has the advantages of strong anti-interference ability and long transmission distance, which can meet the needs of intelligent control system of agricultural unmanned tractor. With the transmission of the upper and lower computer and the connection between the receiving module and the power module, the construction of the wireless transmission link is achieved.

2.1.4 Signal processing of lower computer of agricultural unmanned tractor

The core of the control signal processing system of the lower computer of the agricultural unmanned tractor is the FreescalekS12xs controller with 3.3 V or 5 V power supply. In the FreescalekS12xs controller, TXD and RXD pins are intersecting and connected to TXD and RXD pins in the wireless reception. Other pins control the switch of 5 relays. When the controller receives the control signal of wireless reception and wireless transmission, it responds to the interrupt processing program to control the switch of the relay. The corresponding action of agricultural unmanned tractor is controlled by pneumatic execution system [16, 17].

2.1.5 Pneumatic execution module of agricultural unmanned tractor

If we want to complete the operation of agricultural unmanned tractors in the field, we need to remotely control all kinds of control actions of the system. The pneumatic execution system is used to control the corresponding actions of agricultural unmanned tractor. The pneumatic execution system of the lower computer is composed of a pneumatic system and an execution mechanism at the end. Pneumatic system includes micro air pump, water separator, cylinder, and related pneumatic accessories.

The switch of the solenoid valve can be controlled by the switch of the relay, and further controls the expansion process of the cylinder piston. This is because the solenoid valve is connected to all relays and connected with its corresponding cylinder. The pulling action of the control cables in the main clutch and the left and right clutch of the agricultural unmanned tractor is completed. The pulling force is based on the execution mechanism of the end of the agricultural unmanned tractor, so as to complete a series of manipulation actions of agricultural unmanned tractor.

Through the above process, the construction of pneumatic execution module of agricultural driverless tractor is completed.

2.1.6 Path tracking module of agricultural unmanned tractor

The purpose of the path tracking is to determine the angle of the agricultural unmanned tractor, or the angle of the target, according to the current pose of the agricultural unmanned tractor.

Acquisition of the pose of agricultural unmanned tractor, the acquisition of the pose of agricultural unmanned tractor refers to the use of sensors to collect real-time information on the tractor’s current position, posture, and motion, including the latitude and longitude, the direction angle, the pitch angle, and the speed. Then the deviation of the distance between the agricultural unmanned tractor and the predefined path position is calculated. For the intelligent navigation path algorithm, the distance between the control point of agricultural unmanned tractor and predefined routes and the course deviation of agricultural unmanned tractor. The lateral deviation is the distance from the control point of the agricultural unmanned tractor to the predefined route, and the course deviation reflects the degree of deviation between the agricultural unmanned tractor and the ideal course.

GPS module of agricultural unmanned tractor, in the proposed method, localization of agricultural unmanned tractor is achieved by using the most common GPS. The positioning precision of centimeter level can be obtained by RTK, which can meet the demand of intelligent navigation for agricultural unmanned tractor. Figure 3 shows the GPS parameters in the agricultural unmanned tractor.

Fig.3

GPS parameters in the agricultural unmanned tractor.

In Fig. 3, each index corresponds to the corresponding parameters, which clearly shows that the GPS can meet the high precision intelligent navigation needs of agricultural unmanned tractor.

Through the design of the hardware structure of the intelligent control system of the agricultural unmanned tractor, it provides a feasible basis for the design of the unmanned control system of the tractor. On this basis, the system software design is carried out.

2.2 Software design of agricultural unmanned tractor control system based on intelligent control algorithm

On the basis of the information in Section 2.1, the processing of agricultural unmanned tractor environment image refers to the process of transforming the above land environment, that is, the information reflected by the image of the tillage environment, into data information, and then processing through the computer. In this paper, the method of image processing is modified. The algorithm of low pass filtering and Sobel edge detection is transferred to the field programmable gate array, the image of the tillage environment is properly processed, and the software of the control system based on intelligent algorithm is designed.

2.2.1 Noise reduction of tillage environment image

In this paper, low pass filter is used for denoising of tillage environment image. For a tillage land environment image, the jump part, the noise part, and the edge part represent the high frequency component of the image, and the large-scale background represents the low frequency component. Therefore, the low-pass filtering can effectively eliminate the noise in the image. Some details in the image are in the high frequency region, while low-pass filtering will sharpen the details of the image. Low pass filtering process is: through Fourier transform, low-pass filtering is carried out, and then the filtered tillage environment image is obtained with Fourier inverse transform.

Assume F (u, v) is the Fourier transform of the original tillage land environment image with noise f (x, y), H (u, v) is the transfer function of low pass filter. Then the signal G (u, v) after lower-pass filter is expressed as $G (u, v) = H (u, v) . F (u, v)$ (1)

Where G (u, v) is the obtained filtered tillage environment image after Fourier inverse transform.

2.2.2 Image segmentation of tillage environment based on Sobel edge detection

For image edge detection, every pixel of the image should be calculated, and the computation is very heavy. Compared with other detection algorithms, Sobel edge detection has lower computational complexity and better detection effect. This algorithm is a first order derivative edge algorithm. According to the direction and template of direction gradient, the convolution is completed with the selected pixel template, and then the center point of pixel template is the pixel to be found. The sum of the square of the calculated direction and the gradient value of the direction is carried out, and then the square root is obtained. Compared with the set threshold, the point greater than the threshold is called the edge point, and the point smaller than the threshold is called the common point.

Assume X is the horizontal direction matrix template in the Sobel convolution template, Y is the vertical direction matrix template in the Sobel convolution template, and the remaining one is the image pixel gray value of the 3 × 3 template. The horizontal direction matrix template and the vertical direction matrix template are used to calculate the plane convolution of the environmental pixel template of the agricultural unmanned tractor, so that the difference approximation of the brightness in the horizontal direction and the vertical direction can be obtained. Assume P is the tillage environment pixel, GX is the horizontal gradient of the image, and GY is the vertical gradient of the image, then the horizontal gradient and the vertical gradient is calculated by using. $GX = X * P$ (2) $GY = Y * P$ (3)

In the above image, the gray value of the horizontal direction and the vertical direction of each pixel is given by $G = \sqrt{{GX}^{2} + {GY}^{2}}$ (4)

G is compared with the set threshold. If it is greater than the threshold, it is taken as the edge pixel.

2.2.3 Filtering of tillage land environment image

In the field of image and vision, the new morphological filtering method has been widely applied to many aspects of image processing. Morphological filtering is a nonlinear filtering method. The theory comes from the set theory and it has the parallel structure. Therefore, it is a good choice for the filter to be realized through field programmable gate array. There are four algorithms in morphology, namely corrosion and expansion, open operation and closed operation. In this article, the closed operation is not researched. In open operation, the soil image of the tilling area is first corroded and then expanded. In closed operation, it is first expanded and then corroded. The grayscale corrosion of the image can be defined as $\begin{matrix} (f \oplus G) (X, Y) \\ = min {f (X - i, Y - j) - G (- i, - j)} \end{matrix}$ (5)

Where (f ⊕ G) (X, Y) is the image after the grayscale is corroded, f (X - i, Y - j) are the horizontal and vertical coordinates of a pixel in the original tillage land environment image, and G (- i, - j) is the gradient function of a pixel in the tillage environment image. For each pixel in the tillage environment image, the 3 × 3 structure template is used for scanning. And operation is carried out on the area covered by the structure template. If the total is 1, the corresponding pixel of the image is set to 1, and otherwise it will be 0. The scope of the image will be reduced by a circle after corrosion. For any image in tillage environment image f (x, y), the grayscale expansion of the is given by $\begin{matrix} (f \oplus G) (x, y) = \\ max {f (x - i, y - j) + G (- i, - j)} \end{matrix}$ (6)

For each pixel in the tillage environment image, the 3 × 3 structure template is used for scanning to implement or operation on the area covered by the structure template. If the total is 0, the corresponding pixel of the image is set to 0, and otherwise it will be 1. The image will expand a circle after expansion.

In the open operation, some isolated noises in the tillage environment image can be filtered out. In this paper, open operation is used for filtering in morphology.

In conclusion, the surface of environmental image of agricultural unmanned tractor is maintained in a diffusion state, which is an expansion operation for environmental image of agricultural unmanned tractor. The purpose is to effectively remove the small holes. In order to effectively remove the image points, the pixels of the image can be removed continuously so that the image can be reduced continuously. The image processing method is called corrosion. The image processing method using the above open calculation can improve the image quality and reduce the time spent on image processing.

2.2.4 Acquisition of lateral deviation of agricultural unmanned tractor

From Section 2.1.6, it can be known that, in this paper, the pure path intelligent tracking algorithm is applied to track the preset path of agricultural unmanned tractor. The lateral deviation of agricultural unmanned tractor reflects the degree of deviation of agricultural unmanned tractor from the tracked path, that is, the distance between the control point of agricultural unmanned tractor and the tracked path. In the navigation process of agricultural unmanned tractor, it is necessary not only to define the lateral deviation, but also to determine the sign of the deviation, which can be used to decide the steering direction and steering size of agricultural unmanned tractor. In the following design, when the deviation is set as “+”, the decision wheel turns counter-clockwise. When the deviation is “-”, the decision wheel turns clockwise. Through the distance formula between points, the lateral deviation of agricultural unmanned tractor is given by $d = \frac{| {Ax}_{0} + {By}_{0} + C |}{\sqrt{A^{2} + B^{2}}}$ (7)

Where Ax + By + C = 0 is the path equation for tracking, x₀ and y₀ are the coordinates of the control point, which is obtained after the latitude and longitude projection measured by GPS.

2.2.5 Acquisition of course deviation of agricultural unmanned tractor

The course deviation of the agricultural unmanned tractor reflects the deviation of the agricultural unmanned tractor from the ideal course, that is, the difference between the deviation between the current course θ and the tracking straight path α. The current course is measured by GPS, which is the angle between the forward direction and the north direction of the agricultural unmanned tractor. The straight direction can be obtained by linear equation. When A is 0, α = 90. When A is not 0, α = arctan(- B/A).

2.2.6 Preset path tracking of agricultural unmanned tractor based on pure tracking algorithm

The pure path tracking algorithm is based on the geometric method, according to the current position of the agricultural unmanned tractor and the position of the target, to determine the circular arc required for the agricultural unmanned tractor from the current position to the target position. Assume the coordinate system of agricultural unmanned tractor is xoy. Define X_p is the abscissa of target point in the coordinate system of agricultural unmanned tractor and Y_p is the ordinate of target point in the coordinate system of agricultural unmanned tractor. γ is the turning curvature of agricultural unmanned tractor, which is positive or negative. Set that when agricultural unmanned tractor is driven counterclockwise, γ > 0, and when it is driven clockwise, γ < 0. R is the radius of the instantaneous turning of agricultural unmanned tractor. d is the error of agricultural unmanned tractor relative to lateral position in predefined path with sign. When the deviation is “+”, the decision wheel turns counterclockwise and the deviation is “-”, the decision wheel turns clockwise. L_d is the forward view distance of pure path tracking. Ψ is the deviation between the current course of agricultural unmanned tractor and the path course at the target point. φ is the change of course angle when an agricultural unmanned tractor reaches the target point. According to the above geometric relations, the coordinates of the target point in the coordinate system of agricultural unmanned tractor are given by $X_{P} = - 1 / γ + cos φ / γ$ (8) $Y_{P} = - sin φ / γ$ (9)

In the above geometric environment, there exists a right-angled triangle. This right-angled triangle is generated by the straight line in the predefined path and the vertical and horizontal coordinates of the target point in agricultural driverless tractor. According to the Pythagorean Theorem, it can be obtained as $X_{P}^{2} + Y_{P}^{2} = L_{d}^{2}$ (10)

Using Equations (8), (9), and (9), it can be obtained as $γ = - 2 X_{P} / L_{d}$ (11)

Then $X_{P} = - d cos Ψ + \sqrt{L_{d}^{2} + d^{2}} sin Ψ$ (12)

According to Equation (11), $\begin{matrix} γ = - 2 X_{P} / L_{d}^{2} \\ = 2 (d cos Ψ - \sqrt{L_{d}^{2} + d^{2}} sin Ψ) / L_{d}^{2} \end{matrix}$ (13)

According to the kinematics model of agricultural unmanned tractor, it can be known that $\dot{Ψ} = γ o = o tan δ / L$ (14)

where $\dot{Ψ}$ is the change rate of course of agricultural unmanned tractor, L is the wheelbase of agricultural unmanned tractor [18], o is the speed of agricultural unmanned tractor, and δ is the target angle of agricultural unmanned tractor. According to Equation (14) and Equation (6), the target angle of agricultural unmanned tractor is obtained as $δ = arctan \frac{2 L (d cos Ψ - \sqrt{L_{d}^{2} + d^{2}} sin Ψ)}{L_{d}^{2}}$ (15)

The visual navigation system of the tractor is constructed by constructing the main control and sub-control structure of the agricultural unmanned tractor and combining the image of the land environment. Finally, the intelligent control of agricultural driverless tractor is completed.

3 Results and discussion

In order to prove the overall performance of the proposed design method of intelligent control system of agricultural unmanned tractor based on machine vision, simulation experiment is carried out. An experimental simulation platform for intelligent control system of agricultural unmanned tractor is built under RTL8019AS environment. The experimental data are obtained from Kubota NSD-8 rice agriculture unmanned tractor. During the experiment, the feasibility of the proposed method is observed. Table 1 shows the comparison of the time consuming of image wireless transmission of different methods.

Table 1
Comparison of the time consuming of image wireless transmission

Number of Land environmental image Image transmission time of the reference [14] method/s Image transmission time of the reference [6] method/s Image transmission time of the reference [11] method/s Image transmission time of the proposed

20 6 11 8 3

30 10 19 15 6

40 15 20 19 9

50 21 26 22 11

60 29 31 26 14

Number of Land environmental image	Image transmission time of the reference [14] method/s	Image transmission time of the reference [6] method/s	Image transmission time of the reference [11] method/s	Image transmission time of the proposed
20	6	11	8	3
30	10	19	15	6
40	15	20	19	9
50	21	26	22	11
60	29	31	26	14

From Table 1, it can be seen that, the wireless transmission time of the proposed method is much less than that of other methods. In the reference [14] method, the acquisition circuit in each sensor is aging, which makes the transmission time longer. In the reference [6] method, a lot of time is wasted when measuring the fluctuation of the ground level, and the time used to transmit the detection results to the intelligent control center of the agricultural unmanned tractor is increased. In the reference [11] method, no single image acquisition module is designed, so two or more than two hardware modules are required when the tillage environment images are transmitted to the center of the system, resulting in a longer time for image transmission. The proposed method is based on the IP embedded in the front end bridge of the lower computer of the agricultural unmanned tractor, and constructs the wireless transmission link of the land environment image. This step speeds up the speed of image transmission and saves the time of image transmission, which proves that the proposed method is practicable. Table 2 shows the comparison of image expansion coverage of different methods. The image expansion coverage is calculated by using: $e = \frac{f (x - i, y - i)}{G (- i, - j)} \times 100$ (16)

Table 2

Comparison of image expansion coverage of different methods

Number of images	Image expansion coverage of the reference [14] method/%	Image expansion coverage of the reference [11] method/%	Image expansion coverage of the proposed method/%
100	80	82	97
200	75	77	95
300	70	73	94
400	66	71	95
500	63	68	94

It can be seen from Table 2 that the image extension coverage of the reference method is higher when the number of images is small. However, when the number of images continues to increase, the speed of image expansion coverage decreases very quickly.

It can be seen from Table 2 that the coverage change of the method in this paper shows a stable trend with the increase of the number of images, and the coverage is always higher than 90%, which shows good control performance, but the coverage of the method in reference [11, 14] is much lower than 90%, and its variation fluctuates greatly and the control performance is poor. because this method scans each pixel in the cultivated land environment image according to the structure template, or implements the operation of the template coverage area, improves the image extension coverage. Figure 4 shows the time taken to compare different methods for image filtering.

Fig. 4

Comparison of time consuming of image filtering of different methods.

From Fig. 4, it can be seen that, the reference method has more time for image filtering, and its time curve fluctuates greatly. In the reference method, federal data fusion technology is used to process the sensor data of agricultural unmanned tractor. As the image data obtained by sensors are more and more complex, images cannot be filtered with federal data fusion technology in a short time. In the reference method, there is no hardware specially designed for filter imaging, but the hardware module containing the filter processing technology has a simple filtering operation on the image, which leads to the slower filtering speed and longer time [17]. The proposed method uses open operation for image filtering, which accelerate the filtering speed and saves time for the design of the intelligent control system of agricultural unmanned tractor. Figure 5 shows the comparison of the relative deviation (%) of the intelligent control system of agricultural unmanned tractor between different methods. The relative deviation (%) is given by $\frac{δ}{\bar{x}} = \frac{\sum | x_{i} - x |}{\bar{x}} \times 100$ (17)

Fig. 5

Comparison of the relative deviation of the intelligent control system of agricultural unmanned tractor between different methods.

Where δ is the average deviation of n measurement results, $\bar{x}$ is the arithmetical mean of n measurement values.

From Fig. 5, it can be seen that, in the reference method, when the ridge length is within 300 m, the relative deviation is large, but it is very stable. However, when the ridge length is larger than 300 m, the relative deviation curve rises linearly, which means that the reference method is not suitable for the land with too long ridge. However, the relative deviation of agricultural unmanned tractors in reference method is far higher than that of other methods. This is because the reference method only considers the hardware design of the system, and does not research the hardware module in detail,

Which leads to the relative deviation greater than the other methods. In this paper, the hardware design of agricultural unmanned tractor is comprehensive and specific, and the responsibilities of various parts are refined [16]. In this way, the relative deviation of the operation of agricultural unmanned tractor is reduced. Figure 6 shows the comparison of path tracking efficiency (%) of agricultural driverless tractor control system of different methods. The path tracking efficiency is given by $η = \frac{d cos Ψ - \sqrt{L_{d}^{2} + d^{2}} sin Ψ}{X_{P} Y_{P}} \times 100$ (18)

Fig. 6

Comparison of path tracking efficiency of agricultural driverless tractor control system of different methods.

From Fig. 6, it can be seen that, the path tracking efficiency curve of the proposed method is slower than that of the path tracking curve proposed by other reference methods, and the tracking efficiency is higher. This is because the path tracking of the target point of agricultural unmanned tractor is tracked with the pure tracking algorithm when tracking the target path by the proposed method and the path tracking efficiency is improved. Figure 7 shows the comparison of storage space (GB) of stillage environment image data in the agricultural unmanned tractor intelligent control system of different methods.

Fig. 7

Comparison of storage space of stillage environment image data in the agricultural unmanned tractor intelligent control system of different methods.

From Fig. 7, it can be seen that, the image data of the proposed method has less system storage space. In the proposed method, considering the quality of stillage environment image, a series of processing methods have been carried out. Firstly, the low-pass filtering method is used for denoising of the stillage environment image. Then Sobel edge detection is applied to stillage environment image segmentation. Finally, the processing of stillage environment image is completed based on the corrosion and expansion of the image. These steps not only enhance the quality of the image, but also reduce the storage opportunity of the non-available images, and reduce the storage space of the stillage environment image data. Figure 8 shows the comparison of course deviation of agricultural unmanned tractor of different methods.

Fig. 8

Comparison of course deviation of agricultural unmanned tractor of different methods.

From Fig. 8, it can be seen that, the course deviation of the intelligent control system of agricultural unmanned tractor designed by using the proposed method is least. In the proposed method, when the course deviation of agricultural unmanned tractor is analyzed, the size and the sign of the lateral deviation of the agricultural unmanned tractor are clearly defined. This step not only accelerates the operation of agricultural unmanned tractor, but also decreases the course deviation. Therefore, regardless of the length of the land ridge, the course deviation of the intelligent control system of agricultural unmanned tractor with the proposed method is smaller.

Simulation experiment proves that the proposed method can quickly design the intelligent control system of agricultural unmanned tractor, increase the quantity and quality of the production of agricultural and sideline products, and improve the operation precision of the intelligent control system of agricultural unmanned tractor. It has an expandable value.

4 Conclusions

The main advantages and main methods of this paper are adaptive path tracking algorithm, through the system input information and node detection, and analysis of the farming environment, combined with the actual environment to give different control methods. In addition, the system is divided into main control and sub-control, according to the actual situation, adopt appropriate control method, at the same time, this main control and sub-control method can effectively avoid system control failure. For the design the intelligent control system of agricultural unmanned tractor with the current method, it is unable to carry out high efficiency and low error, which leads to the slow phenomenon of the intelligent control system of agricultural unmanned tractor in the operation process. For this problem, a design method of intelligent control system for agricultural unmanned tractor based on machine vision is proposed in this paper. Simulation experimental results show that the proposed method can design the intelligent control system of agricultural unmanned tractor with high precision, and provides a reference for the research and development of this field. Simulation experimental results show that the proposed method can design the intelligent control system of agricultural unmanned tractor with high precision, Reduce the heading deviation of the unmanned tractor effectively. And the tractor moving path tracking effect is good, and provides a reference for the research and development of this field.

References

Zhang

, Dong

and Liu

, A control strategy for battery-ultracapacitor hybrid energy storage system, Journal of Electrical Technology29 (2014), 334–340.

Kayacan

, Ramon

and Saeys

, Robust trajectory tracking error model-based predictive control for Unmanned Ground Vehicles, IEEE/ASME Transactions on Mechatronics21 (2016), 806–814.

Tian

F.L.

, Guo

Y.Q.

and Li

R.C.

, Design of distributed control system and investigation on bus communication performance of a turbofan engine, Journal of Aerodynamics30 (2015), 2278–2285.

H.H.

, Wang

C.H.

and Li

L.F.

, CSNS control cable information management system based on web, Atomic Energy Science and Technology48 (2014), 545–549.

Xie

H.Y.

, Luo

J.B.

and Yan

Z.Y.

, Research on on-line simulation technology for feedwater control system in nuclear power plant, Atomic Energy Science and Technology48 (2014), 556–557.

J.W.

, Zhao

H.L.

and Li

Z.M.

, Design on automatic control system of CO2 concentration in greenhouse based on STM32, Zhejiang Agricultural Journal27 (2015), 860–864.

Xiao

L.J.

, Su

and Jin

P.X.

, Study on a high-precision digital Temperature-Control System for All-Solid-State Single-Frequency Green Laser, China Laser42 (2015), 72–78.

Cong

and Xue

, Characteristics analysis of stochastic open quantum system models and feedback control, Quantum Electronic Journal32 (2015), 186–197.

Cao

, Process optimization and automatic control of double intermediate frequency induction hardening for cold roller, Metal Heat Treatment40 (2015), 152–154.

10.

Wang

, Wang

and Wang

, Research on the technology of CAN bus on vehicle intelligence terminal for tractor, Journal of Agricultural Mechanization Research38 (2017), 17–27.

11.

Chu

, Huang

and Kong

, Steady state control of MMC-HVDC system based on internal model controller, The Grid Technology39 (2015), 2223–2229.

12.

Teng

, Yang

and Guo

, Study on stability of waterpower-speed control system for hydropower station with two upstream series surge chambers, Hydroelectricity Journal34 (2015), 72–79.

13.

Yin

and Software

S.O.

, Research on unmanned tractor based on embedded PC and internet of things, Journal of Agricultural Mechanization Research35 (2018), 67–69.

14.

Y.F.

, Liu

and Wei

, An h∞filter designed for stabilization control system of airborne EO tracking and pointing platform, Lightning and Control22 (2015), 60–63.

15.

Cui

, Li

Z.J.

and Xiang

, A novel wind power system with dual-rotor and the dual-mode power control strategy thereof, Journal of the Chinese Electrical Engineering Project34 (2014), 6499–6505.

16.

Liu

, Economic analysis of methanol production from coal/biomass upgrading, Energy Sources Part B-Economics Planning and Policy13(1) (2018), 66–71.

17.

Liu

, Feng

and Wang

, Effects of the Sharing Economy on Sequential Innovation Products, Complexity (2019), 3089641.

18.

Liu

, What is the future of solar energy? Economic and policy barriers, Energy Sources Part B-Economics Planning and Policy13(3) (2018), 169–172.