JPH01277408A - Automatic steering controller in agricultural working machine - Google Patents

Abstract

PURPOSE:To enable output of a steering signal with a high accuracy, by entering data from an area sensor in data obtained with a line sensor to provide a steering output signal from the above-mentioned data and carrying out automatic steering control. CONSTITUTION:Image information sensed by respective sensors is processed in an image processing part 45 for an area sensor 39 and an image processing part 46 for a line sensor 40 to specify positions of the respective planting operations. Deviation of transverse shift position and direction from the central position, center line, etc., in the image pickup screen, is calculated from data on positions of planted crop places by the area sensor 39 in an arithmetic part 47 of a central controller 35. In an arithmetic part 48, deviation of transverse shift position from the standard line is calculated from data on positions of planted crop places by the line sensor 40. Furthermore, an instruction extent of steering operation is calculated according to a theory, such as fuzzy control, through a data memory part 49, arithmetic part 50, comparing part 51, etc., of the central controller 35 to output a steering output signal to a steering control valve 34, etc.

Description

[Detailed Description of the Invention] [Industrial Application Field] The present invention is directed to agricultural work using a rice transplanter or the like in a manner substantially parallel to rows of crops, such as so-called planted seedling rows, which are already planted in a field and lined up in rows. This invention relates to an automatic steering system that allows the aircraft to travel.

[Conventional technology]

Conventionally, when planting seedlings in a field using a rice transplanter, the rice transplanter is equipped with a mechanism for planting seedlings at appropriate intervals on the left and right in the direction of movement.
The planting mechanism, which moves up and down as the rice transplanter advances, divides the seedling mat on the seedling platform into appropriate numbers of plants and plants them on the field, so the rice seedlings are placed on the field in an appropriate manner along the direction of movement of the rice transplanter. It is well known that seedlings are planted in multiple rows at appropriate intervals (hereinafter referred to as row spacing) in the left and right directions with respect to the direction of propagation, while the seedlings are lined up at planting intervals (referred to as planting pinch).

Japanese Patent Laid-Open No. 62-61509 discloses a prior art technology for an automatic steering device that allows the rice transplanter to run approximately parallel to rows of seedlings that have already been planted in the field. A color video camera is used to image an appropriate area of the adjacent rows of planted seedlings, and this imaged screen information is binarized to extract areas corresponding to each planted seedling location. After that, a straight line is drawn from the row of the plurality of regions by processing such as Hou g h transformation, and this calculated virtual straight line is combined with an arbitrary reference line such as the vertical and horizontal center line of the imaging screen. It is proposed to perform steering control of the aircraft so that the difference between lateral deviation and inclination deviation with respect to a reference point is within a certain allowable range.

[The problem that the invention attempts to solve]

In the prior art, the imaging means has a two-dimensional plane with an imaging screen extending vertically and horizontally, like a so-called video camera, and therefore, an optical image of the object to be detected is placed on the photoelectric groove surface of the photoelectric sensor. When decomposing pixels in time series, by repeating scanning lines along the horizontal axis (X-axis) of the NANCER plane and sequentially shifting them along the vertical axis (y-axis), - According to this two-dimensional area sensor, it is possible to simultaneously detect multiple plants (seedling locations and their geometric relationships) on a single screen as image information.

However, on the other hand, the amount of data in the above-mentioned image information is enormous, so the capacity of the storage unit must be increased, and image processing calculations are required to reduce the predetermined data necessary for automatic steering to 1. On the other hand, there is a problem in that a computer with a small capacity mounted on a vehicle has a slow calculation speed, and it takes too much time to control the steering from the calculation result.

On the other hand, among the two-dimensional photoelectric sensors, one consisting of a plurality of horizontally arranged pixels corresponding to one row, or a so-called line sensor (also referred to as a -dimensional photoelectric sensor) consisting of a plurality of photoelectric sensors arranged horizontally, is used to perform the above-mentioned implantation. When detecting a location, the line sensors are arranged so that their pixels (or photoelectric elements) are lined up long on the left and right, perpendicular to the direction of movement of the planter, and when one of the line sensors The system uses pixels (or photoelectric elements) to detect the location of planted seedlings, and while the image processing to obtain image information is simple and quick, it is also possible to Since it does not include any information about the direction before or after, it may be the case that the rows of planted seedlings are lined up approximately in a straight line, but if the rows of planted seedlings are deviated significantly from side to side, or if the rows of planted seedlings are uneven (
In the case where one seedling row exists along a curved ridge), there is a drawback that information about the shift and curvature cannot be predicted.

Therefore, when automatic steering control is performed using only either the area sensor or the line sensor, the defects become noticeable.

For planting crops such as seedlings, management machines that fertilize or spray chemicals at appropriate times after work, and harvesting machines for vegetables, etc., carry out work while proceeding along the rows of crops that have already been planted. The same problem as above occurs when an automatic steering device is mounted on an agricultural machine.

The present invention aims to solve the above problem by using both an area sensor and a line sensor.

(Means for Solving the Problems) Therefore, the present invention provides a device for automatically steering an agricultural machine such as a rice transplanter so that it runs substantially parallel to the rows of crops already planted in a field. The planting is performed using a line sensor that scans and images the crop location in the left and right longitudinal directions in the direction of movement of the aircraft, and an area sensor that scans and images the crop row two-dimensionally. It is configured to be in front of the imaging range of the line sensor, and data processing of both the image obtained by the line sensor (image information and the image information captured by the area sensor) is performed almost simultaneously, and the line sensor Data from the area sensor is included in the data, and automatic steering control is performed using the If3 rudder output signal from these data.

[Action/effect of the invention]

According to this configuration, image information data from the line sensor and image information data from the area sensor are
Although data processing is performed simultaneously, the number of times that data can be processed within a fixed time is limited to area sensors, which take longer data processing times, and - data processing time, which is shorter while data processing of image information from the system is executed once. Data processing of the image information from the line sensor system can be performed multiple times.

Therefore, by executing data processing by both sensors almost simultaneously and adding data from the area sensor to data from the line sensor, data that is insufficient from the line sensor alone can be added and the data processing of these data can be improved. Based on the results, a more accurate steering signal can be output.

Since the imaging range of the area sensor is further forward than the imaging range of the line sensor, the image information data from the area sensor can be obtained in advance as data ahead of the agricultural machine in the traveling direction.

Therefore, even if it takes a long time to process the data from the area sensor, by adding the predictive data to the instant data from the line sensor, rich data can be obtained. For example, it has the remarkable effect of being able to output a steering output signal many times in a short period of time and dramatically improving the accuracy of automatic steering using this output signal.

〔Example〕

An example applied to a rice transplanter will be described below. In the figure, 1 is a running body supported by front wheels 3, 3 on both left and right sides of the front part of a frame 2, and rear wheels 4, 4 on both left and right sides of the rear part. At the rear of the body 1, a multi-row seedling planting device 7 comprising a seedling stand 5 and a plurality of planting mechanisms 6 is mounted via a link mechanism 8 so as to be movable up and down.

The power of the engine 9 mounted on the upper surface of the frame 2 of the traveling body 1 is transmitted to both the front and rear axles 3.4 via the clutch 10 and the transmission case 11, and is transmitted via the PTO shaft 12 protruding from the transmission case 11. Power is transmitted to the seedling planting device 7.

In addition, numeral 13 is an actuator for ON-OFF of the clutch 10, numeral 14 is an actuator for traveling speed change, and numeral 15 is a PTO.
Actuator for shaft angular velocity.

A steering gear box 17 is provided on the upper surface of the traveling aircraft 1 in front of the pilot seat 16, and a steering shaft 19 inserted into the steering column 18 is attached to the upper end of a steering column 18 that stands up from the steering gear box 17. The control handle 20 is attached.

Reference numeral 21 denotes a left and right long transmission case in which front wheels 3.3 are attached to both left and right ends via knuckles 22.22, and a power transmission mechanism from the transmission case 11 is housed inside, and the transmission case 21 is attached to it. The bracket I-212, which is U-shaped in plan view, is connected to the swing shaft 2IC rotatably supported by the support member 21b at the center of the lower surface of the frame 2, so that it can swing horizontally and vertically. It is composed of

The steering device includes a steering arm 24 which has a 17-shape in plan view and is horizontally rotatably mounted on a rotational fulcrum shaft 23 erected from one side of the transmission case 21.
The pitman arm 28 is comprised of a pair of left and right tie rods 25, 25 connected to the left and right, a hydraulic cylinder 26, a control valve 27, and a pitman arm 28 that operates the control valve 27, and a steering gear box 17 that is swingable back and forth.

One end of the pair of left and right tie rods 25.25 is connected to the arm portion 24.2 of the steering arm 24, which extends in the front-rear direction of the frame 2, through a ball-and-socket joint. It extends substantially along the case 21 to the left and right as the traveling body 1 moves, and its other ends are swingably connected to knuckle arms 29, 29 extending forward from the knuckles 22, 22.

A control valve 27 is connected rearward via a ball joint to a support shaft 30 of an arm extending inwardly from the steering arm 24 toward the side surface of the frame 2, and a spool at the rear end of the control valve 27 The pier 1 and the man arm 28 are connected through a connecting rod 31.

Further, the rear end of a hydraulic cylinder 26 disposed substantially parallel to the outer surface of the frame 2 is connected to the support shaft 30 via a ball joint, while the front end of the piston rod 26a is connected to the frame via a ball joint. 2 is connected to a horizontal placket shaft 32 protruding from the outer surface.

The control valve 27 and the hydraulic cylinder 26 are each connected by a hydraulic hose, while hydraulic pressure is sent to the electromagnetic solenoid control valve 27 from a hydraulic pump 33 driven by the engine 5.

Then, the pitman arm 28, which swings in accordance with the rotation angle of the control pan 1ζ lever 10, moves the spool of the control valve 27 forward and backward to move the piston in the hydraulic cylinder 26. 2
In response to the rotation of 4, the directions of both left and right front wheels 3.3 are changed.

This hydraulic cylinder 26 is operated by an electromagnetic solenoid type steering control device operated by a central control device 35 for automatic steering and traveling that outputs an output signal in response to a signal from a crop row detection device 37 such as a planted seedling row, which will be described later. It is also driven by a valve 34, in which case the steering angle of the front wheels 3 is determined by detecting the rotation angle of the steering arm 24 with a potentiometer 36 attached to the rotation fulcrum shaft 23.

The actuator 13 for ON/OFF of the clutch 10, the actuator 14 for traveling speed change, the actuator 15 for changing the circumferential speed of the PTO shaft, and the central control device 35 are operated.

Next, a system and method from detecting the arrangement of crop rows by the crop row detection device 37 to automatic steering by the central control device 35 will be explained.

The crop row detection device 37 includes an image capturing means for capturing an image of a target, an image processing section serving as an image processing means for processing the captured image and outputting necessary information (data), a calculation section, and the like.

When detecting an object, the imaging means uses an area sensor 39 whose imaging screen has a two-dimensional plane (xy plane) like a so-called video camera, and a line sensor 40 whose imaging screen is arranged in a row in the horizontal direction. Consists of both.

In the area sensor 39, when the optical image of the object to be detected is resolved in time series on the photoconductive surface of the photoelectric sensor, the repetition of scanning lines scanned along the horizontal axis (X-axis) of the sensor surface is expressed as the vertical axis ( By sequentially shifting the images along the y-axis), − images are formed. According to this two-dimensional area sensor 39, a plurality of crops 12 and 12 are planted on − images. The locations can be detected together with their geometrical relationships, and image information can be obtained.

As an example of this Eliasen"J--39, for example,
In models with a built-in two-dimensional MO3 image sensor or two-dimensional CCD image sensor, the image formed through the lens is sensed by each image sensor (photoelectric element) arranged in a two-dimensional array on the image formation plane. The information on the imaging screen 42 can be output as an electrical signal.

On the other hand, the line sensor 40 may be made up of a plurality of horizontally arranged pixels for one row of the two-dimensional area sensor 2, or may be made up of a plurality of photoelectric sensors arranged side by side.

This line sensor 40 has a built-in -dimensional MO3 image sensor or a one-dimensional CCD image sensor, and the image formed through the lens is formed by each image sensor (photoelectric sensor) arranged in a one-dimensional array on the image forming plane. The information on the imaging screen 43 can be output as an electrical signal.

The sixth line is arranged so that the reference line KO in the imaging direction of the line sensor 40 is parallel to the traveling direction of the traveling aircraft 1, and the pixels in the row are lined up on the left and right of the traveling aircraft 1.
(see figure), and furthermore, the imaging range of the line sensor 40 is
As shown in FIG. 2, a field scene 44 directly below the side of the traveling machine body 1 is photographed, and on the other hand, the imaging range by the area sensor 39 is further in front of the traveling machine body 1 than the imaging range of the line sensor 40. As you can see, the direction of both sensors is cent.

The image processing section 45 for the area sensor 39 and the image processing section 46 for the insensor 40 process the image information detected by each sensor to identify the position of each planting crop. .

Then, in the calculation unit 47 in the central control device 35, based on the data on the position of the planted crops from the area sensor 39, the lateral deviation position deviation ρ2 from the center position 0 of the imaging screen 42, the center line (y-axis), etc., and the direction The calculation unit 48 calculates the reference line K of each line sensor 40 from the data of the position of the crop location for planting using the line sensor 40.
Each planting for O 1) Calculate the lateral shift position deviation ρ1 of the crop location.

In this way, the line sensor 40 and the area sensor 39 obtain image information and process it almost simultaneously according to the flowchart 1 below. Arithmetic unit 50,
A steering operation instruction amount SO is calculated through a comparison section 51 or the like according to a control theory such as fuzzy control, and a steering output signal is output to a steering section, for example, a steering control valve 34 for rotationally driving the steering wheel. be.

In this case, the imaging range by the area sensor 39 is further forward in the traveling direction of the traveling aircraft 1 than the imaging range by the line sensor 40, and the imaging screen by the area sensor 39 will simultaneously capture multiple planting crops. Considering the advantages of having a large amount of image information and being able to predict the arrangement of crop rows, we use area sensors that have a long processing time to process data such as image information processing, and area sensors that can perform real-time processing with a short processing time. After entering the data,
A steering output signal is obtained using these data, thereby performing accurate and rapid automatic steering control that could not be achieved with processing using only line sensors or area sensors.

The flowchart will be explained according to FIGS. 7-a and 7-b. First, the operator on the traveling machine 1 performs a turning operation to start traveling forward along the side of the row of crops that have already been planted. When planting a plurality of crops in a crop row (planted seedling row) located on the side of the traveling machine, the automatic steering device is turned on so that images of the crop locations (NAE) can be taken.

Figure m7-a shows data processing of image information by the line sensor 4o, and in step S1 following start 1 and initial value setting, ρ2. .theta. is input, and set in a counter c-0 as an initial value in step S2. Next, in step S3, the value of C is inputted to step R3 in FIG. 7-b, which will be described later.

In step S4, the line sensor 4o starts detection operation and captures image data. In this case, the line sensor 40 is arranged so that its pixels 43 are lined up in a row on the left and right in the traveling direction of the aircraft, and the reference line KO passing through the pixel 43 at the approximately center position of the line sensor 40 is Since it is arranged so as to roughly follow the direction, when a planting crop spot comes to the position of the reference line KO, the distance from the crop row on the traveling machine 1 next to it is set to the set value (H) (seedling In the case of planting, it is equal to the planting row spacing).

Since the line sensor 40 and the area sensor 39 are arranged on the left side of the traveling aircraft 1, the line sensor 40 and the area sensor 39 are shown in FIGS.
In the figure, when the plant (t LJ crop location) is imaged at a position biased to the right side of the figure (to the right of the y-axis in Figure 5, to the right of the reference line KO in Figure 6), the setting value (H
) is shorter, in other words, the traveling machine 1 is too close to the already planted crop area; When the image is taken at a position that is biased to the left of the reference line KO,
This indicates that the traveling body 1 is too far away from the already planted crop location than the set value (H).

Next, in step S5, a binarization process is performed to distinguish the planting area from other field scenes. In this example, when configuring a captured image on a color screen, the RGB color system [red (R), green (G), and blue (B) color lights are used as primary color lights, and white is obtained by adding light. Red component due to
The characteristics of the field scene are extracted from the signals of each color component, green and blue, and the sum of the signal outputs of these three color components (R+G+
Identify the position of the seedling by identifying it as a seedling when the signal output ratio of the green (G) component to B-1) is equal to or higher than a predetermined value 2
Execute value processing.

In addition, binarization is performed using color difference processing that determines that a portion of the image where the color difference image data (C; -B) obtained by subtracting the blue component from the #3 green component of the color signal is an output of − room or above is a seedling. It's okay.

Furthermore, like a black-and-white TV, the difference in brightness between the seedling spot and other dirt surfaces is detected, and when the threshold of -room or higher is exceeded,
For planting, binarization processing may be performed to determine that it is a seedling location (NAE).

The line sensor 4 has pixels arranged in a line as described above.
Based on the data of the detected seedling location (NAE) according to 0, in step S5, the positional deviation ρ1 of the lateral shift with respect to the reference line KO is calculated.

In this case, for example, in the case of a MOS type, an address readout method is adopted in which a switching pulse of a shift register is applied to photoelectric elements, which are pixels arranged in a column, to sequentially extract signals from each element.

In this way, based on the output signal from one or more of the pixels 43 arranged in a row, it is determined that a seedling is located at that location.

Step S6 is the calculation of the lateral deviation position deviation ρ1 and the difference Δρ1 between the previous detection result and the current detection result of the lateral deviation position deviation ρ1.

In addition, in FIG. 6, (a) is the -th detection (C=0), (b)
) is the second detection (C-1), (C) is the third detection (C=2
), the results of multiple detections are shown, and the shaded areas are the planted crops.

Next, step R, which will be described later, is input in step S1.
The combination of the values of the positional deviation ρ2 and the directional deviation θ of the lateral deviation in 9 and the value of the positional deviation ρ1 of the lateral deviation and Δρ1 (
Then, in step S7, the steering operation instruction 1isQ is calculated, and as a means of calculation, so-called fuzzy control, which will be described later, is executed in this embodiment.

Next, in step S8, a steering signal is output based on the steering operation instruction amount SO. After that, in step S9,
An interrupt flag IF from the area sensor 39, which will be described later, is input, and it is determined in step SIO whether or not this flag rF=0. If IF=0, step SIO is performed.
The value of the counter C is incremented by 1 at 11, and the process returns to before step S3, and the process continues from step S3 to step 3 using the image information data from the line sensor 40.
Repeat the process up to 11.

Therefore, C increases by 1 each time this process is executed.

On the other hand, when IF≠0, return to before step S1,
Input the positional deviation ρ2 and the direction deviation θ of the lateral shift in step R9, which will be described later, and set the counter C= which was the arbitrary number mentioned above.
It will be reset to 0 and the processing up to step SIO will be executed.

Therefore, by using the value of the counter C and the interrupt flag IF, it is possible to perform data processing by the line sensor 40 and data processing by the area sensor 39 almost simultaneously, and to link the two data. be.

The image information processing and arithmetic processing time from step S4 to step S6 in the line sensor 40 is approximately 1 time of the time required to perform one image information processing in the area sensor 39.
Since it is about /10 to 115, it means that while the area sensor 39 executes image information processing once, the line sensor 40 can execute it 5 to 10 times.

On the other hand, in the flowchart of FIG. 7-b, in step R1 following start 2 and initial value setting, the interrupt flag IF is set to 0 by seven degrees, and in step R2, the flag IF is output to the step SIO.

In step R3, the counter C from step S2 is
In step R4, it is determined whether it is C-0 or not, and when it is Cl3, the process returns to step R3 so as not to perform the processes from step R5 onwards.

At the time of C-O, area sensor 3 at step R5
The image information from 9 is input, and binarization processing is executed in step R6. This binarization process is performed in the same manner as the binarization process in step S5 of the line sensor 40 described above. As shown in FIG. 5, the binarized image shows that crop areas in a plurality of planting rows can be distinguished from other field scenes, and each can be divided into two specific areas (A).

Next, in step R7, using a predetermined calculation method by the software 1- incorporated in the calculation unit 47, from the data on the position of the specified area (A), based on the center position or center of gravity position. Execute virtual line calculations of straight line approximations using eight-transform, least squares method, etc. The virtual line KL obtained in this way indicates the expected arrangement of the crop rows imaged by the area sensor 39 in the area ahead of the traveling aircraft 1. This is shown by the solid line in Figure 5.

In step R8, the intersection O of the X-axis and the y-axis, which is a reference point on the imaging screen 42, and the y-axis, which is a reference line parallel to the traveling direction axis of the traveling aircraft 1, are taken, and the above-mentioned virtual line KL is The calculation unit 47 calculates the distance between the point where the axis intersects with the zero point as the positional deviation ρ2 of the lateral shift, and the angle of intersection between the y-axis and the virtual line KL as the direction deviation θ.

After outputting the difference ρ2 and θ between the two parts in step R9 and inputting them to the step S1, the interrupt flag IF=1 is set in the step RIO, and the value of IF=1 is set to the step S9 in the step R11. Output.

At the output stage of step R11, an appropriate waiting time is provided in step R12 so that the line sensor 40 can detect the completion of processing on the IF-〇 or area sensor 39 side in steps SI0 and SI11, and then step R
This will return it to 1.

An example of the time table of these two flowcharts is shown in FIG. As can be understood from this time table, the steering instruction amount S is calculated based on the latest ρ2 and θ data from the area sensor 39, data ρ1 and Δρ1 for each process from the line sensor 40, and the value of C.

Next, fuzzy control using fuzzy theory will be explained.

In general, a control algorithm is implemented using a plurality of information input variables for control, for example, two input variables (x.

y) and the output to the control equipment (operation ff1) 2.

for example, If X is small and y is large, 2 should be in the middle.

If X is large and y is medium, make 2 large.

As in, the control algorithm is expressed by what is called a fuzzy control rule of the (if-tl+en) format (if...then...whatever).

The if... part of the rule is the antecedent part, then...
The part is called the consequent part.

Now, in applying this fuzzy control rule to automatic steering control, in this embodiment, if the label of the fuzzy variable of counter C is 1 greater, the set of data ρ1 and Δρ1 is used as the antecedent part. , the automatic steering operation instruction amount S is assumed to be the consequent part, and when the label of the fuzzy variable of C is "small", the data set of ρ2 and θ is assumed to be the antecedent part,
A total of 18 rules having the operation instruction amount S as the consequent are expressed as shown in Tables 1 and 2 below.

The reason why the counter C is classified by the label of the fuzzy variable is as follows.

The actual processing time by the line sensor 40 is short, and the actual processing time by the area sensor 39 is long. The imaging range of the sensor 39 is the front in the traveling direction of the traveling body 1, and the planting by the line sensor 40 can be predicted in the future more than by the data of crop location detection. and,
As can be understood from step R5 of the flowchart, the area sensor 39 starts inputting image information only when the counter C-0, and when the value of the counter C is small, in other words, the fuzzy variable of the counter C When the label is "small", it means that ρ2 and θ have just been obtained in the previous processing of the area sensor 39, so that data is used, and conversely, the label of the fuzzy variable of counter C is 1 larger. Sometimes, the data of ρ2 and θ from the area sensor is old, so it is decided to use the detection results of the line sensor 40, ρ1 and Δρ1, as data.

In addition, the fuzzy variables taken by the input variables C1ρ1, Δρ1, θ, ρ2 and the output variable S are expressed as discrete fuzzy variables discretized in the integer domain, and Tables 3 to 7
Shown in the table.

The label of the fuzzy variable in counter C in Table 3 is
There are two types of fuzzy variables: "large" and "small", and five types of fuzzy variables as labels: 1 "large right", "small right", "0", "small left", and "large left".

The area of fuzzy variables, which is an ambiguous area, is defined by giving the degree (grade) to which the elements (members) of the entire set of input variables are included in the area (domain). The function that gives l is called the member jump function. For example, the region of the fuzzy variable ρ1 is defined by giving the degree (grade) to which the elements (members) of the entire set of the position deviation ρ1 of the strike slip are included in the region (domain). In the embodiment, the grades falling in each domain are expressed as integers from O to IO.

(The following margin) Table 1 Fascinator LII IL + = Bo "Uji -C is 1 larger" Table 2 C Table 3 Counter C Table 4 of Counter C: JK J Table -E -E -Book Left・-” A-Lake Rule--・Right (blank space below) Table 6 Oxi/30 rad = Table 7 To explain further, for example, the position deviation ρ1 of the lateral shift, the previous result of this position deviation, and the current result. Δ, which is the difference from the result of
In ρ1, the fact that the fuzzy variable is "small and right" is expressed by a membership function with grade r3,5,8°10.8,5,3J in the region "from -1 to +5". .

The inference method in fuzzy control is as follows. For example, when C is large and ρ1 and Δρl are 2
Consider the case of the output S (consequent part) for the antecedent part (from ■ to ■) of a human-powered set.

Membership function A in the i-th rule (from ■ to ■) corresponding to the value of the antecedent part of each input ρ1 and Δρ1
(ρ1)) and A(Δρli) are determined from Table 4, and the degree of fitness ωi =A (ρli) *A (Δρli) is determined by pairing it with the human power ρ1 and Δρ11. Here, * is a symbol for m1ni (intersection in a fuzzy set, the same applies hereinafter)) or multiplication.

Then, the inference result corresponding to the i-th rule is obtained from Table 7, and the fitness degree wi 13 (si)-ωH*B (si) is obtained from the membership function B (si).

Next, in the same way as above, the membership function A(
Find the values of ρlj) and A(ΔρIj) from the above Table 4,
The degree of fitness ωj−8(ρlj) *A (Δρ1j) is determined.

Next, the inference result corresponding to the j-th rule is obtained from Table 7, and the fitness w is calculated from the membership function B(sj).
Find j B (sj)-ωj*B (sj).

These plurality of consequents are, for example, i-th, j-th, k-th,
In the first four cases, the overall inference result SO based on these four rules is wi B (si) + Wj B
(sD, wk B (sk), wl
From B(si), B*-wi B(si)Uwj B(sj)Uwk
Create B(sk)UwL B (sl) (U is the symbol for union) and set the membership of B*33
~ Find as the center of gravity of the Tsubu function.

Similarly, when C is small, the output S(
Also in the case of the consequent part), the membership functions Δ(ρ2i) and A(θi) in the i-th rule (from @l to [phase]) corresponding to the values of the antecedent part of each input ρ2 and θ. The values of ρ21 and θi are obtained from Tables 5 and 6 above, and
The degree of fitness for the pair is determined in the same way as above.

To illustrate this control using an example of actual numerical values, the counter number C=3, ρ1-2. Δρl=1, ρ2=40.
The steering operation instruction NSO when θ-2 (1/30 (rad)) is obtained as follows.

First, from Table 3, the label is "small" for C=3.
When &J, the value of the membership function is "5", and when the label is "large", the value of the membership function is "1".

Now, considering the case where the label is "small" for C=3, the pair of data 34= (antecedent part ■ to ■) of ρ2 and θ corresponds.

Since the detection results are ρ2-40 and θ-2, the labels of the corresponding fuzzy variables are determined from Tables 5 and 6.

Regarding the side-slip position deviation ρ2-40, in Table 5, if you look at the vertical bar part corresponding to the division of "+33 to -+-48J" in the domain, it is "3" in the label "1 greatly to the right",
“10” in “small right” and “5” in “0”
'', so we can take the labels of these three fuzzy variables and find that (S).

Regarding the azimuth deviation θ, if we look at the vertical bar corresponding to the “2” category of the θ range in Table 6, we can see that it is “5” at the label “0” and “10” at “small left”. , and "3" in "large left", so there is a possibility that these three fuzzy variables can be adopted (M).

Therefore, the fuzzy rules based on the data set of ρ2 and θ can be found in Table 2 for the four phases of [phase], ■, ■, and O.
The labels of the consequent part S are respectively "0 (go straight)", "cut small to the right", 1 "cut large to the left", "
It turns out that it becomes "cut sharply to the right."

These membership functions can be determined by referring to each table above.
Regarding the fuzzy control of [phase], see Figure 10-a,
The fuzzy control of ■ is shown in Figure 11-a, Figure 1l-1), and Figure 11-0. -a figure,
The fuzzy control of 0 is shown in FIG. 12-b and FIG. 12-0, and the fuzzy control of 0 is shown in FIG. 13-a, FIG. 13-b, and FIG. 13-0.

Incidentally, the membership function of counter C is shown in FIG.

In this case, for example, regarding the fuzzy control (■), the fitness degree ωj of the operation instruction amount S is limited to the smaller of the ωj values of ρ2 and θ, and the degree of adaptation ωj is limited to the smaller one of the values of ωj of ρ2 and θ, and the degree of adaptation ωj is increased as shown in Fig. 11-0. This is the area below the number "5" surrounded by a thick line.

Similarly, the membership function in each consequent S in the fuzzy control of [phase], [phase], and O is the first O-36
= The area surrounded by thick lines shown in Figure -C, Figure 12-c, and Figure 13-0.

The overall inference result Sol based on the above three rules is B*-ω
i B (si) UO3B(sj)Uωk B(sk
) (U is the symbol for union) and find it as the center of gravity of the surface area surrounded by the outermost area when the membership functions of the three S are overlapped (see Figure 14), the result is The steering operation instruction amount is 5OI-0°67.

Similarly, when considering the fuzzy control rule from the detection results ρ1-2 and Δρ1-1 when the label of C is "large", for ρ1, the range of ρ1 in Table 4 is "
Looking at the vertical bars corresponding to the category ``2'', we see that there is a ``3'' in the label ``Large right'', a ``10'' in the label ``Small right'', and a ``5'' in the label ``0''. Since it straddles, there is a possibility that these three fuzzy variables, that is, the labels "large right", "1 small right", and "0" can be adopted.

On the other hand, if we look at the vertical bar part that corresponds to the "1" category of the range of Δρ1 in Table 4, we see that "8" in the label "small right"
, straddles "8" in the label "0" and "3" in the label "small left". These three fuzzy variables, namely the labels ``small right'', ``1 small left'' and ``0
” may be adopted.

Therefore, the fuzzy rule for inference in Table 1 is that if we search among the combinations ■ to ■ in the antecedent part, then ρ1 is "0" and Δρ1 is "corresponding to 0-1", and ρ1 is " 4 rules: ``Small to the right'' and Δρ1 is 1 small to the left'', ρ1 is ``Small to the right'' and Δρ1 is ``Small to the right'', and ρ1 is 1 large to the right'' and Δρ1 is 0. correspond, and the labels of the subject S are "0" and "0", respectively.
, ``Cut small to the right'' and ``Cut 1 large to the right''.

These membership functions can be determined by referring to each table above.
(However, the diagram of the membership function of counter C is
The one corresponding to the fuzzy rule ■ shown in the figure is the 16th rule.
-a, 16-b, and 16-0, and those corresponding to fuzzy rule ■ are shown in 17-a and 17-b.
18-a, 18-b, and 18-0, and those corresponding to fuzzy rule ■ are shown in FIG. Chapter 19-a
19-b and 19-0, respectively.

In this case, each degree of fitness ωi = A (ρli) *A (
θi), * is the symbol of m1ni, so the fitness degree in the fuzzy rule ■ is ω-5, and correspondingly, the membership function in the consequent part S has a height of `` as shown in Figure 16-0. 5" is the area surrounded by the thick line.

Similarly, the fitness for the fuzzy rule ■ is ωj−3, and the corresponding membership function in the consequent part S is surrounded by a thick line below the height “3” as shown in Figure 17-0. This is an area where

Furthermore, the fitness degree ω in the fuzzy rule ■ becomes −8, and the corresponding membership function in the consequent part S is surrounded by a thick line below the height “8” as shown in Figure 18-0. This is an area where

Also, the fitness degree ωl in the fuzzy rule ■ is 3,
The membership function in the consequent part S corresponding to this is a region surrounded by a thick line below the height "3" as shown in FIG. 19-0.

The overall inference result SO■ based on these four rules is found as the center of gravity of the area surrounded by the outermost area when the four membership functions of S are overlapped (see Figure 20). Steering operation instruction amount 5OII--1,18
becomes.

Judging from FIG. 9, the value CI=5 of the membership function when the label is "small" for C=3, and similarly from FIG. 15, the label is "large". Considering the membership function value CII=1 when
1+CII).

From the calculation formula, SO=0.35.

In accordance with this value, an output signal is sent to the central control device 35, the electromagnetic solenoid of the steering control valve 34 is activated, and the hydraulic cylinder 26 that changes the rotation angle of the steering arm 24 in the steering mechanism is driven to perform corrective steering, On the side of a predetermined row of crops, automatic steering control is executed in which the traveling machine 1 moves forward in parallel along the row of crops. In this case, the potentiometer 36 that detects the rotation angle of the steering arm 24 determines whether or not the front wheels 3 already have a steering angle that tilts to the right or left with respect to the direction of travel. This can be useful for convergence control that progresses to a parallel state.

With this configuration, when planting in a wide area, the area sensor has the disadvantage that the calculation speed is slow because it detects the crop area at once and processes the image, and it takes too much time to control the steering from the calculation result. This problem can be solved by using a line sensor that requires a short calculation time for detection and image processing, and has the effect of enabling accurate and quick automatic steering control.

[Brief explanation of the drawing]

The drawings show an embodiment of the present invention, in which Fig. 1 is a plan view of a riding rice transplanter, Fig. 2 is a side view, Fig. 3 is a plan view of main parts of the steering device, and Fig. 4 is a steering/traveling automatic machine. A block diagram of the control device and an action explanatory diagram including the hydraulic circuit, Figure 5 is a diagram of the image capture screen by the area sensor, Figures 6 (a), (b), (
C,) are diagrams of the image capture screen by the line sensor, respectively.
Figures -a and 7-b are for flowchart 1, Figure 8 is a control time chart, Figures 9 and 15 are diagrams showing the membership function of counter C, and Figures 10-a and 15 are diagrams showing the membership function of counter C. Figure 10-b and Figure 1 Q-0 are fuzzy control rules [phase]
Diagrams of membership functions in the case of , Figures 11-a and 1
Figures 1-b and 11-0 are membership function diagrams for fuzzy control rule O, and Figures 12-a, 12-b, and 12-0 are membership functions for fuzzy control rule ■. Ship function diagrams, Figures 13-a, 13-b, and 13
-0 is a diagram of the membership function in the case of fuzzy control rule O, FIG. 14 is an explanatory diagram showing the method for determining the steering operation instruction amount Sol, and FIGS. 16-a, 16-b, and 16-0.
The figure is a diagram of the membership function in the case of the fuzzy control rule ■. Figures 17-a, 17-b, and 17-0 are diagrams of the membership function in the case of the fuzzy control rule ■.
Figures 8-a, 18-b, and 18-0 are membership function diagrams for fuzzy control rule ■, and Figures 19-a, 19-b, and 19-0 are for fuzzy control. FIG. 20, which is a diagram of the membership function in the case of rule (2), is an explanatory diagram showing a method for determining the steering operation input amount son. 1... Traveling body, 2... Frame, 3.4...
... Wheels, 5 ... Seedling stand, 6 ... Planting mechanism,
7...Seedling planting device, 8...Link mechanism, 9...
Engine, 11...Minsion case, 17...
・Steering gear box, 20... steering hand 1
NΔE...Seedling location, K...Virtual line, 23...Rotation fulcrum axis, 24...Steering arm, 26...Hydraulic cylinder, 34... ... Steering control valve, 35... Central control device, 36... Potentiometer, 37... Crop row inspection device, 39...
... Area sensor, 40 ... Line sensor, 4
5.46... Image processing section, 47.48.50...
- Arithmetic section. Patent applicant Yanmar Agricultural Machinery Co., Ltd. Agent
People Patent Attorney Akio Hoben

Claims (1)

[Claims] (1). In a device that automatically steers an agricultural machine such as a rice transplanter so that it runs approximately parallel to rows of crops that have already been planted in a field, the machine scans the area of the planted crops in the left and right longitudinal directions in the direction of movement of the machine. The sensor is configured to include a line sensor that captures an image by scanning a row of crops two-dimensionally, and an area sensor that captures an image by two-dimensionally scanning a row of crops. The image information obtained from the line sensor and the image information obtained from the area sensor are both processed almost simultaneously, the data from the area sensor is added to the data from the line sensor, and a steering output signal is obtained from these data. An automatic steering control device for an agricultural machine, characterized in that it is configured to perform automatic steering control.