JPH0575334B2 - - Google Patents

Description

[Detailed Description of the Invention] [Industrial Application Field] The present invention is provided with an imaging means for imaging the boundary between an untreated work area and a treated work area in two-dimensional directions, and uses the imaging information to , relates to an imaging-type boundary detection device provided with an image processing means for determining a straight line corresponding to the boundary based on the brightness of each quantized pixel arranged in a two-dimensional direction.

[Conventional technology]

This type of imaging-type boundary detection device described above takes advantage of the fact that untreated work areas and treated work areas appear to have different brightness, and processes the captured image information to detect boundaries. It is designed to find a straight line.

Since the information on the obtained straight line is mainly used as control information for automatically driving a work vehicle along the boundary, conventionally, all pixel rows lined up in the longitudinal direction of the boundary are subject to image processing. was.

[Problem that the invention seeks to solve]

The detected straight line information is mainly used as control information to automatically drive the work vehicle along the boundary between the untreated work area and the treated work area, so it can be used over a wide range along the longitudinal direction of the boundary. It is desired to process image information of

In other words, if the state in front of the vehicle can be detected before the vehicle travels to that point, deviations from the boundary can be quickly corrected.

Therefore, in order to increase the number of pixels to be processed that line up in the longitudinal direction of the boundary, it is possible to image a wide range of work areas in two dimensions and increase the number of pixels to be processed. However, when processing all the pixels arranged in the longitudinal direction of the border as in the above conventional method, increasing the number of pixels arranged in the longitudinal direction of the border results in a very large number of pixels to be processed. This has the disadvantage that it becomes difficult to process images at a speed that corresponds to the traveling speed of the work vehicle.

The present invention has been made in view of the above-mentioned circumstances, and its purpose is to enable rapid image processing while using image information in a wide range along the longitudinal direction of the boundary.

[Means for solving problems]

The characteristic configuration of the imaging type boundary detection device according to the present invention is as follows.

That is, the image processing means is configured to process a plurality of pixel rows or a group of pixel rows located apart from each other among a plurality of pixel rows arranged in the longitudinal direction of the boundary. .

[Effect]

The boundary between the untreated work area and the treated work area exists continuously, and even if image information is thinned out along its length, the detection accuracy of the straight line corresponding to the boundary will not be affected. .

Therefore, by processing a plurality of pixel rows or groups of pixel rows that are located apart from each other among a plurality of pixel rows arranged in the longitudinal direction of the boundary, the number of pixels to be processed can be thinned out, and the total number of pixel rows can be reduced. Although it uses image information over a wide range along the longitudinal direction of the boundary,
This suppresses an increase in the number of pixels to be processed in image processing.

〔Effect of the invention〕

Therefore, it is possible to use a wide range of image information along the longitudinal direction of the boundary without increasing the number of pixels to be processed, so it is possible to speed up image processing without reducing detection accuracy. It came to this.

〔Example〕

Embodiments of the present invention will be described below based on the drawings.

As shown in Fig. 6, a pair of left and right front and rear wheels 1, 2
However, each of them is provided so that the steering can be operated freely, and a lawn mowing device 3 is suspended on the lower abdomen of the vehicle body V so as to be movable up and down, thereby forming a work vehicle used for cutting grass, weeds, etc. There is.

As shown in FIGS. 6 and 7, an image sensor _S1 serving as an imaging means that images the work area in two-dimensional directions images a predetermined range of the work area on the front side with respect to the vehicle body V. It is set up like this.

To explain the imaging field of view of the image sensor _S1 , the vehicle body V is in an appropriate state with respect to the boundary L between the unmown land B as an untreated work land and the mowed land C as a treated work land. In this state, the boundary L is located at the center of the field of view in the width direction.

To explain the structure of the vehicle body V, as shown in FIG. 5, there are hydraulic cylinders 4 and 5 for steering to operate the front and rear wheels 1 and 2 separately, and
Control valves 6 and 7 are provided therefor, and a hydraulic continuously variable transmission 8, which can freely switch between forward and backward directions and can freely change speeds in both forward and backward directions, is interlocked with the engine E.
A speed change motor 9 is operatively connected to a speed change arm 10 of the speed change device 8.

Further, in order to detect the traveling distance of the vehicle body V,
A rotation speed sensor S ₂ that outputs a set number of pulse signals per unit rotation speed is provided so as to be rotationally driven by the output of the transmission device 8 .

A steering position detection potentiometer that detects the steering position of each of the front and rear wheels 1 and 2.
R ₁ , R ₂ , and a shift position detection potentiometer R ₃ that detects the operating state of the transmission 8 are provided, and the detection information thereof and the image sensor S ₁ are provided.
Based on information on _a straight line corresponding to the boundary L determined based on the imaging information of A control device 11 using a microcomputer is provided to control the vehicle to automatically travel along the same route.

In the figure, H is a steering handle for boarding operation, R ₀ is a potentiometer for detecting its operation position, and 12 is a speed change pedal for boarding operation.

However, the image processing means 100 for detecting a straight line corresponding to the boundary L by image processing the image information captured by the image sensor _S1 is not connected to the control device 1.
It is configured using 1.

To explain automatic driving of the vehicle body V, as shown in FIG. Steering based on the information detected by the image processing so that V automatically travels along the boundary _L1 between the uncut area B and the cut area C along the length direction of the work process. to be controlled, and the boundary L ₂ on the end side of one working stroke
In other words, by repeating the process of automatically turning toward the starting end of the next working stroke that intersects with the opposite side of the unmoved land B, a predetermined area is cut in a so-called circular cutting style. The lawn mowing work is done automatically.

Adding an explanation while describing the operation of the control device 11, as shown in _FIG . As shown in _FIG . The steering will be controlled so that the vehicle automatically travels along ₁ .

Based on the detection information of the rotation speed sensor _S2 , the traveling speed is decelerated to a preset speed as the vehicle travels within a set range for a distance corresponding to the length of one working process. After that, the end detection process is performed to detect the boundary _L2 on the end side of the work process using the full screen information, and the end position relative to the current position of the vehicle body V is calculated based on the position information on the image. Based on the information on the calculated end position, when the end of the work stroke is reached, the machine turns toward the next work step based on a preset and stored turn pattern.

After the turn, for example, it is determined whether the work has been completed by determining whether the preset number of strokes has been traveled or whether the set travel distance has been reached. If the work is not completed, the processes from the boundary detection process described above will be repeated.On the other hand, if the work is completed, the vehicle body V will be stopped and the entire process will be completed.

In addition, to explain the reason for decelerating when approaching the terminal end, the boundary _L2 on the terminal end side is
It will exist in the width direction on the screen,
Also, since its position is unknown, it is necessary to process the image using full-screen information, but in that case, the number of pixels targeted for image processing increases,
The speed of image processing is lower than that when detecting the boundary _L1 on the stroke side. Therefore, the traveling speed of the vehicle body V is reduced to correspond to the processing speed of image processing.

To explain the calculation of the deviation with respect to the boundary L ₁ on the stroke side, there is a deviation in the position toward the horizontal axis of the vehicle body and a deviation in the inclination with respect to the direction of the boundary L ₁ , but these are calculated as quantitative values. Alternatively, only the direction of deviation may be calculated.

Also, if we add an explanation about the steering control process,
The parallel steering type corrects positional deviations in the width direction of the vehicle body, and the four-wheel steering type corrects directional deviations. However, the steering amounts of the front and rear wheels 1 and 2 may be differentiated to correct the positional and directional deviations at the same time.

Next, the boundary detection process will be described in detail based on the flowchart shown in FIG.

However, the boundary detection process described below is performed by utilizing the phenomenon that the uncut area B appears darker than the cut area C.

When detecting the boundary L ₁ on the stroke side, high-speed image processing can be performed in accordance with the traveling speed of the imaging V while preventing the detection accuracy of the straight line corresponding to the boundary L ₁ from decreasing. In order to do this, the screen range to be processed is limited by thinning out the pixel rows lined up in the longitudinal direction of the detection boundary _L1 , that is, in the y-axis direction on the screen, every few rows.

However, when detecting the boundary _L2 on the terminal side, as described above, since the entire screen information is to be processed, the screen range is not limited.

That is, as the boundary detection process starts, the image sensor S ₁ performs imaging processing, and imaging information over the entire field of view is captured and quantized in accordance with a preset pixel density. be done.

The captured image information will be smoothed to remove small variations in brightness.

After smoothing, it is determined whether or not to limit the screen to be processed based on whether the boundary _L1 on the work process side or the boundary _L2 on the end side is detected.

In the case of screen restriction, as shown in Fig. 2, within the setting range (m ₁ to n ₁ , m ₂ to _{n 2} ; however, m ₁ < n ₁ < m ₂ <
By excluding only the pixel columns located at n ₂ ) from the differential processing to obtain the differential value SX of the brightness change, the pixel columns to be processed in the length direction of the boundary L ₁ on the stroke side are thinned out. .

In this differential processing, the image processing means 10
0 is configured to process a plurality of pixel row groups located apart from each other among the plurality of pixel rows lined up in the longitudinal direction of the boundary L.

To explain why it is possible to thin out the pixel columns to be processed in the length direction of the boundary L ₁ on the process side, the boundary L ₁ exists continuously in the vertical direction on the screen. Therefore, even if continuous pixel rows are not used, the detection accuracy does not deteriorate.

On the other hand, when the screen is not limited, that is, when detecting the boundary _L2 on the terminal side, a differential value is calculated for each pixel of the entire screen.

To explain the differential processing, eight pixels a to d, which are adjacent to the pixel e to be processed, are
Using a 3x3 pixel mask (see Figure 4) that covers 9 pixels including f to i, based on the following formula (i),
Differential value of brightness change in the x-axis direction on the image
The process of calculating SX is performed for each pixel arranged in the two-dimensional direction.

SX(x,y)=(a+d+g)−(c+f+i)
...(i) However, in this differentiation process, the sign of the differential value SX is determined based on whether the boundary L is located on the left or right side with respect to the vehicle body V, and it is determined whether it is positive or negative. Only the value that serves as the sign will be used.

In other words, in equation (i), the brightness of the pixel located on the right side of the image is subtracted from the brightness of the pixel located on the left side of the image, so if the left side is brighter, the differential value SX is It becomes a positive value,
If the right side is brighter, it will be a negative value.

Therefore, when driving with the boundary _L1 located on the left side with respect to the vehicle body V, the differential value
If a vehicle with a positive SX value is extracted and the vehicle is run while positioned on the right side, a vehicle with a negative differential value SX is extracted.

Then, by performing binarization so as to extract only pixels for which the absolute value of the differential value SX is greater than or equal to a preset threshold value, pixels at locations where brightness changes are large are extracted.

Next, using Hough transform processing, for each binarized pixel, a large number of straight lines that pass through that pixel and are divided at each set angle are divided based on the following formula (ii) in a polar coordinate system (ρ, θ), take a histogram of the frequency of taking the same coordinate value, find a straight line with the maximum frequency from the histogram, and connect that straight line to the boundary L ₁ in the direction along the work process.
Detected as a straight line corresponding to .

ρ=x・cosθ+y・sinθ...(ii) However, in this boundary detection process, the screen used for explanation is a virtual one, and the straight lines corresponding to each of the boundaries L ₁ and L ₂ are It is not actually drawn, but the value ρ found by equation (ii) above, which is the most frequent value, and the value of the angle θ, which is the value, are used as information corresponding to the detected straight line. . Then, in the process of calculating the deviation with respect to the boundary and calculating the end position, from those values ρ and θ,
By mapping to the coordinate system of the actual ground, the actual positions of the boundaries L ₁ and L ₂ with respect to the vehicle body V are calculated.

Then, by controlling the steering based on the information mapped on the actual ground, the vehicle body V is moved to the boundary.
The robot automatically travels along _L1 , and when it reaches the terminal end boundary _L2 , it is guided to automatically turn toward the starting end of the next work stroke.

By the way, when detecting the boundary _L2 on the terminal side, the brightness changes in the front and back direction with respect to the vehicle body V, and changes from bright to dark from the top to the bottom on the screen. In the above-mentioned differential processing, the differential value of the brightness change in the y-axis direction is calculated and used in place of the differential value SX in the x-axis direction, and the terminal side boundary L ₂ is detected in the same process. I will do it.

In addition, since the image sensor S ₁ is installed so as to image the work area on the front side of the vehicle body from diagonally above, the position change of each boundary L ₁ and L ₂ on the screen obtained from the captured image information is is inversely proportional to the distance to the actual positions of boundaries L ₁ and L ₂ with respect to the vehicle body V, and becomes smaller as the distance increases.

Therefore, in each of the determination of the terminal side boundary _L2 , the calculation of the distance to the terminal side boundary _L2 , and the calculation of the deviation from the detected stroke side boundary _L1 , the detected position on the screen is Corresponding to this, the actual position with respect to the vehicle body V is converted.

[Another example]

In the above embodiment, only a group of pixel rows located far apart in the y-axis direction from among the pixel rows arranged in the y-axis direction on the screen are subjected to differential processing. It is also possible to thin out one row of pixels located every other column as targets for differential processing. In addition, as shown in FIG. 8, in addition to the process of thinning out the pixel rows aligned in the y-axis direction, only the pixels located in a predetermined range on the center side of the pixels aligned in the x-axis direction are subject to differential processing. By doing so, the number of pixels to be processed may be further thinned out. In other words, the vehicle body V
Since the vehicle is controlled to automatically travel along the detected boundary L ₁ , the boundary L ₁ will not be imaged at a position away from the center in the width direction of the imaging field of view. Therefore, even if only the pixels located at the center in the width direction of the imaging information are used, the boundary _L1 will not be detected incorrectly, and image processing can be further speeded up.

Furthermore, in the above embodiment, the pixels to be subjected to image processing are limited by thinning out the pixels to be subjected to differential processing. The binarization range after processing, the Hough transform target range, etc. may be limited, and the specific configuration for limiting the processing target pixels and the image processing means 10
The specific configuration of 0 can be changed in various ways.

Further, in the above embodiment, the detection information of the rotation speed sensor S ₂ is used to illustrate the case where the process is performed when the vehicle body V approaches the end of the work process, but for example, the process of boundary detection Detection of the termination may be performed alternately each time.

In addition, in the above embodiment, the case where the binarization is performed based on the magnitude of the differential value SX in the x-axis direction is exemplified, but for example, the differential value in the y-axis direction may also be used to determine the brightness change. The direction may also be considered.

Furthermore, in the above embodiment, a case has been exemplified in which Hough transform processing is used as the straight line calculating means for determining the straight line corresponding to the boundary L, but the specific configuration of the straight line calculating means can be modified in various ways.

Further, in the above embodiment, the present invention is used as a means for automatically driving a lawn mowing work vehicle, but the present invention can also be applied as a means for detecting the boundaries of various work areas. The specific configuration of each part including the image processing means can be changed in various ways.

Incidentally, although reference numerals are written in the claims section for convenient comparison with the drawings, the present invention is not limited to the structure shown in the accompanying drawings by the reference numerals.

[Brief explanation of the drawing]

The drawings show an embodiment of the imaging type boundary detection device according to the present invention, in which FIG. 1 is a flowchart of boundary detection processing, FIG. 2 is an explanatory diagram of a processed image, and FIG. 3 is a flowchart showing an outline of control operation. Chart, Figure 4 is an explanatory diagram of differential processing, Figure 5 is a block diagram showing the control configuration, Figure 6 is an overall side view of the work vehicle, Figure 7 is an explanatory diagram of the work area, and Figure 8 is a separate implementation. It is an explanatory diagram of an example. B... Untreated working area, C... Treated working area, L
... Boundary, S ₁ ... Imaging means, 100 ... Image processing means.

Claims (1)

[Claims] 1 Boundary L between untreated working area B and treated working area C
An image capturing means _S1 for capturing an image of the part in two-dimensional directions is provided, and an image processing means calculates a straight line corresponding to the boundary L based on the brightness of each quantized pixel arranged in the two-dimensional direction from the image capturing information. 100, the image processing means 100 detects a plurality of pixel columns or pixel columns located apart from each other among a plurality of pixel columns arranged in the longitudinal direction of the boundary L. An imaging type boundary detection device configured to process groups.