JP2947141B2 - Weed recognition device for weeding robot - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention analyzes a regular pattern pattern appearing in an image photographed by a TV camera by a texture analysis method which recognizes the features of the entire group as a pattern without focusing on individual form recognition. The present invention relates to a weed recognition device for a weeding robot that recognizes the presence or absence of weeds.

FIG. 2 shows an example of an image obtained by extracting a portion showing only weeds from an image taken by a TV camera.
3 is a chart in which the density values at the observation line a in FIG. 2 are plotted. FIG. 4 shows an example of an image obtained by extracting a portion showing only grass from an image taken by a TV camera, and FIG.
It is the chart which plotted the density value in observation line b inside. Each of these images has a unique pattern, and in the case of turf, black and white change is particularly severe. Further, in the chart in which the density values on the observation line a and the observation line b are plotted, there is a difference in the distance between pixels belonging to the same density value, and in the case of weeds, the variation in the distance between pixels is particularly large. The weed recognition device of the present invention recognizes the presence or absence of weeds mixed in grass or the like from the knowledge that the distance between pixels belonging to the same density value is different for weeds.

[0003]

The use of herbicides must not harm main crops such as turf. For this reason, weeding robots that remove weeds using herbicides
In order to minimize the area where the herbicide is injected, it is necessary to specify the place where the weeds grow in a narrow area.

According to the present invention, an image area is subdivided into area units necessary for specifying a place where a herbicide is to be injected by a herbicide, and the presence or absence of a weed in each subdivision area is determined to specify an area containing a weed. The purpose is to do.

[0005]

In order to achieve the above object, the present invention is configured as follows.

That is, in a system in which a regular pattern pattern appearing in an image captured by a TV camera is analyzed to recognize the presence or absence of weeds, a density value level in which all density values of the image are divided into a plurality of density value levels Classifying means, area dividing means for equally dividing the image into a plurality of sub-areas, and inter-pixel distance calculating means for calculating a distance between reference points of a group of pixels belonging to the same density value level for each of the sub-areas;
A distance variation calculating unit that calculates a variation in the distance between the pixels; and a weed presence determining unit that determines the presence or absence of a weed in each subdivided area according to the magnitude of the variation in the distance. The subdivision area determined to be present is further divided, and the presence or absence of weeds is determined for the subdivision area after division in the same manner as the subdivision area before division.
By repeating the operation of further dividing the subdivided area determined to have weeds an arbitrary number of times, the presence or absence of weeds is determined for the subdivided areas obtained by dividing the image taken by the TV camera any number of times, and thereby the area containing the weeds is determined. And a weed recognition device for a weeding robot.

[0007]

Embodiments of the present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram of a weed recognition apparatus according to the present invention. The weed recognition device of the present invention includes a density value level classifying unit 1 for inputting an image of a TV camera and classifying all the density values of the image into a plurality of density value levels, and similarly inputting an image of the TV camera to convert the image. An area dividing means 2 for equally dividing into a plurality of subdivided areas; an inter-pixel distance calculating means 3 for calculating a distance between reference points (such as center points) of a group of pixels belonging to the same density value level for each subdivided area; It is composed of a distance variation calculating means 4 for calculating a variation in the distance between pixels, and a weed presence determining means 6 for determining the presence or absence of a weed based on the magnitude of the variation in the distance.

The weed recognizing device of the present invention is configured as described above, receives an image from a TV camera, and classifies all the density values of the image into several tens of density value levels by the density value level dividing means 1. Next, the image of the TV camera is equally divided by the area dividing means 2 into several tens of subdivided areas. Then, the following processing is performed for each of the above-described subdivision areas.

First, the inter-pixel distance calculating means 3 calculates the distance between the reference points of the pixels belonging to the same density value level for each of the above-described subdivision areas. Then, the distance variation calculating means 4 calculates the variation in the distance between the pixels. Finally, the weed presence determining means 6 compares the distance variation calculated by the distance variation calculating means 5 with the distance variation of an image of only the main crop such as turf, which is stored in advance, and determines the size of the weed. Is determined.

Here, as shown in FIG. 6, for the subdivided area determined to have a weed, the subdivided area is further equally divided into several tens of areas, and each subdivided area belongs to the same density value level described above. The process from calculating the distance between pixels to determining the presence or absence of weeds is repeated.

By performing the above processing, the subdivision area determined to have a weed is subdivided into a minimum area necessary for specifying the place where the herbicide injects the herbicide, and the subdivision area including all the weeds in the image area is obtained. The processing is terminated at the point in time when is specified.

[0013] Pixels in parts such as leaves, soil, and shades of an image captured by a TV camera have almost the same density value. Therefore,
When the distance between pixels belonging to the same density value level is calculated in these portions, the answer is "1" in most cases, and the distance between pixels does not vary, and the pattern pattern cannot be recognized.

As a countermeasure for this, a group of neighboring pixels belonging to the same density value level is labeled with the same number, and the distance between reference points having the same label number is calculated to obtain the variation. , An appropriate pattern pattern can be recognized.

The label processing will be described below.
First, as shown in FIG. 7, all density values of an image are divided into several density value levels, and label names such as A, B, C. Next, this image is equally divided into several tens of sub-areas, and the following processing is performed on each of the sub-areas.

First, as shown in FIG. 8, for each pixel in the subdivision area, if there is a pixel belonging to the same density value level in the vicinity of 8, the same label name is assigned to each pixel. Then, as shown in FIG. 9, the distance between the reference points having the same label name is calculated for the subdivision areas color-coded by the labels, and the variation is calculated. Hereinafter, the presence or absence of weeds is determined based on the size.

[0017]

The weed recognition apparatus according to the present invention comprises a density value classifying means for classifying all density values of an image into a plurality of density value levels, and an area dividing means for equally dividing the image into a plurality of subdivided areas. An inter-pixel distance calculation unit that calculates a distance between reference points of a pixel group belonging to the same density value level for each of the subdivision areas; a distance variation calculation unit that calculates a variation in the distance between the pixels; Weed presence / absence determination means for determining the presence or absence of weeds for each subdivision area according to the magnitude of the variation, further dividing the subdivision area determined to have weeds by the weed presence / absence determination means, By repeating the operation of determining the presence or absence of weeds in the same manner as the subdivision area and further dividing the subdivision area determined to have the weed, an arbitrary number of times,
The presence / absence of weeds is determined for a subdivided area obtained by dividing the image captured by the TV camera an arbitrary number of times, thereby identifying an area having weeds. Therefore, a camera for general use can be used without requiring a filter or a special camera requiring sensitivity in the near infrared region, so that the cost of the weeding robot can be reduced. Also, regardless of whether the surroundings of the weed are soil or grass, there is a difference in the distance between the pixels belonging to the same density value level, so that there is an effect that the weed can be identified.

[Brief description of the drawings]

FIG. 1 is a block diagram of a weed recognition device according to an embodiment of the present invention.

FIG. 2 is an example of an image in which a portion where only grass is photographed is extracted.

FIG. 3 is a table in which density values of a line a in FIG. 2 are plotted.

FIG. 4 is an example of an image in which a portion showing only weeds is extracted.

FIG. 5 is a table in which density values of a line b in FIG. 4 are plotted.

FIG. 6 is an explanatory diagram of processing for equally dividing a subdivision area to determine the presence or absence of a weed;

FIG. 7 is an example in which all density values of an image are divided into several levels of density values.

FIG. 8 is an example in which pixels belonging to the same density value level near eight have the same label name.

FIG. 9 is an example of a subdivision area color-coded by a label.

[Explanation of symbols]

 DESCRIPTION OF REFERENCE NUMERALS 1 density value level dividing means 2 area dividing means 3 inter-pixel distance calculating means 4 distance variation calculating means 5 weed existence determining means

──────────────────────────────────────────────────続 き Continued on the front page (58) Fields surveyed (Int. Cl. ⁶ , DB name) G06T 7/00 A01M 7/00 G06T 1/00

Claims (1)

(57) [Claims] 1. A method for recognizing the presence or absence of weeds by analyzing a regular pattern pattern appearing in an image photographed by a TV camera, wherein a density value level for dividing all density values of the image into a plurality of density value levels Classifying means; area dividing means for equally dividing the image into a plurality of sub-areas; inter-pixel distance calculating means for calculating a distance between reference points of a group of pixels belonging to the same density value level for each of the sub-areas; A distance variation calculating means for calculating a variation in the distance between pixels; and a weed presence determining means for determining the presence or absence of a weed in each of the subdivided areas based on the magnitude of the variation in the distance, wherein the weed presence determining means has a weed. The subdivision area determined as above is further divided, and the presence or absence of weeds is determined for the divided subdivision area in the same manner as the subdivision area before division, and By repeating the operation of further dividing the rear any number of times, it is possible to determine the presence or absence of weeds in the subdivided area obtained by dividing the image taken by the TV camera any number of times, thereby identifying an area having weeds. Weed recognition device of a weeding robot.