KR20180096966A - Automatic Counting Method of Rice Plant by Centroid of Closed Rice Plant Contour Image - Google Patents

Abstract

The present invention relates to a method for automatically counting the number of rice seedlings using morphological and connection operations and, more specifically, a method capable of automatically counting the number of transplanted rice seedlings by applying an image processing technique to detect rice using a field situation image of a farmland obtained from a low altitude unmanned aerial vehicle and automatically count the number of rice. The method comprises the following steps. (a) An unmanned aerial vehicle flies over a farmland and takes a digital image of the farmland through a camera mounted on the unmanned aerial vehicle and transmits the digital image to a central processing unit. (b) The central processing unit counts the number of rice seedlings using the image transmitted from the unmanned aerial vehicle by the central processing unit. (c) A screen output unit generates the number of rice seedlings received from the central processing unit on the screen.

Description

{Automatic Counting Method of Rice Plant by Centroid of Closed Rice Plant Contour Image}

The present invention relates to a method of automatically counting rice seedlings using morphological operations and connection operations. More particularly, the present invention relates to a method and apparatus for detecting a rice paddy using a general agricultural field scene image obtained from a low altitude unmanned aerial vehicle And a method of automatically counting seedlings of rice seedlings using a morphological operation and a connection operation which automatically counts the number of transferred rice seedlings by applying a treatment technique.

Rice is an important crop that is a large part of the world's population, especially in Asia. However, growth, quality and yield of rice are affected not only by relics but also by environmental factors and cultivation methods. That is to say, the planting density and the planting density and the planting density of the rice plant have a great influence on the rice yield. The competitive characteristics of the plant vegetation and the germination development are greatly influenced by the plant density which greatly affects the growth and yield of rice.

Thus, from transplantation to harvesting, rice monitoring is critical for achieving the desired growth and yield, and the number of plants per area is crucial to yield.

However, manual field-based plant monitoring is labor intensive, extensive and time consuming, so it is easier, simpler and cheaper.

 On the other hand, remote sensing and digital image processing techniques are already used in agriculture, but their use is limited due to difficulties in extracting objects from the background, separating objects from overlapping and touch, and processing digital images for automatic aggregation.

Nevertheless, research on counting the number of trees or flowers using an unmanned aerial vehicle has been carried out, and although the accuracy is not high, the possibility has been shown.

However, this method has problems that can not be applied to rice detection and aggregation because rice morphological structure is very irregular and growth rate is faster than that of general kinds of rice, And there are various variables such as the density of breakfast depending on the cultivation methods of the farmers.

Therefore, it is necessary to develop a method of quantitative determination that can be applied to plants having characteristics such as rice, and further, by applying the thus-developed method to plants having morphological characteristics more easily than rice, It is a necessary point.

Patent Document 1: Korean Patent No. 10-1207098

SUMMARY OF THE INVENTION The present invention has been made in order to solve the above problems, and it is an object of the present invention to provide a method of automatically counting rice seedlings using morphological calculation and connection calculation according to the present invention. We propose an image processing technique that compiles.

In order to achieve the above object, there is provided a method of counting rice seedlings using a morphological operation and a connection operation according to the present invention, comprising the steps of: (a) taking an image of a digital image through a camera mounted on a non- To the central processing unit; (b) counting the number of rice seedlings using the image transmitted from the unmanned airplane by the central processing unit; And (c) generating a number of rice seedlings transferred from the central processing unit on the screen.

In addition, the digital image of step (a) is RGB (Red Green Blue) image.

The step (b) may further include: (b-1) converting the digital image into a grayscale image; (b-2) removing noise from the grayscale image; (b-3) converting the noise-removed grayscale image into a binary image; (b-4) performing a morphological operation on the binary image; (b-5) forming contour elements by connecting contours of shapes present in the morphologically processed binary image; (b-6) counting the number of the structural elements, wherein the number of the seedlings is the number of the structural elements.

The method may further include removing noise from the binary image subjected to the morphological operation processing between steps (b-4) and (b-5).

In addition, an adaptive median filter is applied to the noise removal.

Further, the method may further include forming a center point of the structural element in the step (b-5).

Further, the method further includes the step of comparing the outline and the center point and recognizing any one of the outline or the center point as a structural element.

The method of automatic counting of rice seedlings using morphological calculation and connection calculation according to the present invention having the above characteristics can detect and count rice without mechanical interference and can be used as an automated tool for other crops have.

Figure 1 shows morphological changes through morphological image processing techniques.
Figure 2 shows an experimental study area
Figure 3 shows the experimental equipment and operating relationship including unmanned aerial vehicles
Figure 4 is a flow chart of the rice aggregation algorithm
FIG. 5 is a view showing an RGB image
FIG. 6 shows a gray-
FIG. 7 is a graph showing a gray shaded image
Fig.
FIG. 9 shows a binary image
FIG. 10 is a view showing an image
Fig. 11 is a view showing an image
FIG. 12 shows a case where a false positive case
Figure 13 shows the correlation graph of rice counts between the proposed method and actual data

The term used in the present invention is a general term that is widely used at present. However, in some cases, there is a term selected arbitrarily by the applicant. In this case, the term used in the present invention It is necessary to understand the meaning.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the drawings.

The method for counting rice seedling seedlings according to the present invention comprises the steps of: (a) capturing a digital image of a discussion through a camera mounted on an unmanned aerial vehicle and transmitting the digital image to a central processing unit; (b) counting the number of rice seedlings using the image transmitted from the unmanned airplane by the central processing unit; And (c) generating a number of seedlings transferred from the central processing unit on the screen.

Here, the central processing unit may be a computer or a notebook computer including a main process, and the screen output unit may be a monitor connected to each computer or a notebook computer.

On the other hand, the digital image of step (a) is preferably RGB (Red Green Blue) image.

(B) counting the number of seedlings by counting the number of rice seedlings using the image received from the unmanned airplane by the central processing unit (b), the step (b) (B-2) removing noise of the grayscale image; (b-3) converting the noise-removed grayscale image into a binary image; (b-4) performing a morphological operation on the binary image; (b-5) forming contour elements by connecting contours of shapes present in the morphologically processed binary image; (b-6) counting the number of the structural elements.

In particular, the method may further include removing noise from the binary image subjected to the morphological operation processing between steps (b-4) and (b-5), wherein the noise removal includes an adaptive median filter (Adaptive median filter) is applied.

The adaptive median filter is characterized by adaptive mask shape and area segmentation, and then processed with a modified switching median filter, thereby reducing sunlight reflection, image blurring , Various types of noise caused by environmental factors such as distortion, unmanned airplane shadows, and other conditions such as obstacles of rice cultivated land can be effectively removed.

Particularly, by performing noise removal once more on the binary image subjected to the morphological operation processing between steps (b-4) and (b-5) through an adaptive median filter, the accuracy There is an effect to optimize.

The structural element is the morphological contour of the rice, which is the number of rice to be counted.

This is achieved by morphological operations on the binary image of step (b-4).

The morphological operation adopted in the present invention is a kind of image processing method. Morphological image processing is a nonlinear processing group related to the shape or form of an image, and is used in an image processing method such as shape detection, division, noise removal and the like. Morphological operations are particularly suitable for binary image processing because they depend only on the relative order of pixel values, not on mathematical values. In such morphological image processing, expansion and erosion are basic morphological processing operations and are applied as a basic element of the algorithm. This operator is used for images with a set of pixels of interest, called Structure Elements (SE).

Referring to FIG. 1, the expansion and erosion can be seen in the basic image (a) in which the structural element is not morphologically processed, and its shape changes through expansion (b) and erosion (c).

In particular, the morphological operations employed in the present invention only use closures of morphological operations. A closed operation is simply an expansion followed by erosion using the same construction element (SE) in both operations.

The method may further include forming a center point of the structural element in the step (b-5), and comparing the outline and the center point to recognize any one of the outline or the center point as a structural element .

Referring to FIG. 9, even if a contour of a shape existing in a morphologically processed binary image is connected to form a structural element in step (b-5), a center point of the structure element may not be formed, Because the accuracy of the number of seedlings can be changed according to the decision on whether to accept the seedlings.

Therefore, in order to increase the degree of purification of the seedlings conservatively, it is possible to additionally select a method of counting the center point after recognizing the center point as a structural element.

Hereinafter, the present invention will be described in detail by way of examples. However, the following examples are illustrative of the present invention, and the present invention is not limited to the following examples.

Materials and methods

Research Area

 The study area for this experiment is shown in Fig. 2, which is latitude 35 ° 1'34.24 "N and Longitude 126 ° 49'14.21" E, Jeollanam-do Agricultural Research Center in Naju, Korea. The main experimental field area was 12580m2 and the crop was rice. (Oryza sativa L.) rice transplantation date was May 25, 2016. During the rice growing season, images were taken every two weeks using an RGB camera mounted on a drone.

Obtain information

Referring to FIG. 3, the image taken from the field is taken immediately after implantation. Use a Sony alpha a5100 digital camera (lens: Sony E 16mm f 2.8). The camera was 24.3MP with CMOS image and the maximum resolution was 6000 * 4000. The rice image was taken on a clear sunny day and the height of the camera was 10m from the crop. The camera was mounted on a rotor-type unmanned aerial vehicle (model: DJI-S1000; Geospatial Information Co. Ltd.). The forward flight speed was 2.5 m / s. The notebook was connected to the unmanned aircraft with a wireless data link connector (model: LK 900, dji, 900MHz).

Parameters used by the UAV and camera to collect data from the field For Rotary wing UAV For Camera (Sony alpha a5100) Parameters Parameters Altitude (m) 10 Image sensor CMOS Flying speed (m / s) 2.5 F-stop f / 2.8 Ground sampling distance (cm / pixel) 0.2 Exposure time 1/2500 sec Time for image acquisition (min / ha) 15 ISO 200 Max. flight time (min) 15 Focal length 16 mm

Image processing

4 is a flowchart of the rice counting algorithm. First, the RGB image (see FIG. 5) taken from the camera mounted on the unmanned airplane is read as a MATLAB workspace and converted into a grayscale image (see FIG. 6). An adaptive median filter was applied to the gray shade image (see FIG. 7) for noise reduction. Then, since the morphological operation is applied to the binary image, the grayscale image is converted into the binary image (see FIG. 8). In this method, closed morphological techniques were used for binary images (see FIG. 9). The Closed operation softens the contour portion and combines the narrow and long thin valleys to remove small tiny holes and fill the gaps in the contour.

Again, adaptive intermediate filters are applied to the image to remove more noise. This process is followed by linking contours of existing shapes in a morphologically processed binary image to form a structural element by drawing the boundary (see FIG. 10). Then, a center point (see FIG. 11) is created for each connected component and the center point is counted in rice paddies. Aggregate the total number and display the results. If one of the elements in the 8 direction of Equation 1 and the current pixel value are connected through the connection operation from the rice seedling candidate obtained by the morphological operation, the rice is detected by treating it as a connecting element.

1, x + 1, y + 1, x + 1, y + 1, (x + 1, y), (x + 1, y + 1)}. ... ... ... ... . (Equation 1)

Here, x and y are coordinate values of the respective axes.

To obtain the number of rice seedlings thus obtained, the outline of rice seedlings is obtained and the center point of the rice seedlings is obtained based on the outline. The number of rice seedlings is obtained through the number of center points of the rice seedlings thus obtained.

Performance evaluation

We chose accuracy, recall, and F-Measure for performance measurements. Calculate with the following equation.

 Where TP is the number of correctly detected rice, TN is the exact number of non-rice detection, and FP is the number of non-rice in the detected rice.

This performance measurement format can be used to measure the performance of the method.

Experiment result

Image processing

 The first four steps of the image processing algorithm are to use a filter for the basic transformation of the image format and noise reduction. All of these procedures were used prior to morphological processing. Gray shades and binary images showed good results with noise reduction. After the closure of the morphological operation was performed, an adaptive median filter was applied and very good results were obtained. At this stage, the binary pixels (0s and 1s) represent the components of the image. 1 indicates a black pixel in rice, and 0 indicates a background of an image. By this, 1 shows the center of gravity of individual rice. If the filter is applied after performing a close operation on the image, the center of gravity of the plant is shown clearly and clearly as shown in the figure below. At the end of this process, each connected black pixel is bounded by a connected component. It then creates a center of gravity for each connected component, recognizes it as rice, and counts the center points. Aggregate the total number and display the results.

Performance evaluation

 The proposed algorithm was tested by randomly selecting 10 test images.

Table 2 shows the proposed method and the number of rice detected and counted visually. True positive, false positive, and false negative numbers of rice are also calculated. Recall, precision, and F-measure were calculated using equations (2), (3), and (4). The results show that the algorithm can perform 87% of the performance with the F-measure corresponding to 85% accuracy and 89% recall. The best-fit image shows 93% performance according to F-measure, 90% accuracy and 97% recall. In Figure 12, some errors are shown due to false positive errors. The correlation of rice counts between the proposed method and actual data is shown in Fig. The R ² value is about 0.47.

Results of visual handwriting calculations and image-processed methods according to the present invention. Sample Ground truth No of plant TP FP FN Recall Precision F-measure Image 01 37 38 28 6 5 0.84 0.82 0.84 Image 02 43 48 38 6 6 0.86 0.86 0.86 Image 03 43 37 31 5 6 0.83 0.86 0.85 Image 04 36 38 32 6 5 0.86 0.84 0.85 Image 05 35 37 33 4 3 0.91 0.89 0.90 Image 06 40 44 38 6 4 0.90 0.86 0.88 Image 07 48 45 35 8 5 0.87 0.81 0.84 Image 08 42 35 29 6 One 0.96 0.82 0.89 Image 09 51 42 38 4 One 0.97 0.90 0.93 Image 10 49 51 40 9 3 0.93 0.81 0.86 Average 0.89 0.85 0.87

It will be apparent to those skilled in the art that various modifications and variations can be made in the present invention without departing from the spirit or scope of the present invention. Various modifications and variations will be possible without departing from the spirit of the invention. Therefore, the scope of the present invention should be construed as being covered by the scope of the appended claims, and technical scope within the scope of equivalents thereof should be construed as being included in the scope of the present invention.

Claims (7)

(a) taking a digital image of a discussion through a camera mounted on an unmanned aerial vehicle and transmitting the digital image to a central processing unit;
(b) counting the number of rice seedlings using the image transmitted from the unmanned airplane by the central processing unit; And
(c) generating on the screen the number of rice seedlings transferred from the central processing unit by the screen output unit.
The method according to claim 1,
Wherein the digital image of step (a) is RGB (Red Green Blue) image.
The method according to claim 1,
The step (b)
(b-1) converting a digital image into a grayscale image;
(b-2) removing noise of the grayscale image;
(b-3) converting the noise-removed grayscale image into a binary image;
(b-4) performing a morphological operation on the binary image;
(b-5) forming contour elements by connecting contours of shapes present in the morphologically processed binary image;
(b-6) counting the number of the structural elements,
Wherein the number of the rice seedlings is the number of the structural elements.
The method of claim 3,
And removing noise from the binary image subjected to the morphological operation processing between the step (b-4) and the step (b-5).
The method of claim 4,
Wherein an adaptive median filter is applied to the noise removal.
The method of claim 3,
And forming a center point of the structural element in the step (b-5).
The method of claim 6,
Comparing the contour with a center point to recognize either the contour or the center point as a structural element.

Images