TW200836803A - Sphere identification system - Google Patents

Abstract

An sphere identification system is provided. The method includes retrieving an original image of the subject; converting into a gray-level image data; proceeding edge detecting to retrieve a shape of the subject; amplifying the shape of the subject; proceeding the threshold to obtain an appearance of the subject; proceeding the convex hull to fill with the inner space of the appearance of the subject; removing the noise outside the image; cutting off the point portion of the edge. Therefore, the identification system determines whether the subject is the predetermined ball or not through the calculated center, x-coordinate and y-coordinate of the subject.

Description

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image processing system, and more particularly to a method for identifying a sphere for detecting tennis balls. [Prior Art] With the advancement of science and technology civilization, the leisure service industry has developed rapidly. If some of the service-oriented work can be performed by robots, it will be automated and have the effect of entertainment. In recent years, the government has been researching and developing robotic support robots. Therefore, this project will be born. At present, the scope of use of robots is not limited to applications in the industrial world. In fact, in the service fields such as leisure and entertainment, many intelligent machine people are also used to serve the public in daily life; for example, robots that are automatically cleaned, robots that pour tea, etc. Modern people are busy, nervous, and stressed. Many people must use sports to relieve tension. Therefore, many ball games are increasingly loved by office workers. Taking tennis as an example, during the practice, the stadium is full of many nets, waiting for the caddies to pick them up. However, in caddy picking up tennis balls, they often suffer from frictional tears due to long-term weight on the back, causing severe back injury. To do this, it is to reduce the load on the caddy and prevent the caddy from being physically injured. Therefore, there will be a robot that can play the role of a caddy to replace the person who used the 200836803 to pick up tennis balls and cause serious damage to the body. However, in order to accurately judge whether the object to be picked up is a sphere to be picked up by %, the following invention is produced. SUMMARY OF THE INVENTION It is therefore an object of the present invention to provide a method for identifying a sphere for taking in the shape of a target and confirming whether it is a preset sphere for the purpose of accurately identifying the tennis ball. According to the above object of the present invention, a method for identifying a sphere for ingesting a shape of a target and confirming whether it is a preset sphere is selected. The method for identifying the sphere includes the following steps: Step (A): Take the original image of the target. Step (B): Convert the original image into grayscale image data. Step (C): Gray-level image data is used for edge detection (Detection) to extract the appearance of the object. Step (D): After edge detection (Edge Detecti〇n), the data is expanded to expand the outline of the target. ° Step (Ε) • Perform a binarization (Thresh〇ld) process to get the appearance of the target. Step (F): Perform a convex hull (c_exHull) process to fill the internal vacancy of the object's appearance. Step (G): Remove noise from the periphery of the image object. Step (H): Cut off the sharp part of the edge. Step (I) ························································································ In the identification of the sphere of the present invention, by taking the color original image of the target object, the machine edge detection and expansion effect of the tennis ball, the second step is sequentially passed through the gray scale, and the processing is performed. The tennis ball that you want to pick up. """^ and sure to determine whether it is a robot

[Embodiment] Referring to Fig. 1, a flow chart of this step is shown. Method for Identifying the Lunar Body - Example Referring to Figure 2, it is a graph of the original image of the color captured by the library 110 in the flowchart. μ楚 The two brothers are shown in grayscale graphics. For color graphics, please refer to Appendix 1. ... the identification method of the sphere of this month, which is used to ingest a target object/shape and to determine whether it is a preset sphere (10) (see Figure u). In this embodiment, it is used to identify the shape of the tennis ball. The method for identifying the sphere includes the following steps: Step · Refer to the process 110 and the process 111 of FIG. 1 , the second figure and the annex 1 'first capture the color original image of the object 1 , including the color, shape and size of the tennis ball and Edge pattern, and chroma analysis. Referring to Fig. 3, it is a graph corresponding to the color-graded grayscale image of the flow 120 in Fig. 1. Figure 10 is a statistical diagram after processing by process 120. Step 2: Convert the original image of the object 100 into grayscale image data by referring to the process 120 and the process 121 of FIG. 1 and the third figure of 200836803 and the annex 1. Since processing a color image is three times longer than processing a grayscale or black-and-white image, this step turns the color image into a grayscale image, which speeds up the execution of the program. In addition, the image contains three colors of R, G, and B. Therefore, it is divided into RGB-Red Plane (color to red level) and RGB-Green Plane (color to green level) in the gray scale of the image (Vision). RGB-Blue Plane (color to blue level). Through the prototype of Gray Morphology, it can be seen that the Blue Plane has a deeper color, and the Green Plane is more moderate and clear. Therefore, in this embodiment, the grayscale image is taken. Grayscale image in green. Referring to Fig. 4, there is a graph corresponding to the edge expansion of the flows 130, 140 in Fig. 1. Step 3: Referring to the process 130, FIG. 4 and the annex 1 of FIG. 1 , gray detection image data is used for edge detection to extract the appearance contour of the object 100. Step 4: Refer to the process 140, FIG. 4 and Annex 1 of FIG. 1 to expand the edge detection data to expand the outline of the object. In addition, Edge Detection is mainly a description of feature extraction. The amount of data has been greatly reduced compared with the original image, so it is more convenient and convenient to describe this image and post-processing. The present invention seeks the edge of the image by using the Sobel Filter. This operation contains two operators as follows: '1 2 1 _

Gy = 0 0 0 1 1 1 1 1 8 200836803

Gx -1 0 Γ -2 〇 2 1 1 1 1 iGl"Approximate M = |叫+ |办|

Gx and Gy are the X, Y changes of the image object, EGx=< and Σ Gy=〇, respectively. At this time, the pixel (pixei) of one pixel (Pixel) has the same value, which means there is no edge. Variety. Assuming that the left and right values are the same, it is guaranteed that the GX output will not be zero, but Gy will still be different from the value of the driving value.

ΛΛΛ J2 is 4 4ΛΛ _ I = G Using the above points, the approximation can be written as follows: Η = |(4 + 2χ^2 + Λμ(Λ + 2χΛ + Λ^+ \(Α3 + 2χΑ6 + Α9)-( Αι+2χΑ4 + ΑΊ] From the above point (kernel) will get the 4th picture, it can be clearly seen that the edge of this object (target) is clearly expressed. There are two basics in Gray Morphology. Operation Erosion and Dilatation (Dilati〇n). Two steps are required to complete both operations, an image called Activeimage is used to perform the image of the invasion or expansion, and can also be used in Figure 4. Seeing the effect of expansion, the advantage is that in 9 200836803, Edge Deteet_ can be seen that the edge of the tennis ball... is not enough to be highlighted, so it needs to be expanded to make subsequent image processing more convenient for calculation. Figure 5 is a graph corresponding to the flow of the process in Figure j. Step 5. The process of the reference, the figure of the figure 15〇, the figure of Figure and the annex I, the process of binarization (Threshoid), The appearance of the target 丨(9)

Shape 0 Binarization can also be called grayscale division (Th gift. M). Generally speaking, when processing graphics image binarization, the values are mostly concentrated at 28 and 225 to 255. Divide the grayscale division (Thresh〇ld) into two kinds of gray values, so that the gray value of the object image itself is higher than the set value, then make it a bright spot, and if the gray value is lower than the set value, then make it dark spot. The most frequently encountered problem is the distortion noise caused by the filtering of the image. Fortunately, this problem can greatly reduce these problems when spatial filtering is done after binarization. Assuming m is the threshold of binarization (Thresh〇lding Value), set the image grayscale score value to be: η mzzYJf{^y) /=1 where f: input image; η: number of all pixels ; f(x, y): Gray value of the pixel coordinate (X, y). The pixel values are divided into 〇 and 255 from 0 to 255, and the image pixels have only two gray levels of 0 and 255. Figure 5 is the prototype after binarization. 200836803 According to the experimental results, the original image file size is 75·5Κ, and the file size is reduced to 21.3Κ after binarization. Since the above-mentioned binarized image is easier to store, process and identify, the binary image signal processing is in (4) In image recognition processing, 'has a very important position. The reference Fig. 6 is a graph corresponding to the convex package processing corresponding to the flow 160 in Fig. 1. Step 6 · Process 160, Figure 6 and Annex I of Figure 1 of the photo

The convex hull (C〇nvexHu11) is processed to fill the internal vacancy of the object 1 〇〇 appearance. Referring to Fig. 7, a figure corresponding to the process of removing the external noise of the image corresponding to the flow 170 in Fig. 1 is used. Step VII. Referring to the process 17G, Fig. 7 and the attachment - in Fig. 1, the noise 101 on the periphery of the image object is removed (see Fig. 6). The reference Fig. 8 is a pattern corresponding to the process 180 of Fig. 1 which is subjected to the de-angular processing to cut off the sharp portion of the edge. Step 8: Referring to the process 180, FIG. 8 and Annex 1 of FIG. 1, the sharp portion of the edge of the object 100 is cut off. Referring to Fig. 9, the figure obtained by taking the center point, the X seat and the Y coordinate corresponding to the flow 190 in Fig. 1 is shown. Step 9: Refer to the process 190, Fig. 9 and Annex 1 of Figure 1, and find the center point 1〇2 of the target 1〇〇, and the X coordinate and the Y seat '' to determine whether it is pre- Set the sphere. See Fig. 10 for the statistical diagram after processing by process 120. Using the Color Threshold state of Visi〇n 11 200836803 The graph shows the pixel intensity distribution of the image. The grayscale and contrast of the entire image are seen from the graph. This state analyzes the chart. The color model used is HSI (Hue Saturation Intensity), and the horizontal axis represents grayscale values. Depending on the brightness or grayscale, from 0 to 255, the smaller the value. The darker the table, the brighter the opposite. The HSI can be used to analyze the image captured by the global CCD camera 230 (see Fig. 11) and detected via the edge, and the current grayscale distribution is treated as a post-binarization. Referring to Fig. 11, it is an external perspective view of the robot 200 for picking up the tennis ball. Referring to Figures 1 and 11, the identification method of the sphere of the present invention is composed of soft and hard bodies, and the foregoing describes the execution of the software. The hardware portion includes a microcomputer unit 210, an image capture card 220, and one mounted on the body of the robot 200. €0 camera 230 〇 When the CCD camera 230 captures an image of the object, such as the tennis 100' color, shape, size and fringe, the microcomputer unit 210 drives the CCD camera 230 lens as the tennis ball 100' moves, making this The tennis 100' is kept in the center of the FOV to properly capture the image of the tennis 100'. In summary, the method for identifying the sphere of the present invention, thereby determining the research of tennis, and the data obtained through the experiment, we set the identification of the object parameters to a similarity of 850 or more, high color sensitivity, and a moderate search strategy. It has been experimentally confirmed that the pre-set object (tennis 100,) can be quickly identified and the time taken is only 0.640 seconds, which is much lower than the 0.973 seconds required for the CCD to capture a picture. 12 200836803 In addition, because the pattern match in this system is the main function of tracking objects, the functions provided by Labview 7.1 will accurately identify objects and image templates. Although it has the advantages of precise comparison, it has relatively high requirements for light and identification of object images. In the experiment, it was found that the factors such as day and night, whether the window is covered by curtains and so on will affect the success rate of the model comparison. In particular, if the identified object is another ball, the model comparison fails, causing the _ 》 system to lose the object to be locked. Therefore, the identification method of the sphere of the present invention can obtain an accurate _ effect by using a robust calculation. In the present invention, the above-mentioned sphere 1 is used for picking up tennis balls. The method for identifying a sphere according to the present invention can also be used to pick up various spheres such as table tennis, golf balls, and the like. The present invention has been disclosed in the above-described embodiments, but it is not intended to limit the invention, and any person skilled in the art can make various changes and retouches within the spirit and scope of the present invention. The scope of the invention is defined by the scope of the appended claims. BRIEF DESCRIPTION OF THE DRAWINGS The above and other objects, features, advantages and embodiments of the present invention will become more <RTIgt; Step 13 of the identification method 200836803 Fig. 2 is a diagram of the image of the original image which is subtracted from the color of the grayscale image. The image of the original color obtained by the process 110 in Fig. 1 corresponds to the flow in the first figure. In the first diagram, the flow 130, 140 passes through the edge &gt; [the measured and expanded graphics.

The graph ^5 is a graph corresponding to the binarization of the flow 150 in Fig. 1. Fig. 6 is a diagram corresponding to the convex envelope processing of the flow 160 in Fig. 1 to remove the image external noise corresponding to the flow 170 in Fig. 1. Figure 8 is a graph corresponding to the process 180 of Figure 1 which has been subjected to a de-angular treatment to cut off the sharp portion of the edge. The figure 9 is a graph obtained by taking the center point, the X seat and the Y coordinate corresponding to the flow 19 in Fig. 1 . Figure 10 is a statistical diagram after processing by process 120. The third diagram is a stereoscopic view of the appearance of the sphere for the player who picks up the tennis ball. Annex I: Color pictures of 2nd, 3rd, 4th, 5th, 6th, 7th, 8th, 9th, and 1st. [Main component symbol description] 1G0: target 101: noise 102: center point 100': tennis 14 200836803 110: process 111: process 120: process 121: process 130: process 140: process 150: process 160: process 170: Process 180: Process 190: Process 200: Robot 210: Microcomputer Unit 220: Image Capture Card 230: CCD Camera 15

Claims (1)

200836803 X. Patent application scope: 1. A method for identifying a sphere for ingesting the shape of a target object and confirming whether it is a preset sphere. The method for identifying the sphere includes the following steps: Step (A): Capture the original image of the target; Step (B): Convert the original image into grayscale image data; Step (c): Use grayscale image data for Edge Detection to extract the outline of the target Step (D) ··Edge detection (Edge Detecti〇n) data, then expand to expand the outline of the target; Step (E) for binarization (Thresh〇ld) processing The appearance of the target is prototyped; Step (F): a convex hu11 treatment is performed to fill the internal vacancy of the appearance of the target; Step (G): removing the noise of the periphery of the image object; Step (H): Cut off the sharp part of the edge; and step (I) · Find the center point of the target, and the X coordinate and the Y coordinate respectively, to determine whether it is the preset sphere. 1. 2. For the patent application scope 1 Said Identification of the spheres, wherein, in step (A), the color-based object captured the original image. 3. The sphere of application of the method of identification of patentable scope of paragraph 1, which 16,200,836,803 In the step (B), the grayscale image is a grayscale image in green. 4. The method for identifying a sphere as described in claim 1, wherein the sphere identification method is used to identify tennis balls. 17