KR101673056B1 - sorting method for grains - Google Patents

Abstract

A technique relating to a grain defect detection method is disclosed. In one embodiment, the grain defect detection method includes a grain preparation process for preparing a grain to be sorted, an image acquisition process for acquiring a digital image of the grain to be sorted, And a good article judgment process for judging whether or not the grain is good article. Wherein the quality determining step comprises the steps of: extracting brightness distributions of adjacent pixels according to the morphological deformation of the surface of the grain to be selected from the digital image of the obtained grain to be selected; The brightness distribution of each pixel is compared with a reference value extracted from the brightness distribution of each adjacent pixel according to the morphological deformation of the surface of the good grain previously extracted from the good grain to determine whether or not the selected grain is good.

Description

Methods for detecting grain defects {sorting method for grains}

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention generally relates to a grain defect detection method, and more particularly, to a grain defect detection method based on a change in brightness per adjacent pixel due to a morphological deformation of a grain surface.

Today, the development of the machinery industry has contributed to the automation of agricultural production, and the development of various sensor industries is accelerating the automation of agricultural product selection. Especially, the appearance of agricultural products is an important factor that determines its commerciality. Most of the screening works related to this have been handled through subjective judgment of the worker. Thus, not only was the screening inconsistent, but the results were unsatisfactory compared to the considerable labor consumption. To solve these problems, automated agricultural product sorting devices using various sensor devices have been developed and sold.

Rice sorting machine is a typical automation device of agricultural product appearance inspection which is popular in the market and is a method of measuring transparency by using photo-diode and CCD (charge coupled device). On the contrary, the fruits should be judged complexly considering internal factors such as sugar content and acidity, external factors related to size and color. In the case of grains, it is mainly determined by the appearance, but the selection through the complicated image processing is inefficient because the number of the grains is too large. Even in the case of the same kind of grain, the size, color and morphological characteristics of the target grain may differ due to the climate and regional characteristics of the producing region

Conventional methods for judging whether good or defective grains are available include a method of comparing the brightness component of the sample image to be filtered with a threshold value in the obtained image, a method of comparing the color information of the sample image to the threshold The method of comparing with the thresholds of red, green and blue, the method of comparing the size of the sample with the size of the prototype and the spectroscopic analysis using the near infrared (NIR) And a method of judging the state of the inside of the sample to be screened through the method. However, in the case of grains, the color, contrast distribution, size, and internal properties of the surface may be regarded as defective by the morphological deformation (wrinkled) of the sample surface even though they have the same properties as good products. These objects are difficult to select by the above methods.

In one embodiment, a method for detecting grain defects is disclosed. The method for detecting defects in grains includes a grain preparation process for preparing grain to be sorted, an image acquiring process for acquiring a digital image of the grain to be sorted, and a determination process for determining whether the selected grain is a good product from the digital image of the selected grain And a good quality judging process to be judged. Wherein the quality determining step comprises the steps of: extracting brightness distributions of adjacent pixels according to the morphological deformation of the surface of the grain to be selected from the digital image of the obtained grain to be selected; The brightness distribution of each pixel is compared with a reference value extracted from the brightness distribution of each adjacent pixel according to the morphological deformation of the surface of the good grain previously extracted from the good grain to determine whether or not the selected grain is good.

The foregoing provides only a selective concept in a simplified form as to what is described in more detail hereinafter. The present disclosure is not intended to limit the scope of the claims or limit the scope of essential features or essential features of the claims.

1 is a view showing a grain sorting apparatus according to an embodiment.
2 is a view for explaining good grain and defective grain.
FIG. 3 is a diagram for explaining a grain defect detection method using a histogram method utilizing brightness distribution of pixels extracted from a digital image of a grain.
FIGS. 4 and 5 are diagrams for explaining a grain defect detection method using a VS (valley size) method that utilizes differences in brightness values between adjacent pixels on a grain surface extracted from a digital image of a grain.
FIG. 6 is a view for explaining a grain defect detection method using the histogram method and VS method described above with reference to FIGS. 3 to 5. FIG.
7 is a flow chart illustrating a grain defect detection method according to one embodiment disclosed herein.

Hereinafter, embodiments disclosed in this specification will be described in detail with reference to the drawings. Like reference numerals in the drawings denote like elements, unless the context clearly indicates otherwise. The exemplary embodiments described above in the detailed description, the drawings, and the claims are not intended to be limiting, and other embodiments may be utilized, and other variations are possible without departing from the spirit or scope of the disclosed technology. Those skilled in the art will appreciate that the components of the present disclosure, that is, the components generally described herein and illustrated in the figures, may be arranged, arranged, combined, or arranged in a variety of different configurations, all of which are expressly contemplated, As shown in FIG. In the drawings, the width, length, thickness or shape of an element, etc. may be exaggerated in order to clearly illustrate the various layers (or films), regions and shapes.

It is to be understood that the singular " include " or " have " are to be construed as including the stated feature, number, step, operation, It is to be understood that the combination is intended to specify that it is present and not to preclude the presence or addition of one or more other features, numbers, steps, operations, components, parts or combinations thereof.

In each step, the identification codes (e.g., 110, 120, 130, ...) are used for convenience of explanation and the identification codes do not describe the order of the steps, Unless the specific order is described, it may occur differently from the order specified. That is, steps may occur in the same order as described, may be performed substantially concurrently, or may be performed in reverse order.

All terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the disclosed technology belongs, unless otherwise defined. Terms defined in commonly used dictionaries should be interpreted to be consistent with meaning in the context of the relevant art and can not be construed as having ideal or overly formal meaning unless expressly defined in the present application.

1 is a view showing a grain sorting apparatus according to an embodiment. 2 is a view for explaining good grain and defective grain. FIG. 3 is a diagram for explaining a grain defect detection method using a histogram method utilizing brightness distribution of pixels extracted from a digital image of a grain. FIGS. 4 and 5 are diagrams for explaining a grain defect detection method using a VS (valley size) method that utilizes differences in brightness values between adjacent pixels on a grain surface extracted from a digital image of a grain. FIG. 6 is a view for explaining a grain defect detection method using the histogram method and VS method described above with reference to FIGS. 3 to 5. FIG. 7 is a flow chart illustrating a grain defect detection method according to one embodiment disclosed herein.

Referring to FIG. 1, the grain sorting apparatus 100 includes a grain sending unit 10, a grain image obtaining unit 20, an image processing unit 30, a good grain sorting unit 40, and a control unit 50. In some other embodiments, the grain sorting apparatus 100 may further include a grain separating outlet 60 for selectively discharging the selected grain.

The grain conveyor 10 conveys the grains to be sorted to be defective and provides the grains to the grain image acquiring unit 20. The grain conveyor 10 may include a conveyor belt 12 on which the grain is conveyed and a feeder 14 for feeding the grain to the conveyor belt 12. [ In addition, the grain conveyor 10 may further include a vibrating portion 16 optionally. The amount of grains supplied to the conveyor belt 12 or the feeding speed of the grain can be adjusted by controlling the vibration of the vibrating unit 16 through the control of the control unit 50 according to a user's input command or a previously inputted command. In one embodiment, the conveyor belt 12 may include a transfer channel (not shown) that allows the grains fed from the feeder 14 to be spaced apart from one another. The conveyance channel may be provided in the form of a groove so that the grain can be seated on the conveyor belt 12. [ The shape of the groove may have various shapes such as U-shape, V-shape, and the like. As an example for the sake of understanding, the above example is not limited as long as the grain can be seated. Also, the number of the transport channels is not limited. If a plurality of feed channels are used, the number of grains that can be conveyed at one time through the conveyor belt 12 can be increased and the grain defect detection time can be shortened by processing them all at once through the grain sorting apparatus 100 have.

The grain image acquiring unit 20 acquires an image of the grain to be sorted, which is provided through the conveyor belt 12. The grain image acquiring unit 20 may include a scan camera 22, a front illuminator 24, and a rear illuminator 26. The scan camera 22 may be a facial scan camera or a line scan camera. The image of the grain obtained from the scan camera 22 may be a face scan image or a line scan image. The facial scan image may be obtained through the facial scan camera or may be obtained by sequentially scanning the target grain several times through the line scan camera. The backlight 26 can be used to easily separate the image of the grain obtained through the front illuminator 24 from the background.

The image processing unit (30) processes the image of the target grain obtained through the grain image acquiring unit (20). The image processing unit 30 may include a frame-grabber 32 and a processing unit 34 for acquiring images received from the scan camera 22. The processing unit 34 can extract the brightness value of each pixel, the position of the pixel column having a specific brightness, and the number of specific brightness values from the image of the grain to be sorted provided from the image obtaining unit 32. Through this, the processing unit 34 can extract the difference in brightness value between adjacent pixels of the grain surface from the image of the grain. It is also possible to extract values obtained by normalizing cumulative values of differences in brightness values of adjacent pixels on the grain surface and differences in brightness values of adjacent pixels on the grain surface. On the other hand, the grain sorting apparatus 100 calculates the cumulative value of the difference in brightness value between adjacent grains on the surface of the good grain extracted from the good grain in advance, the cumulative value of the difference in brightness value between adjacent grains on the surface of the good grain, And a database (not shown) in which at least one selected from a combination of the values is normalized. The information stored in the database may serve as a reference value (hereinafter, referred to as a reference value using a valley size (VS)) for determining whether the selected grain is good. In other words, the processing unit 34 calculates a brightness value difference between adjacent pixels on the surface of the grain to be selected, a brightness value per adjacent pixel on the surface of the grain to be selected, from the brightness distribution of adjacent grains on the surface of the grain to be selected A value obtained by normalizing an accumulation value of differences of brightness values of adjacent pixels on the surface of the grain to be sorted and a combination thereof can be extracted. The processing unit 34 can determine whether the selected target grain is a good product or a defective product by comparing the value extracted from the selected grain with the value stored in the database beforehand extracted from the good grain.

In the figure, four signal processors (DSPs) for processing 256 bytes are shown as an example of the processor 34. [ The above example is an example for understanding, and the number of signal processing apparatuses included in the processing section 34 can be changed according to the amount of the grain to be sorted, the discrimination speed of the desired grain to be sorted, and the like. In the drawing, a signal processing apparatus for processing input data of 256 bytes is shown as an example. The above example is an example for the sake of understanding, and the bytes of the input data that can be processed by the signal processing apparatus can be changed as needed.

The good grain selection unit 40 notifies the image processing unit 30 of whether the target grain is a good product or a defective product, and separates the selected grain into a good product and a defective product. The good grain selecting unit 40 may include an air-gun. For example, when the selected grain is judged to be defective by the image processing unit 30 from the image processing unit 30, the defective product can be sent to the defective product outlet of the grain separation outlet 60 by operating the air gun. In this case, the good grain selection unit 40 does not operate the air gun when the selection target grain is judged to be good from the image processing unit 30, and the grain to be selected judged as a good product falls freely into the grain separation outlet ( 60). Alternatively, unlike the case shown in the drawing, the non-defective grain selection unit 40 does not operate the air gun when receiving the determination that the selected target grain is defective from the image processing unit 30, It can fall freely to the defect outlet. In this case, when the good grain is judged to be good from the image processing unit 30, the good grain selection unit 40 operates the air gun to feed the grain to be selected, which is judged to be good, to the good outlet of the grain separation outlet 60 . Further, in the figure, a good grain discriminating device 40 including one air gun is shown as an example. As another example, unlike the case shown in the drawing, when a plurality of grains having a plurality of feed channels are simultaneously fed through the feeder 10, the good grain selector 40 corresponds to the number of the feed channels And a plurality of air guns. In this case, the good grain selecting unit 40 includes the air guns corresponding to the plurality of feed channels and the individual feed channels, so that the good and defective item selection amount for the selected grain can be improved. The air gun corresponding to the individual conveyance channels may be constituted by a plurality of air guns for easily conveying the selected grain to the grain separating outlet 60 selected as good or defective.

The control unit 50 may control operations of the grain conveyor 10, the grain image acquiring unit 20, the image processing unit 30, and the grains sorting unit 40. [ The control unit 50 may control the operation of the motor of the conveyor belt 12 to adjust the feeding speed of the grain to be sorted. In addition, the control unit 50 may control operations of the grain image acquiring unit 20 and the image processing unit 30. [ That is, the control unit 50 may operate the grain image acquisition unit 20 in consideration of the processing speed of the image processing unit 30. [ In addition, the control unit 50 can control the operation of the good grains sorting unit 40 in consideration of the processing speed of the image processing unit 30. For example, a predetermined period of time is required to determine whether the grain to be sorted through the image processing unit 30 is good or defective. In addition, the grain image acquiring unit 20 and the good grain discriminating unit 40 are disposed in the grain discriminating apparatus 100 at a predetermined distance. In this process, a certain amount of time may be required for the grain to be sorted to move from the grain image acquiring unit 20 to the good grain selecting unit 40. That is, the control unit 50 controls the grains image obtaining unit 20 and the grains screening unit 40 such that the time to delay the determination of good or defective products through the image processing unit 30 for the grain to be sorted, It is possible to control the operation of the good grain selecting unit 40 in consideration of the time taken to travel.

The grain separation outlet (60) can separate and discharge the selected grain. In this case, the grain to be sorted can be separated into good and defective products.

1 shows a grain sorting apparatus 100 as an example. It will be appreciated that the example of FIG. 1 may be used in various ways other than the manner illustrated in FIG. 1 as the grain sorting apparatus 100 as an example for the sake of understanding. Hereinafter, the operation of the grain sorting apparatus 100 will be described with reference to FIG. Referring to FIG. 1, the grain to be sorted which is fed through the feeder 10 can be dropped at the end of the feeder 10, that is, at the end of the conveyor belt 12. The image of the grain to be sorted falling through the grain image acquiring unit 20 can be acquired. In this case, the image of the grain to be sorted can be obtained through the line scan camera 22 having 1024 pixels. A line scan image for the selected grain can be obtained through the line scan camera 22. In addition, the surface scan image for the target grain can be obtained by sequentially scanning the falling target grain through the line scan camera 22 several times. The line scan image for the selected target grain obtained through the line scan camera 22 may be acquired through the image acquisition unit 32 and processed through the processing unit 34. [ In order to increase the image processing speed, the processing unit 34 may be composed of several signal processing devices. In the figure, a processing section 34 composed of four DSP (digital signal processor) devices is shown as an example. In this case, each of the signal processing apparatuses can perform image processing on 256 pixels out of 1024 pixels provided by the line scan camera 22. For example, when four pieces of grain to be sorted are fed through the feeding unit 10 at a time, an image of the grain to be sorted can be obtained through a line scan camera 22 having 1024 pixels. The image provided by the line scan camera 22 is divided and processed through the four signal processing apparatuses so that the image processing for the four sorting objects to be fed through the feeding unit 10, that is, the individual sorting target grains, And the signal processing apparatus can process each. Through this, it is possible to perform image processing for each selected grain, and at the same time, it is possible to analyze several selected grains, thereby improving the detection speed. The image processing unit 30 compares the analyzed analysis result of the image of the selected target grain with the good quality determination reference value stored in the database of the grain sorting apparatus 100 to determine whether the selected target grain is a good product . The image processing unit 30 provides the determination result to the good grain selecting unit 40. The good grain selecting unit 40 operates a good and defective product sorting device such as an air gun to distinguish good products from defective products. A method for comparing the analysis result of the image of the grain to be sorted through the image processing unit 30 and the analysis result of the analysis with the good judgment reference value stored in the database of the grain sorting apparatus 100 will be described below. A grain defect detecting method disclosed in this specification will be described using a grain image obtaining unit 20 using a line scan camera 22 as a grain image obtaining unit 20 for convenience of explanation.

Referring to FIG. 2, (a) of FIG. 2 shows a state of a grain corresponding to a good product, and (b) to (d) shows a state of a grain corresponding to a defective product. Fig. 2 (b) shows the defective grain by morphological deformation (wrinkled), (c) shows the defective grain due to the overall discoloration, and (d) It is appearance. In the present specification, a grain defect detection method of separating cereals corresponding to (a) of Fig. 2 from cereals corresponding to (b) to (d) of Fig. 2 is proposed.

FIG. 3 is a diagram for explaining a grain defect detection method using a histogram method utilizing brightness distribution of pixels extracted from a digital image of a grain. 3 (a) to 3 (d) are diagrams showing a profile of a contrast value for grain corresponding to FIGS. 2 (a) to 2 (d) (this is referred to as a histogram). The histogram-based grain defect detection method is a method for judging whether a grain to be sorted is good or defective on the basis of a reference value using a histogram obtained from a brightness distribution previously extracted from good grain. In other words, the grains image acquiring unit 20 is used to acquire images of the good grains. The image processing unit 30 provides information on a lightness value profile, that is, a lightness value distribution, in the image from the image of the acquired good grains. For example, the image processing unit 30 may provide information on the number of pixels having a specific brightness value as the lightness value profile of the non-defective grain. A method of determining whether a grain to be sorted is a good product or a defective product from the lightness value profile obtained from the image of the good grain will be described below.

First, Th_L, Th_H, and Th_D are determined from the histogram obtained from the digital image of the obtained good grain. Th_L is a threshold value for determining a boundary on the dark side, Th_H is a threshold value for determining a boundary on the bright side, and Th_D is a ratio of the number of pixels between Th_L and Th_H to the total number of pixels of the good grain obtained from the digital image . In the figure, TH_L, Th_H and Th_D, which are set in the range corresponding to 98% of the good grain, are shown as an example. Th_L, Th_H, and Th_D may be utilized as reference values using the histogram.

Then, the image processing unit 30 provided with the image of the grain to be sorted through the grain image acquiring unit 20 proceeds to determine the quality of the selected grain. The good judgment can be proceeded by the following process. First, the number of pixels satisfying the condition that is equal to or larger than Th_L and equal to or smaller than Th_H is added to GPs with respect to the pixel of the object to be sorted for each line image. Next, the number of pixels of the grain to be sorted detected for each line image is added to TPs. The above process repeats the process of adding the line image to the completion of the plane image for the selection target grain. Then, GPs are divided by TPs and stored in Ratio 1. If Ratio 1 is Th_D or more, it is good. If it is below Th_D, it is judged as bad. Here, GPs means the number of pixels satisfying the condition of the good product in the grain image to be sorted, and TPs means the total number of pixels of the grain image to be sorted.

The grain defect detection method of the histogram method utilizes the lightness value profile, and therefore, as shown in FIGS. 2 and 3, the grain to be sorted corresponding to (b) to (d) There is a problem that the possibility of being judged to be large is large. Particularly, in the case of the grain to be sorted which should be judged as defective due to the abnormality of the surface shape as in the case of FIG. 2 (b), the abnormality of the surface shape may cause the abnormality of the histogram. However, It often happens to have a contrast value profile similar to good. Therefore, in the present specification, a detection method based on a surface shape capable of solving or supplementing such a problem, that is, a method using a change in brightness per adjacent pixel due to morphological deformation of a grain surface (hereinafter referred to as brightness Quot; value difference ").

FIGS. 4 and 5 are diagrams for explaining a grain defect detection method using a VS (valley size) method that utilizes differences in brightness values between adjacent pixels on a grain surface extracted from a digital image of a grain. 4 (a) and 4 (b) show the defective grains 1 and the defective grain 2 separated from the background by the backlight 26, respectively. 4 (c) and 4 (d) show the brightness distribution (4) of the pixel obtained from the line scan image scanned along the AA 'line with respect to the good grains and the defective grains, respectively. The x-axis represents the relative position of each pixel in the line scan image, and the y-axis represents the brightness value of each pixel. In the drawing, a line scan image having 256 pixel brightness values is described as an example. 4 (e) and 4 (f) show the brightness distribution (4) graph of the pixel obtained from the line scan image scanned along the AA 'line with respect to the non-defective grain 1 and the defective grain 2, respectively, In this case, VD (4a, valley depth) represents the amount of water filled in the valley corresponding to the depth of each pixel with respect to each horizontal line (3). Fig. 5 is a diagram for explaining a method of calculating a valley size (VS) from VD (4a) and VD (4a). 4 (a) to 4 (d), it can be seen that the pixel brightness of the scanned line image varies depending on the surface shape of the object. In addition, the pixel brightness of the scanned line image may vary depending on the intensity of illumination applied to the object, the color of illumination, and the like. In addition, the pixel brightness of the scanned line image may vary depending on the color of the image extracted from the scanned line image. The color of the illumination applied to the object can be selected during the scan of the line image. The extraction of images having different colors from the scanned line images can be performed in the process of analyzing the scanned line images. For example, the pixel brightness distribution of the scanned line image obtained when the white light is provided to the object during the scanning process and the pixel brightness distribution of the scanned line image obtained when the red light is provided to the object may have different distributions. Accordingly, illumination having different colors depending on the surface shape, surface color, etc. of the object to be analyzed can be utilized. At least one illumination selected from red illumination, green illumination, blue illumination, and combinations thereof may be used as the illumination depending on the surface shape, surface color, etc. of the object to be analyzed. As another example, the color of the image may be selected in the course of image processing of the scanned line image through the image processing unit 30. In this case, the pixel brightness of the scanned line image obtained when the colors of the selected image are red images and green images, respectively, may have different values. At least one image selected from a red image, a green image, a blue image, and a combination thereof may be used as the image that can be extracted from the scanned line image according to the surface shape, surface color, etc. of the object to be analyzed.

As shown in FIG. 4 (c), in the case of the non-defective grain 1, the brightness of a pixel located closer to the center of the scanned line image has a larger brightness value, and a smaller brightness value toward the edge . ≪ / RTI > That is, in the case of the non-defective grain 1, the brightness distribution of pixels of the line image obtained from the line scan image scanned along the line AA 'takes the form of a quadratic curve. However, when the surface of the grain is not smooth as in the case of the defective grain (2), the brightness distribution of pixels of the line image obtained from the line scan image scanned along the AA 'line is irregular This takes the form of a large graph. The VS defect detection method disclosed in this specification focuses on the brightness distribution of pixels of a scanned line image different from that of the non-defective grains 1 exhibited by the defective grains 2. That is, the grain defect detection method disclosed in the present specification is a method for detecting the grain defect (1) and the defective grain (2) based on VD (4a) . 4 (c) and 4 (d), there is a large change in the brightness value of each pixel of the line image as the number of defective grains 2 increases, (d), as shown in the graph of FIG. Assuming that the valley is poured at the upper part of (c) and (d) of FIG. 4, the amount of water to be filled in the valley in case of the defective grain (2) It seems. In view of this point, the grain defect detection method disclosed in the present specification is a method for detecting the grain defect in the horizontal direction represented by (c) and (d) in FIG. 4 corresponding to the height of the water filled in the valley, The pixel-by-pixel VD 4a of the line image scanned on the line 3 is extracted. Thereafter, VS is calculated from the extracted VD (4a), and the VS is compared with VS extracted from the non-defective grains (1) to determine whether the selected target grain is good or defective. Normalized VS may be used in this process. Hereinafter, the grain defect detection method of the VS system will be described in detail with reference to FIG.

5 is a view for explaining a grain defect detection method of a VS (valley size) method. The method of detecting grain defect by VS method is a method of judging from the digital image of grains based on the difference of brightness value of adjacent grains on the surface of the grain to be sorted. 5 is a graph showing the brightness distribution of adjacent pixels on the surface of the grain to be selected, which is obtained from the line scan image. For convenience of description, a method of detecting a VS with respect to one line scan image will be described.

The image processing unit 30 can extract a brightness value difference between adjacent pixels on the surface of the grain to be selected from the digital image which is a line scan image of the grain to be sorted. From this, the image processing unit 30 can obtain VS. The process in which the image processing unit 30 obtains the VS from the difference in brightness value between adjacent pixels on the surface of the grain to be sorted may be performed, for example, as follows.

First, brightness distributions of adjacent pixels on the surface of the grain are extracted from the digital image of the obtained grain based on individual pixels of the digital image.

Next, a pixel having a maximum brightness value and a pixel having a brightness maximum value are extracted from the brightness distribution of the adjacent pixels on the extracted grain surface.

Hereinafter, pixels adjacent to each other among the pixels having a maximum value of brightness and the local maximum value of brightness are referred to as a first pixel and a second pixel, and a brightness value of the second pixel is defined as a brightness value of the first pixel Is greater than the brightness value of the light source. For example, L_LPR1 and L_LPRm shown in FIG. 5 may correspond to the first pixel and the second pixel, respectively. Next, there is a method in which a pixel having a maximum brightness value and a pixel having a brightness value equal to the brightness value of the first pixel are present between the pixels adjacent to each other, Three pixels are extracted.

Next, brightness values for extracting brightness values of pixels existing between the first pixel and the third pixel with respect to the brightness value of the first pixel on the basis of the brightness value of the first pixel The difference extraction process is performed. For example, when L_LPR1 and L_LPRm shown in FIG. 5 are referred to as the first pixel and the second pixel, respectively, L_LPR1 + w1 corresponds to the third pixel. A difference value between the brightness values of the pixels L_LPR1 to L_LPR1 + w1 is extracted based on the brightness value L_LP1 of the first pixel L_LPR1. Hereinafter, the difference value of the brightness value of each of the pixels existing between the first pixel and the third pixel with respect to the brightness value of the first pixel will be referred to as a brightness value difference. The difference value of the brightness value of each of the pixels existing between the first pixel and the third pixel with respect to the brightness value of the first pixel is expressed by VD (4a ). The brightness value of the first pixel corresponds to the brightness value corresponding to the horizontal line (3) shown in (c) and (d) of FIG.

Next, a summing process is performed after the brightness difference extraction process, and the sum of the brightness difference values is summed up. The sum of brightness differences corresponds to VS. The sum VS, which is the brightness value difference, can be normalized by dividing the number of pixels of the target grain in the digital image.

The above process may be performed on the entire surface of the grain to be sorted by changing the line, thereby obtaining the VS for the entire surface of the grain to be sorted. The VS obtained for the whole surface of the grain to be sorted can be normalized by dividing the number of pixels of the entirety of the grain to be sorted by the digital image corresponding to the entire surface of the grain to be sorted. In other words, the line scan image of the grain may be a surface scan image including a plurality of line scan images obtained for the entire grain surface. The summation may be performed for each of the plurality of line scan images of the grain. The sum of the brightness differences obtained from each of the plurality of line scan associations is added to each other through the summing process, and the sum of the brightness differences (hereinafter, referred to as cumulative brightness difference values) Dividing the digital image by the number of pixels of the digital image to normalize the cumulative brightness difference value with respect to the number of pixels of the digital image of the grain.

Finally, the process of determining whether the grain is good is performed by comparing the extracted cumulative brightness difference value or the normalized cumulative brightness difference value with a reference value using VS. Wherein the reference value using the VS includes a brightness distribution of the adjacent grains on the surface of the grains of the good that has been extracted in advance from the line scan image or the face scan image of the grains of the good grain, And a cumulative value of a brightness value difference in the distribution of the brightness per unit time, VS, a normalized VS, and a combination thereof. The image processing unit 30 directly utilizes the extracted brightness value difference or converts the extracted brightness value difference into VS or normalized VS in the above-described manner, and then converts the VS or the normalized VS into the reference value using the VS It is possible to determine whether the selected grain is a good product or a defective product. In addition, the image processing unit 30 may compare the brightness distribution of adjacent pixels on the extracted surface of the selected grain with the reference value using the VS to determine whether the selected target grain is good or defective.

The process of extracting VS using FIG. 5 will be described as an example. The image processing unit 30 extracts brightness distributions of adjacent pixels on the surface of the grain based on the individual pixels of the digital image from the digital image of the obtained grains. A pixel having a maximum value of brightness (Global Peak) and a pixel having a local maximum value of brightness (Local Peak) are extracted from the brightness distribution of the adjacent pixels on the extracted grain surface. In FIG. 5, GPR is a position of a pixel or a pixel in which a maximum value of brightness exists, and L_LPR1, L_LPRm, R_LPRn, and R_LPR1 are positions of pixels or pixels having a local maximum value of brightness. The location of the maximum value of the brightness is determined by setting the pixel having the largest brightness in the left side of the GPR to the pixel having the maximum local value (L_LPRm) based on the GPR, which is the position of the maximum value of brightness, The pixel having the largest brightness on the left side is selected as the pixel having the local maximum value based on the existing pixel L_LPRm. Through this process, a pixel (L_LPR1) whose local maximum value starts from the left side of the GPR is selected based on the GPR. In the case of the right side of the GPR, a pixel in which the local maximum value exists is set in the same manner. FIG. 5 represents a pixel having a maximum brightness value and a pixel having a maximum brightness value set in the above-described manner. 5, Hmin is the column position of the starting pixel at which the grain image starts in the line image, which is the digital image of the grain, Hmax is the column position of the last pixel at which the image of the grain ends in the line image, to be. In other words, the grain image is separated from the background and starts at Hmin and ends at Hmax.

From Fig. 5, VS can be obtained through the following equation. Equation (1) represents VS, and Equation (2) corresponds to a normalized VS divided by the total number of pixels of the grain image after summing VS obtained through a plurality of line scans.

Here, m and n are positive integers, which means the number of pixels having a local maximum value existing on the left and right sides based on GPR, respectively. wj is the width of the jth valley. In other words, wj is the number of pixels corresponding to the number of pixels having a brightness value less than or equal to the horizontal line (3) in the GPR direction when the brightness value of the pixel having the maximum value of the jth region is the horizontal line (3) Value. k is a positive integer and corresponds to the number of lines scanned or scanned lines. L_LP1, L_LRm, GP, R_LRn and R_LR1 denote the brightness values of the pixels located at L_LPR1, L_LPRm, GPR, R_LPRn and R_LPR1, respectively. And Pj is the brightness value of each pixel belonging to the width of the j-th valley.

(L_LPi - Pj) and (R_LPi - Pj) are the differences between the brightness values of the pixels located in the valley with respect to the brightness value of the horizontal line (3) having the maximum value in each region. Therefore, the VS for each line scanning image can be obtained by adding the differences of the brightness values from the respective line scanning images. In addition, normalized VS can be obtained by adding VS obtained from each line scanning image and dividing it by the number of pixels of the entire grain image.

Thereafter, the image processing section 30 extracts from the good grain in advance and compares the reference value using the VS stored in the database of the grain discriminating apparatus 100 or the memory of the image processing section 30 with the VS of the to-be-selected grain, It is possible to judge whether the selected grain is good or defective.

FIG. 6 is a view for explaining a grain defect detection method using the histogram method and VS method described above with reference to FIGS. 3 to 5. FIG. In the figure, a determination is first made as to whether a grain to be sorted is a good product or a defective product by a histogram method, and then detection using a VS method is further performed for those judged to be good products among the grain to be sorted. The histogram method and the VS method have been described above with reference to FIGS. 3 to 5, and a detailed description thereof will be omitted for convenience of explanation. This makes it possible to improve the accuracy of grain defect detection for the selected grain. Of course, it is natural that VS can detect grains defect only.

7 is a flow chart illustrating a grain defect detection method according to one embodiment disclosed herein. Referring to FIG. 7, the grain defect detection method 200 first prepares a grain to be selected through a grain preparation process (210). For example, the process of preparing the selected grain may include preparing a plurality of grains spaced apart from each other, and dropping the plurality of grains spaced apart from each other.

Next, a digital image of the selected grain is obtained through an image acquisition process (220). The digital image obtained through the image acquisition process 220 may be a line scan image of the selected target grain separated from the background image. In addition, the digital image of the selected target grain obtained through the image acquisition process 220 may include a color image. The color image may be extracted from at least one selected from a red image, a green image, a blue image, and a combination thereof through an image conversion process 250 to be utilized in the method of detecting a grain defect described herein. In one embodiment, the image acquisition process 220 may include providing illumination to the background of the plurality of selected target crops to be dropped, and acquiring digital images of the plurality of selected target crops falling during the provision of the illumination . ≪ / RTI > Whereby the digital image of the plurality of selection subject grains can be easily separated from the background image by the illumination.

Next, it is determined whether the selected target grain is a good product (230) from the digital image of the selected target grain obtained through the good product determination process. The good product determination process 230 may be configured to determine, from the digital image of the obtained target grain, the good quality determination process, from the digital image of the obtained target grain, brightness distribution of adjacent pixels according to the morphological transformation of the surface of the target grain And comparing the brightness distribution of adjacent pixels according to the morphological transformation of the extracted grain surface to the reference value extracted from the brightness distribution of adjacent pixels according to the morphological transformation of the surface of the good grain previously extracted from the good grain And determines whether the selected grain is a good product. For example, the digital image obtained through the image acquisition process may be a line scan image of the selected target grain separated from the background image. In this case, the quality determination process may include extracting the brightness distribution of adjacent pixels according to the morphological transformation of the surface of the grain to be selected on the basis of the individual pixels of the digital image from the digital image of the obtained selection target grain Extracting a pixel having a maximum value of brightness and a pixel having a maximum value of brightness from the brightness distribution of adjacent pixels according to the morphological transformation of the extracted grain surface of the extracted object; The first pixel and the second pixel are adjacent to each other among the pixels having the maximum value of the pixel and the brightness, and the brightness of the second pixel is greater than the brightness of the first pixel - extracting a third pixel having a brightness value equal to the brightness value of the first pixel, the brightness value of the first pixel being A difference value of brightness values of pixels existing between the first pixel and the third pixel with respect to the brightness value of the first pixel - hereinafter referred to as brightness difference difference, A summing process of summing the difference in brightness values, and a determination process of determining whether the selected target grain is good by comparing the sum of the brightness value differences with a reference value using the VS.

The reference value using the VS may be calculated from the brightness distribution of the adjacent pixels according to the morphological deformation of the surface of the good grain previously extracted from the line scan image of the good grain, the difference in brightness value between adjacent pixels according to the morphological deformation of the surface of the good grain, A VS, a valley size, a normalized valley size (VS), and a combination thereof, depending on the morphological deformation of the surface of the good grain. In addition, the reference value using the VS may be calculated from a cumulative brightness difference value (VS) using the difference in brightness value between adjacent pixels on the surface of the good product, which has been previously extracted from the face scan image of the good product grain, A normalized value of the cumulative brightness difference value using the brightness difference value, and a combination thereof. The good product determination process 230 has been described with reference to FIGS. 3 to 6, and a detailed description thereof will be omitted for the sake of convenience.

In one embodiment, the grain defect detection method 200 is performed after the summing process and divides the sum of brightness values by the number of pixels of the digital image of the grain to be sorted, And the normalization process of the brightness difference difference added to the number of pixels. In this case, the determining step may determine whether the selected target grain is good by comparing an average of the sum of the brightness differences obtained through the normalization process with a reference value using the VS. The detailed description thereof has been given above with reference to FIGS. 3 to 6, and will be omitted for convenience of explanation.

In another embodiment, in the grain defect detection method (200), the line scan image of the grain to be sorted is a surface scan image including a plurality of line scan images obtained for the entire surface of the grain to be sorted, The determining process 230 and the summation process may be performed for each of the plurality of line scan images of the grain to be sorted. In this case, the sum of the brightness difference values obtained from each of the plurality of line scan images through the summing process is added to each other, and the sum of the brightness difference values, hereinafter referred to as cumulative brightness difference values, Dividing the selected target grain by the number of pixels of the digital image of the target grain to normalize the cumulative brightness difference value with respect to the number of pixels of the digital image of the target grain. In this case, the determining step may determine whether the selected target grain is good by comparing the average of the cumulative brightness difference values obtained through the normalization process with a reference value using the VS. The detailed description thereof has been given above with reference to FIGS. 3 to 6, and will be omitted for convenience of explanation.

In some other embodiments, the grain defect detection method 200 may be performed before the good quality determination process 230 and may be performed from the digital image of the selected target grain obtained in the image acquisition process 220, Extracting a profile, comparing the lightness profile of the grain to be discriminated with a reference value using a histogram to determine whether the grain to be sorted is a good article, and removing grain from the grain to be discriminated (240). That is, the grain defect detection method 200 may further include a histogram type selection process performed before the good quality determination process 230. In this case, the reference value using the histogram can be obtained from the brightness / darkness profile extracted beforehand from the good product, that is, the histogram. The detailed description thereof has been given above with reference to FIGS. 3 to 6, and will be omitted for convenience of explanation.

In some further embodiments, the digital image of the selected subject matter obtained through the image acquisition process 220 may include a color image. The grain defect detection method 200 is performed before the good product determination process 230 and performs an image transformation process 250 for extracting at least one selected from a red image, a green image, a blue image, and a combination thereof from the color image . In this case, the good product determination process 230 may determine whether the selected target grain is a good product from the image extracted through the image transformation process 250. The detailed description thereof has been given above with reference to FIG. 4, and will be omitted for convenience of explanation.

The grain defect detection method 200 disclosed in this specification combines the VS method or the histogram method with the VS method, thereby solving the problem of the discrimination power of the grain defect detection method which was dependent on the conventional manual and mechanical sorting. In addition, we were able to automatically select the sorting criteria that were difficult to determine uniformly according to the user's criteria. In other words, the brightness distribution of the adjacent grains on the surface of the good grain previously extracted from the good grains selected by the user as a good product, the cumulative value (VS) of the difference in brightness value between adjacent grains on the surface of the good grain, The normalized value of the cumulative value of the difference in the brightness value of each star, and the combination thereof, is set as a reference value using VS, so that it can be adjusted to the user's criterion based on the grain defect selection criteria. Also, in order to select real time, several pieces of grain to be sorted can be supplied at a time through a plurality of feed channels, and the selected target grain is easily separated from the background through background illumination. In addition, the image processing of the surface morphology of the selected grain to be obtained for each channel is performed through independent signal processing apparatus, thereby enabling real-time sorting. Above all, using the VS method has improved the accuracy of the determination of grain defect detection. According to this advantage, the grain defect detection method 200 disclosed in this specification will be applied to a variety of grain grains.

From the foregoing it will be appreciated that various embodiments of the present disclosure have been described for purposes of illustration and that there are many possible variations without departing from the scope and spirit of this disclosure. And that the various embodiments disclosed are not to be construed as limiting the scope of the disclosed subject matter, but true ideas and scope will be set forth in the following claims.

Claims (9)

Cereal preparation process to prepare selected grain;
An image acquiring step of acquiring a digital image of the grain to be sorted; And
And determining whether the selected target grain is a good product from the digital image of the obtained selected target grain,
The digital image obtained through the image acquisition process is a line scan image of the selected target grain separated from the background image
The good-
Extracting brightness distributions of adjacent pixels according to the morphological transformation of the surface of the grain to be selected from the digital image of the obtained selection target grain;
Extracting a pixel having a maximum value of brightness and a pixel having a maximum value of brightness from the brightness distribution of adjacent pixels according to a morphological transformation of the extracted grain surface of the extracted object;
The first pixel and the second pixel are adjacent to each other among the pixels in which the maximum value of the brightness exists and the local maximum value of the brightness exist and the brightness value of the second pixel is the brightness of the first pixel A third pixel having a brightness value equal to the brightness value of the first pixel;
A difference value of a brightness value of each of pixels existing between the first pixel and the third pixel with respect to the brightness value of the first pixel on the basis of the brightness value of the first pixel, A brightness difference extraction process for extracting a brightness difference;
Summing the difference of brightness values; And
And comparing the sum of the brightness values with a reference value extracted from the brightness distribution of adjacent pixels according to the morphological deformation of the surface of the good grain previously extracted from the good grain to judge whether the selected grain is good or not Method of detecting defective grain. The method according to claim 1,
Wherein the step (b)
Further comprising the step of: normalizing the sum of the difference in brightness values with respect to the number of pixels of the digital image of the selection subject grain by dividing the summed brightness difference by the number of pixels of the digital image of the selection subject grain,
Wherein the determining step determines whether the selected target grain is good by comparing an average of the sum of the brightness differences obtained through the normalization process with the reference value. The method according to claim 1,
Wherein the line scan image of the grain to be sorted is a surface scan image including a plurality of line scan images obtained for the entire surface of the grain to be sorted,
Wherein the good product determination process and the summation process are performed for each of the plurality of line scan images of the grain to be sorted,
The sum of the brightness differences obtained from each of the plurality of line scan images through the summing process is added to each other, and the sum of the brightness difference values, hereinafter referred to as cumulative brightness difference values, Further comprising a step of normalizing the cumulative brightness difference value with respect to the number of pixels of the digital image of the target grain by dividing the number of pixels of the digital image of the target grain by the number of pixels of the digital image of the target grain,
Wherein the determining step determines whether the selected target grain is good by comparing an average of the cumulative brightness difference values obtained through the normalization process with the reference value. 4. The method according to any one of claims 1 to 3,
Wherein the reference value includes at least one of brightness distribution of adjacent pixels according to morphological deformation of the surface of the good grain previously extracted from the line scan image of the good grain, difference in brightness value between adjacent pixels according to morphological deformation of the surface of the good grain, At least one of a brightness value difference of adjacent pixels according to a morphological transformation of a grain surface, a VS, a valley size, a normalized valley size (VS), and a combination thereof, A cumulative brightness difference value using the difference in brightness value between adjacent pixels according to the morphological deformation of the surface of the good grain, and a cumulative brightness difference value using the difference in brightness value between adjacent pixels according to morphological deformation of the surface of the good grain. A normalized value, and a combination thereof. 4. The method according to any one of claims 1 to 3,
The grain preparation process
Preparing a plurality of selection subject grains to be placed apart from each other; And
And dropping the plurality of selection subject grains to be separated from each other,
The image acquisition process
Providing illumination behind the falling plurality of selected target crops; And
And acquiring a digital image of the plurality of selected target grains to be dropped in the course of providing the illumination,
Wherein the digital image of the plurality of selection subject grains is separated from the background image by the illumination. 4. The method according to any one of claims 1 to 3,
Wherein the digital image of the selected target grain obtained through the image acquisition step includes a color image,
Further comprising an image conversion step of performing at least one of a red image, a green image, a blue image, and a combination thereof from the color image,
Wherein the good quality determination process determines whether the selected target grain is good from the image extracted through the image conversion process. 4. The method according to any one of claims 1 to 3,
The method of claim 1,
Extracting a brightness value profile of the selected target grain from the digital image of the target grain obtained in the image acquiring process;
Determining whether the selected target grain is good by comparing the lightness profile of the extracted target grain with a reference value using a histogram; And
And removing grains judged to be defective among the grains to be sorted,
Wherein the reference value using the histogram is obtained from a profile of lightness and darkness extracted in advance from the good grain. Cereal preparation process to prepare selected grain;
An image acquiring step of acquiring a digital image of the grain to be sorted; And
And determining whether the selected target grain is a good product from the digital image of the obtained selected target grain,
Wherein the quality determining step comprises the steps of: extracting brightness distributions of adjacent pixels according to the morphological deformation of the surface of the grain to be selected from the digital image of the obtained grain to be selected; Determining whether the selected target grain is a good product by comparing the brightness distribution of each pixel with a reference value extracted from a brightness distribution of adjacent pixels according to morphological deformation of a surface of a good grain previously extracted from a good grain,
The method of claim 1,
Extracting a brightness value profile of the selected target grain from the digital image of the target grain obtained in the image acquiring process;
Determining whether the selected target grain is good by comparing the lightness profile of the extracted target grain with a reference value using a histogram; And
And removing grains judged to be defective among the grains to be sorted,
Wherein the reference value using the histogram is obtained from a profile of lightness and darkness extracted in advance from the good grain. Cereal preparation process to prepare selected grain;
An image acquiring step of acquiring a digital image of the grain to be sorted; And
And determining whether the selected target grain is a good product from the digital image of the obtained selected target grain,
Wherein the quality determining step comprises the steps of: extracting brightness distributions of adjacent pixels according to the morphological deformation of the surface of the grain to be selected from the digital image of the obtained grain to be selected; Determining whether the selected target grain is a good product by comparing the brightness distribution of each pixel with a reference value extracted from a brightness distribution of adjacent pixels according to morphological deformation of a surface of a good grain previously extracted from a good grain,
Wherein the reference value includes at least one of brightness distribution of adjacent pixels according to morphological deformation of the surface of the good grain previously extracted from the line scan image of the good grain, difference in brightness value between adjacent pixels according to morphological deformation of the surface of the good grain, At least one of a brightness value difference of adjacent pixels according to a morphological transformation of a grain surface, a VS, a valley size, a normalized valley size (VS), and a combination thereof, A cumulative brightness difference value using the difference in brightness value between adjacent pixels according to the morphological deformation of the surface of the good grain, and a cumulative brightness difference value using the difference in brightness value between adjacent pixels according to morphological deformation of the surface of the good grain. A normalized value, and a combination thereof.

Images