JP2021072048A - Image processing system and image processing method - Google Patents

Abstract

To provide an image processing technique capable of improving the accuracy to determine the presence or absence of defects based on ultrasonography images.SOLUTION: An image processing system 1 includes: an image acquisition unit 10 that acquires an image G based on reflection of ultrasonic wave U by transmitting the ultrasonic wave U to an object Q to be inspected; a development extraction unit 11 that extracts an appeared image E that appears in the image G according to the state that the object Q has; a vector extraction unit 12 that extracts a vector V whose start point and end point are two ends in the long-axis direction of the appeared image E; and a determination processing unit 13 that performs a series of processing to determine the cause of the appearance of the appeared image E based on the vector V.SELECTED DRAWING: Figure 2

Description

Embodiments of the present invention relate to image processing systems and image processing methods.

Conventionally, ultrasonic inspection technology has been widely used as a means for inspecting structures. It is the inspector who judges whether or not there is a defect from the cross-sectional image obtained by the ultrasonic inspection, and makes the judgment based on his expertise and experience. Therefore, the judgment result may vary depending on the skill of the inspector. Therefore, there is known a technique for automatically determining from ultrasonic inspection using signal processing or AI. For example, there is a technology that tracks the pixel movement of the output development (echo image) that appears in the image based on the reflection of ultrasonic waves, and automatically determines the output development that appears continuously in the images before and after the continuation as a defect candidate. Has been done. However, with this technique, there is a possibility that development due to the shape of the object to be inspected may appear continuously, and these developments may also be determined as defect candidates. In addition, in an ultrasonic image obtained from a layered object to be inspected, a technique is known in which a specific output development is deleted, an area is divided from the position of the remaining output development, and a layer boundary or an area of interest is automatically extracted. Has been done. Although the efficiency of the inspection process is expected to be improved by using this technique, it does not lead to the improvement of the variation in the determination of the presence or absence of defects.

Japanese Patent No. 4728762 JP-A-2018-190388

In the case of ultrasonic inspection of a welded portion, development occurs in the image based on various causes such as the shape of the inspection target, the structure of the welded portion, and internal objects in addition to the defect. Depending on how the ultrasonic waves are applied, the intensities or shapes of the plurality of output developments may be similar, or the output developments may be superimposed on each other. Therefore, there is a possibility that an erroneous judgment may be made because a difference cannot be obtained in the feature amount that is a reference for determining the presence or absence of a defect.

An embodiment of the present invention has been made in consideration of such circumstances, and an object of the present invention is to provide an image processing technique capable of improving the accuracy of determining the presence or absence of a defect based on an image of an ultrasonic inspection. And.

The image processing system according to the embodiment of the present invention has an image acquisition unit that transmits ultrasonic waves to an object to be inspected and acquires an image based on the reflection of the ultrasonic waves, and a state that the object has. Based on the vector, the output development extraction unit that extracts the output development appearing in the image, the vector extraction unit that extracts the vector having the two end points in the long axis direction of the output development as the start point and the end point, and the vector. It is provided with a determination processing unit that performs processing capable of determining the cause of the appearance of development.

An embodiment of the present invention provides an image processing technique capable of improving the accuracy of determining the presence or absence of a defect based on an image of an ultrasonic inspection.

The whole block diagram which shows the image processing system of this embodiment. A block diagram showing an image processing system. A cross-sectional view showing an object when the major axis direction of a defect is parallel to the incident direction of a vertically incident ultrasonic wave. FIG. 3 is a cross-sectional image showing an object corresponding to FIG. A cross-sectional view showing an object when the major axis direction of a defect is inclined with respect to the incident direction of vertically incident ultrasonic waves. FIG. 5 is a cross-sectional image showing an object corresponding to FIG. A cross-sectional view showing an object when the major axis direction of a defect is perpendicular to the incident direction of vertically incident ultrasonic waves. FIG. 6 is a cross-sectional image showing an object corresponding to FIG. 7. A cross-sectional view showing an object when the major axis direction of the defect is perpendicular to the incident direction of the ultrasonic wave incident at an angle of incidence. FIG. 9 is a cross-sectional image showing an object corresponding to FIG. A cross-sectional view showing an object when the major axis direction of the defect is inclined with respect to the incident direction of the ultrasonic wave incident at an angle of incidence. FIG. 6 is a cross-sectional image showing an object corresponding to FIG. Explanatory drawing which shows the state which surrounded the output development with a bounding box. An explanatory diagram showing a state in which a vector is extracted by a bounding box. Explanatory drawing which shows the state which extracted the edge of output development. Explanatory drawing which shows the state which the vector was extracted by the edge. Explanatory drawing which shows the angle between a virtual line extending in the incident direction of an ultrasonic wave and a vector. A cross-sectional image diagram showing a state in which output development due to defects and output development due to noise are specified. A cross-sectional image diagram showing a state in which the output development caused by defects is deleted and the output development caused by noise is left. A cross-sectional image diagram showing a state in which output development due to noise is deleted while output development due to defects is left. A cross-sectional image diagram showing a state in which development due to noise is highlighted. Top view showing the state of scanning the phased array probe along the welding line. Multiple cross-sectional image views used for image comparison processing. A flowchart showing an image processing method.

Hereinafter, embodiments of the image processing system and the image processing method will be described in detail with reference to the drawings. In the cross-sectional view, hatching is omitted to aid understanding.

Reference numeral 1 in FIG. 1 is the image processing system of the present embodiment. This image processing system 1 is used to transmit ultrasonic waves to an object Q to be inspected and to nondestructively inspect the inside of the object Q based on the reflection of the ultrasonic waves. For example, a cross-sectional image G (FIG. 18) of the object Q is generated based on the reflection of ultrasonic waves, and it is determined from this cross-sectional image G whether or not there is a defect inside the object Q. The presence or absence of defects may be determined by automatic processing of the image processing system 1 or by the user who visually recognizes the cross-sectional image G displayed by the image processing system 1.

The image processing system 1 includes a phased array probe 2 and an inspection computer 3. The phased array probe 2 is connected to the inspection computer 3. In this embodiment, the phased array probe 2 is used for electronic scanning. In addition, you may scan mechanically using a single probe.

The phased array probe 2 includes a plurality of probes 4. The probe 4 is a sensor incorporating an oscillator that transmits or receives ultrasonic waves. In this embodiment, a phased array probe 2 in which a plurality of probes 4 are arranged in a row is illustrated. A phased array probe 2 in which a plurality of probes 4 are arranged in a plane may be used.

The ultrasonic wave U is transmitted in a state where the phased array probe 2 is in contact with the surface of the object Q, and the reflected wave of the ultrasonic wave U is acquired by the phased array probe 2.

As shown in FIG. 3, the defect D1 of the object Q includes cavities, cracks or flaws formed inside or on the surface of the object Q. As shown in FIG. 4, in the ultrasonic inspection, the development E1 (echo) suspected to be the defect D1 appears in the cross-sectional image G of the object Q. That is, the development E1 appears in the cross-sectional image G according to the state of the object Q.

Next, the system configuration of the image processing system 1 will be described with reference to the block diagram shown in FIG.

The inspection computer 3 includes a display 5, an input operation unit 6, a storage unit 7, a communication unit 8, and a main control unit 9.

The display 5 displays the cross-sectional image G acquired by using the phased array probe 2. Further, the display 5 displays the determination result of the presence or absence of the defect D. Further, other information may be displayed on the display 5. That is, the image processing system 1 of the present embodiment includes a display device such as a display that outputs an image.

The input operation unit 6 is operated by a user who uses the inspection computer 3. The input operation unit 6 includes an input device such as a mouse or a keyboard. That is, predetermined information is input to the inspection computer 3 according to the operation of these input devices.

The storage unit 7 stores the cross-sectional image G acquired by using the phased array probe 2. Further, the storage unit 7 stores the determination result of the presence / absence of the defect D. Further, other information may be stored in the storage unit 7.

The communication unit 8 communicates with another computer via a communication line. The communication unit 8 may communicate with another computer via a WAN (Wide Area Network), an Internet line, or a mobile communication network. In addition, a LAN cable or a USB cable may be used to connect to another computer for communication. The image processing system 1 may be configured by using one inspection computer 3 or may be configured by using a plurality of inspection computers 3 connected to each other by a communication line.

The main control unit 9 comprehensively controls the image processing system 1. The main control unit 9 includes an image acquisition unit 10, a development extraction unit 11, a vector extraction unit 12, and a determination processing unit 13. These are realized by executing the program stored in the memory or the HDD by the CPU.

The image acquisition unit 10 generates a cross-sectional image G of the object Q based on the information acquired by the phased array probe 2. That is, the phased array probe 2 transmits the ultrasonic wave U to the object Q to be inspected, and the image acquisition unit 10 acquires the cross-sectional image G based on the reflection of the ultrasonic wave U.

As will be described later, the output development extraction unit 11 extracts the output development E (FIG. 18) that appears in the cross-sectional image G according to the state of the object Q. Some of the developed development E are caused by the defect D or some are caused by noise. In the image processing system 1 of the present embodiment, it is possible to distinguish between those caused by the defect D and those caused by noise.

As will be described later, the vector extraction unit 12 extracts (calculates) a vector V having two end points in the long axis direction of the development E as a start point T1 and an end point T2 (FIGS. 13 to 16).

As will be described later, the determination processing unit 13 performs processing capable of determining the cause of the appearance of the development E based on the vector V. The determination processing unit 13 includes a defect determination unit 14, a noise determination unit 15, a comparison processing unit 16, an automatic determination unit 17, and a display control unit 18.

As will be described later, the defect determination unit 14 compares the incident direction N of the ultrasonic wave U with the direction of the vector V, and determines whether or not the cause of the appearance of the development E is a defect. In this way, it is possible to determine whether or not the cause of the appearance of the development E is a defect based on the direction of the vector V.

As will be described later, the noise determination unit 15 compares the incident direction N of the ultrasonic wave U with the direction of the vector V, and determines whether or not the cause of the appearance of the development E is noise. In this way, it is possible to determine whether or not the cause of the appearance of the development E is noise based on the direction of the vector V.

As will be described later, the comparison processing unit 16 performs a process capable of comparing at least two adjacent cross-sectional images G among a plurality of positionally or temporally continuous cross-sectional images G. By doing so, since the presence or absence of the defect D is determined based on the plurality of cross-sectional images G, the determination accuracy can be improved.

As will be described later, the automatic determination unit 17 performs a process of determining the cause of the appearance of the development E. In this way, the image processing system 1 can automatically determine the cause of the development E.

The display control unit 18 performs a process of displaying the cross-sectional image G on the display 5. Further, the display 5 is controlled to display the determination result of the presence or absence of the defect D. Further, control for displaying other information on the display 5 may be performed.

The display control unit 18 performs a process of superimposing the vector V on the development E displayed on the display 5, for example. This makes it easier for the user to determine whether the development E is due to a defect or noise. In this way, the cause of development E can be manually determined.

That is, the display control unit 18 controls the image displayed on the display 5. The display 5 may be separate from or integrated with the main body of the inspection computer 3. Further, the display control unit 18 may control the image displayed on the display of another computer connected via the network.

The "process capable of determining the cause of the appearance of the development E" of the present embodiment includes a process of displaying the cross-sectional image G on the display 5 by the display control unit 18. Based on the cross-sectional image G displayed on the display 5, the cause of the appearance of the development E can be determined manually by the user. The user can input the result of the determination made by himself / herself by using the input operation unit 6.

Further, in the "process in which at least two adjacent cross-sectional images G among a plurality of cross-sectional images G that are continuous in position or time can be compared", the display control unit 18 can compare the plurality of cross-sectional images G. The process of displaying on the display 5 is included. Based on the cross-sectional image G displayed on the display 5, the cross-sectional image G can be manually compared by the user.

In this embodiment, the display 5 is illustrated as the display device, but other modes may be used. For example, a printer that prints information on a paper medium may be used instead of the display 5. That is, a printer may be included as an object controlled by the display control unit 18.

The inspection computer 3 of the present embodiment has hardware resources such as a CPU, ROM, RAM, and HDD, and when the CPU executes various programs, information processing by software is realized by using the hardware resources. It consists of a computer. Further, the image processing method of the present embodiment is realized by causing a computer to execute various programs.

Next, the image processing method implemented using the image processing system 1 of the present embodiment will be described in detail with reference to FIGS. 3 to 23.

As shown in FIG. 3, in the phased array probe 2, when ultrasonic waves U are simultaneously transmitted from a plurality of probes 4 (simultaneous excitation), the ultrasonic waves U are propagated in the vertical direction from the surface of the object Q. The phased array probe 2 can change the incident angle of ultrasonic waves by changing the timing (delay time) of transmitting ultrasonic waves U from a plurality of probes 4. For example, as shown in FIG. 9, when the ultrasonic waves U are sequentially transmitted in the order in which the plurality of probes 4 are arranged (time difference excitation), the ultrasonic waves U are propagated diagonally from the surface of the object Q. ..

The image acquired by using the phased array probe 2 may be a pixel image (planar image) or a voxel image (stereoscopic image). In the following description, a cross-sectional image G as a pixel image will be illustrated. The cross-sectional image G of the object Q is an image when viewed from a direction perpendicular to the incident direction N of the ultrasonic wave U. That is, the cross-sectional image G is an image when the object Q is virtually cross-sectioned along the incident portion of the ultrasonic wave U.

FIGS. 3, 5, and 7 show how the ultrasonic wave U at the defect D is reflected or scattered when the ultrasonic wave U is vertically incident on the object Q from the phased array probe 2. 9 and 11 show how the ultrasonic wave U at the defect D is reflected or scattered when the ultrasonic wave U is incident on the object Q from the phased array probe 2 at an angle of incidence.

A cross-sectional image G (FIGS. 4, 6, 6, 8) in which a predetermined range R is imaged by attaching a phased array probe 2 to the upper surface side of the object Q, transmitting an ultrasonic wave U, and obtaining a reflected wave thereof. 10 and 12) can be generated. In the cross-sectional image G, the ultrasonic wave U is reflected by the defect D and the lower surface of the object Q, so that the development E caused by the defect D and the development B caused by the lower surface of the object Q appear.

As shown in FIG. 3, when the major axis direction of the defect D1 is parallel to the incident direction N of the vertically incident ultrasonic wave U, the ultrasonic wave U transmitted from the predetermined probe 4 Is scattered at the upper end of the defect D1. Further, the ultrasonic wave U transmitted from the probe 4 next to it does not hit the defect D1 and is reflected on the lower surface side of the object Q. Therefore, the intensity of the ultrasonic wave U that is scattered or reflected by the defect D1 and is received by the phased array probe 2 is weak, and appears in the cross-sectional image G (FIG. 4) of the range R that is imaged based on the ultrasonic wave U. The image E1 is also weak and small.

As shown in FIG. 5, when the major axis direction of the defect D2 is inclined with respect to the incident direction N of the vertically incident ultrasonic wave U, even if the ultrasonic wave U hits the defect D2, the reflected wave is generated. It is difficult to return to the phased array probe 2. For example, when the defect D2 is tilted at 45 °, the ultrasonic wave U transmitted from the phased array probe 2 bends at a right angle to the incident direction and travels. Therefore, the development E2 appearing in the cross-sectional image G (FIG. 6) of the range R imaged based on the ultrasonic wave U is also weak and small. For example, two small undeveloped E2s appear scattered at the top and bottom of the defect D2.

As shown in FIG. 7, when the major axis direction of the defect D3 is perpendicular to the incident direction N of the vertically incident ultrasonic wave U, the reflected wave of the ultrasonic wave U hitting the defect D3 is phased. Return to the array probe 2. Further, if the defect D2 has a predetermined width, the reflected wave of the ultrasonic wave U transmitted by the other probe 4 can be obtained by the width. Therefore, the development E3 appearing in the cross-sectional image G (FIG. 8) of the range R imaged based on the ultrasonic wave U is strong and large.

Even if the shape has the same shape as the defects D2 and D3 shown in FIGS. 5 and 7, the output development E of the cross-sectional image G changes by changing the incident angle θ'of the ultrasonic wave U.

As shown in FIG. 9, when the ultrasonic wave U is incident at an angle perpendicular to the defect D2, the reflected wave of the ultrasonic wave U hitting the defect D2 returns to the phased array probe 2. Therefore, the development E4 appearing in the cross-sectional image G (FIG. 10) of the range R imaged based on the ultrasonic wave U is strong and large. The orientation of the output development E4 also appears strongly in the same orientation as the actual defect D2.

As shown in FIG. 11, even if the angle of incidence is incident, if the major axis direction of the defect D3 is inclined with respect to the incident direction N of the ultrasonic wave U, even if the ultrasonic wave U hits the defect D2. , It becomes difficult for the reflected wave to return to the phased array probe 2. Therefore, the development E5 appearing in the cross-sectional image G (FIG. 12) of the range R imaged based on the ultrasonic wave U is also weak and small.

In this way, the long axis direction, shape, and intensity of the developed development E change depending on the incident direction N of the ultrasonic wave U and the shape of the defect D. In particular, the long axis direction of the developed development E may be electrical noise or the like when the reflected wave of the ultrasonic wave U is physically oriented in a direction in which the reflected wave cannot be obtained. This is one useful feature.

Therefore, in the present embodiment, attention is paid to the fact that the reflection characteristic of the ultrasonic wave U changes depending on the incident direction N (FIG. 17) of the ultrasonic wave U and the direction of the defect D in the long axis direction, and the direction of the output development E in the long axis direction. The presence or absence of defects is determined using the vector V (FIG. 18) representing the above. The vector V of the present embodiment is a plane vector in the case of a pixel image (plane image) and a space vector in the case of a boxel image (stereoscopic image).

The output development extraction unit 11 performs a process of extracting the output development E included in the cross-sectional image G when specifying the major axis direction of the output development E. For example, there are a method of surrounding the development E with a bounding box 19 (FIG. 13) and a method of extracting the edge of the development E (FIG. 15). Then, the vector extraction unit 12 performs a process of extracting the vector V (FIGS. 14 and 16) based on the output development E extracted by the output development extraction unit 11.

When the bounding box 19 is used, as shown in FIG. 13, the development extraction unit 11 generates a bounding box 19 in which the development E is inscribed, and the bounding box 19 surrounds the development E. Do.

As shown in FIG. 14, the vector extraction unit 12 extracts the vector V for specifying the orientation of the development E in the long axis direction. For example, the long side L1 and the short side L2 of the bounding box 19 used for the extraction of the development E are specified. Then, among the four vertices of the bounding box 19, the vector V is calculated with the two vertices closer to the portion inscribed by the development E as the start point T1 and the end point T2, respectively. Of the two vertices (two ends), the one farther from the incident side of the ultrasonic wave U is set as the start point T1, and the closer one is set as the end point T2.

For example, assuming that the ultrasonic wave U is incident from above the paper surface of FIG. 14, the lower left apex away from the phased array probe 2 is the start point T1, and the upper right apex near the phased array probe 2 is the end point T2. .. In this way, the vector V from the start point T1 to the end point T2 is obtained.

When extracting the edge of the output development E, as shown in FIG. 15, the output development extraction unit 11 performs a process of extracting the edge 20 (contour) of the output development E. For example, the portion where the change gradient of the pixel value is equal to or greater than the threshold value may be set as the edge 20, or another method may be used.

As shown in FIG. 16, the vector extraction unit 12 extracts the vector V for specifying the orientation of the development E in the long axis direction. For example, the long side L1 and the short side L2 of the edge 20 used for the extraction of the development E are specified. Then, a vector V is calculated in which the two end points of the edge 20 in the long axis direction are the start point T1 and the end point T2, respectively. Of the two end points, the one farther from the incident side of the ultrasonic wave U is set as the start point T1, and the closer one is set as the end point T2.

In the extracted edge 20, the end point away from the phased array probe 2 is the start point T1, and the end point close to the phased array probe 2 is the end point T2. In this way, the vector V from the start point T1 to the end point T2 is obtained.

The display control unit 18 displays the vector V extracted by the vector extraction unit 12 on the display 5 in a manner of superimposing it on the corresponding output development E.

As shown in FIG. 17, the noise determination unit 15 corresponds to this vector V when the angle θ formed by the virtual line K extending in the incident direction N of the ultrasonic wave U and the vector V is in the range of 0 ° to 30 °. It is determined that the output development E to be performed is caused by noise. In this way, noise can be determined based on the angle θ formed.

For example, when the cross-sectional image G is acquired using the ultrasonic U of the longitudinal wave, if the incident angle θ with respect to the defect D of the ultrasonic U is 30 ° or less, most of the longitudinal wave is mode-converted to the transverse wave. Or, the reflected ultrasonic wave U does not return to the incident side, so that the reflection intensity of the longitudinal wave is lowered. That is, if it is caused by the defect D, the ultrasonic wave U is hardly reflected. Therefore, it is presumed that the undeveloped development E in which the above-mentioned angle θ formed in the range of 0 ° to 30 ° is not caused by the defect D but is caused by noise. The formed angle θ used for the determination is the smaller of the four formed angles θ formed at the intersection of the virtual line K extending in the incident direction N of the ultrasonic wave U and the vector V (inferior angle).

Further, the automatic determination unit 17 determines that the angle θ formed by the virtual line K extending in the incident direction N of the ultrasonic wave U and the vector V is in the range of 30 ° to 90 ° and has an intensity equal to or higher than a specific threshold value. It is determined that the output development E corresponding to this vector V is caused by the defect D.

The display control unit 18 may display the virtual line K extending in the incident direction N of the ultrasonic wave U on the display 5. Then, the cause of the appearance of the development E may be manually determined by the user based on the virtual line K displayed on the display 5. Further, the display control unit 18 may display the numerical value of the angle θ used for the determination on the display 5.

As shown in FIG. 18, when a plurality of development Es appear in the cross-sectional image G, each vector V of these developments E is extracted. Then, based on the vector V, it is determined whether the development E is caused by the defect D or the noise. For example, when displaying the cross-sectional image G on the display 5, post-determination image processing based on the determination is performed.

The noise determination unit 15 causes noise based on the direction of the vector V, for example, when the development E6 caused by noise and the development E7 caused by defect D appear in the cross-sectional image G. The output development E6 is specified.

For example, as shown in FIG. 19, the display control unit 18 may perform a process of deleting the output development E7 caused by the defect D and leaving the output development E6 caused by noise on the cross-sectional image G. Further, as shown in FIG. 20, the cross-sectional image G may be subjected to a process of removing the output development E6 caused by noise while leaving the output development E7 caused by the defect D. Further, as shown in FIG. 21, the cross-sectional image G may be subjected to a process of highlighting the development E6 caused by noise.

By performing these processes, it becomes easier to recognize the development E7 caused by the defect D. Further, the visibility can be improved by removing the development E6 caused by noise or a predetermined shape.

As shown in FIG. 22, when the phased array probe 2 is scanned along the surface of the object Q, a plurality of cross-sectional images G arranged along the scanning direction W are obtained. Here, the comparison processing unit 16 performs processing capable of comparing adjacent cross-sectional images G among a plurality of positionally continuous cross-sectional images G.

The size of the vector V of the output development E extracted by the vector extraction unit 12 represents the size of the defect D that is the source of the output development E. For example, in the inspection of the weld line 21 of the object Q, when scanned phased array probe 2, a plurality of cross-sectional images G that are sequentially acquired over the position P ₁ to P _N in the object Q is obtained .. The position _P 1 to P _N in FIG. 22 shows the positions arranged in the scanning direction W.

FIG. 23 (A) shows the output development E8 of the cross-sectional image G corresponding to the _{position P n-1.} FIG. 23B shows the output development E9 of the cross-sectional image G corresponding to the _{position Pn.} FIG. 23C shows the development E10 of the cross-sectional image G corresponding to the _{position P n + 1.} The position C in FIG. 23 indicates a predetermined position in a direction perpendicular to the scanning direction W.

For example, when the _{vector V is obtained for the output development E9 appearing in the cross-sectional image G corresponding to the position Pn} , the orientation and size of the output development E9 in the long axis direction can be found (FIG. 23 (B)). In the cross-sectional images G (FIGS. 23 (A) and 23 (C)) that are positioned before and after the cross-sectional image G, when the developed developments E8 and E10 are located at the same position C, the defect D is caused. It is presumed to be a thing.

In particular, in _{the case of the defect D that continuously extends from the position P n-1} to the position P _{n + 1} , the position does not change significantly, and the vectors V of the developed developments E8, E9, and E10 intersect with each other. .. In that case, it can be determined that the developments E8, E9, and E10 are caused by the defect D. On the other hand, if the positions of the _{defects D that continuously extend from the position P n-1} to the position P _{n + 1} have changed significantly, or if the vectors V do not intersect each other, the developments E8 and E9 , E10 can be determined to be due to noise.

In general, the size of the defect D that continuously spreads in a position such as a crack changes depending on the position, so that the size of the vector V can also change in the front and rear cross-sectional images G. By recording the information of the vector V of the cross-sectional images before and after this and observing the change from the continuous cross-sectional image G, it is possible to know the characteristics of the reflection source that is the source of the development E.

The cause of the appearance of the development E can be determined based on the comparison of these continuous cross-sectional images G. This determination may be performed by automatic processing by the automatic determination unit 17, or may be performed manually by a user who visually recognizes the cross-sectional image G displayed on the display 5.

When recording the vector V, a recording method such as recording as data corresponding to the coordinates of the cross-sectional image G or drawing a locus of change of the vector V on the cross-sectional image G may be used.

In the present embodiment, the cause of the appearance of the development E is determined by comparing the adjacent cross-sectional images G among the plurality of positionally continuous cross-sectional images G, but other embodiments may be used. .. For example, when making a judgment using a voxel image (stereoscopic image), the cause of appearance of development E is determined by comparing adjacent voxel images among a plurality of voxel images acquired continuously in time. You may.

Next, the image processing method executed by the image processing system 1 will be described with reference to the flowchart of FIG. The block diagram shown in FIG. 2 will be referred to as appropriate.

First, in step S11, the image acquisition unit 10 acquires the cross-sectional image G based on the information acquired by the phased array probe 2.

In the next step S12, the development extraction unit 11 extracts the development E (FIG. 18) that appears in the cross-sectional image G according to the state of the object Q.

In the next step S13, the vector extraction unit 12 extracts the vector V having the two end points in the long axis direction of the development E as the start point T1 and the end point T2 (FIGS. 13 to 16).

In the next step S14, the display control unit 18 executes the image display process. By this image display processing, the cross-sectional image G is displayed on the display 5. Here, the display control unit 18 displays the vector V extracted by the vector extraction unit 12 on the display 5 in a manner of superimposing it on the corresponding output development E.

In the next step S15, the automatic determination unit 17 of the determination processing unit 13 determines whether or not the development E is caused by the defect D based on the direction of the vector V. It should be noted that, based on the cross-sectional image G displayed on the display 5, it may be manually determined by the user whether or not the development E is caused by the defect D.

In the next step S16, the automatic determination unit 17 of the determination processing unit 13 determines whether or not the development E is caused by noise based on the direction of the vector V. It should be noted that, based on the cross-sectional image G displayed on the display 5, it may be manually determined by the user whether or not the development E is caused by noise.

In the next step S17, the comparison processing unit 16 executes the image comparison processing. By this image comparison processing, the adjacent cross-sectional images G among the plurality of cross-sectional images G that are continuous in position or time are compared.

In the next step S18, the automatic determination unit 17 of the determination processing unit 13 determines whether or not the development E is caused by the defect D based on the comparison of the plurality of cross-sectional images G. The user may compare a plurality of images displayed on the display 5 and determine whether or not the development E is caused by the defect D.

In the next step S19, the automatic determination unit 17 of the determination processing unit 13 determines whether or not the development E is caused by noise based on the comparison of the plurality of cross-sectional images G. The user may compare a plurality of images displayed on the display 5 and determine whether or not the development E is caused by noise.

In the next step S20, the display control unit 18 executes image processing after the determination. In this post-determination image processing, the display of the cross-sectional image G is updated based on the determination result. For example, a process is performed in which the output development E7 caused by the defect D is deleted and the output development E6 caused by noise is left (FIG. 19). A process of deleting the output development E6 caused by noise is performed while leaving the output development E7 caused by the defect D (FIG. 20). A process of highlighting the output development E6 caused by noise is performed (FIG. 21). Then, the process ends.

Although the flowchart of the present embodiment illustrates a mode in which each step is executed in series, the context of each step is not necessarily fixed, and even if the context of some steps is exchanged. good. Also, some steps may be executed in parallel with other steps.

The inspection computer 3 may include a computer having artificial intelligence (AI) for performing machine learning. Then, the artificial intelligence may determine the cause of the development E appearing in the cross-sectional image G. Further, the inspection computer 3 may include a deep learning unit that extracts a specific pattern from a plurality of patterns based on deep learning.

For the analysis using the computer of the present embodiment, an analysis technique based on learning of artificial intelligence can be used. For example, a learning model generated by machine learning by a neural network, another learning model generated by machine learning, a deep learning algorithm, a mathematical algorithm such as regression analysis can be used. Further, the form of machine learning includes forms such as clustering and deep learning.

The system of this embodiment includes a computer having artificial intelligence that performs machine learning. For example, the system may be configured by one computer provided with a neural network, or the system may be configured by a plurality of computers provided with a neural network.

Here, the neural network is a mathematical model that expresses the characteristics of brain function by computer simulation. For example, we show a model in which artificial neurons (nodes) that form a network by synaptic connection change the synaptic connection strength by learning and have problem-solving ability. In addition, neural networks acquire problem-solving abilities through deep learning.

For example, a neural network is provided with an intermediate layer having six layers. Each layer of this intermediate layer is composed of 300 units. In addition, by having a multi-layer neural network learn in advance using learning data, it is possible to automatically extract features in a pattern of changes in the state of a circuit or system. In the multi-layer neural network, an arbitrary number of intermediate layers, an arbitrary number of units, an arbitrary learning rate, an arbitrary number of learning times, and an arbitrary activation function can be set on the user interface.

A reward function may be set for various information items to be learned, and deep reinforcement learning for extracting the information item having the highest value based on the reward function may be used for the neural network.

For example, CNN (Convolution Neural Network), which has a proven track record in image recognition, is used. In this CNN, the intermediate layer is composed of a convolutional layer and a pooling layer. The convolution layer acquires a feature map by filtering nearby nodes in the previous layer. The pooling layer further reduces the feature map output from the convolutional layer into a new feature map. At this time, by obtaining the maximum value of the pixels included in the region of interest in the feature map, it is possible to absorb a slight deviation in the position of the feature amount.

The convolution layer extracts the local features of the image, and the pooling layer performs a process of summarizing the local features. In these processes, the image is reduced while maintaining the characteristics of the input image. That is, in CNN, the amount of information contained in the image can be significantly compressed (abstracted). Then, using the abstracted image stored in the neural network, the input image can be recognized and the image can be classified.

There are various methods for deep learning such as autoencoder, RNN (Recurrent Neural Network), RSTM (Long Short-Term Memory), and GAN (Generative Adversarial Network). These methods may be applied to the deep learning of the present embodiment.

The system of this embodiment includes a control device in which a dedicated chip, a controller such as an FPGA (Field Programmable Gate Array), a GPU (Graphics Processing Unit), or a CPU (Central Processing Unit) is highly integrated, and a ROM (Read Only). Storage devices such as Memory) or RAM (Random Access Memory), external storage devices such as HDD (Hard Disk Drive) or SSD (Solid State Drive), display devices such as displays, and input devices such as mice or keyboards. , With a communication interface. This system can be realized by a hardware configuration using a normal computer.

The program executed by the system of the present embodiment is provided by incorporating it into a ROM or the like in advance. Alternatively, the program may be a computer-readable, non-transient storage medium such as a CD-ROM, CD-R, memory card, DVD, or flexible disk (FD) with a file in an installable or executable format. It may be stored and provided in.

Further, the program executed by this system may be stored on a computer connected to a network such as the Internet, and may be downloaded and provided via the network. The system can also be configured by connecting separate modules that independently perform the functions of the components to each other via a network or a dedicated line, and combining them.

According to the embodiment described above, the presence or absence of defects is determined based on the image of the ultrasonic inspection by providing the vector extraction unit that extracts the vector whose start point and end point are the two end points in the long axis direction of output development. It is possible to improve the accuracy of the extraction.

Although some embodiments of the present invention have been described, these embodiments are presented as examples and are not intended to limit the scope of the invention. These embodiments can be implemented in various other forms, and various omissions, replacements, changes, and combinations can be made without departing from the gist of the invention. These embodiments and modifications thereof are included in the scope and gist of the invention, as well as in the scope of the invention described in the claims and the equivalent scope thereof.

1 ... image processing system, 2 ... phased array probe, 3 ... inspection computer, 4 ... probe, 5 ... display, 6 ... input operation unit, 7 ... storage unit, 8 ... communication unit, 9 ... main control unit, 10 ... Image acquisition unit, 11 ... Output development extraction unit, 12 ... Vector extraction unit, 13 ... Judgment processing unit, 14 ... Defect judgment unit, 15 ... Noise judgment unit, 16 ... Comparison processing unit, 17 ... Automatic judgment unit, 18 ... Display control unit, 19 ... Bounding box, 20 ... Edge, 21 ... Welding line, B ... Image development, C ... Position, D (D1-3) ... Defects, E (E1-10) ... Output development, G ... Cross section Image, K ... virtual line, L1 ... long side, L2 ... short side, N ... incident direction, P ... position, Q ... object, R ... imaged range, T1 ... start point, T2 ... end point, U ... super Sound wave, V ... vector, W ... scanning direction.

Claims (8)

An image acquisition unit that transmits ultrasonic waves to an object to be inspected and acquires an image based on the reflection of the ultrasonic waves.
A development extraction unit that extracts output development that appears in the image according to the state of the object, and a development extraction unit.
A vector extraction unit that extracts a vector having two end points in the long axis direction of development and development as a start point and an end point,
A determination processing unit that performs processing capable of determining the cause of the appearance of development based on the vector, and a determination processing unit.
To prepare
Image processing system. The determination processing unit includes a defect determination unit that compares the incident direction of the ultrasonic wave with the direction of the vector and determines whether or not the cause of the appearance of the development is a defect.
The image processing system according to claim 1. The determination processing unit includes a noise determination unit that compares the incident direction of the ultrasonic wave with the direction of the vector and determines whether or not the cause of the appearance of the development is noise.
The image processing system according to claim 1 or 2. The noise determination unit determines that the noise is when the angle formed by the virtual line extending in the incident direction of the ultrasonic wave and the vector is in the range of 0 ° to 30 °.
The image processing system according to claim 3. The image is a cross-sectional image of the object,
The determination processing unit includes a comparison processing unit that performs processing capable of comparing at least two adjacent cross-sectional images among a plurality of continuous cross-sectional images.
The image processing system according to any one of claims 1 to 4. The determination processing unit includes an automatic determination unit that performs processing for determining the cause of the appearance of development.
The image processing system according to any one of claims 1 to 5. The determination processing unit includes a display control unit that performs a process of superimposing the vector on the output development displayed on the display.
The image processing system according to any one of claims 1 to 6. A step of transmitting ultrasonic waves to an object to be inspected and acquiring an image based on the reflection of the ultrasonic waves, and
A step of extracting the output development appearing in the image according to the state of the object, and
A step of extracting a vector having two end points in the long axis direction of development and development as a start point and an end point,
A step of performing a process capable of determining the cause of the appearance of the development based on the vector, and
including,
Image processing method.

Images