JPWO2012108278A1 - X-ray inspection apparatus and method - Google Patents

Abstract

An X-ray irradiation apparatus 10 that irradiates the object 1 with X-rays from a plurality of directions and captures a plurality of X-ray transmission images with parallax, and an image processing apparatus 20 that inspects the object 1 from the plurality of X-ray transmission images. With. For example, the image processing apparatus using the evolution calculation method determines the correspondence between the edges of the object 1 and both ends of the edges of the object 1 (S2 to S7), and then three-dimensionally measures the position of the object 1 (S8). ).

Description

  The present invention relates to an X-ray inspection apparatus and method for inspecting an object based on a plurality of X-ray images having large parallax.

  The detection of corresponding points on each image based on a plurality of images with parallax is called “stereo matching”. In addition, an algorithm that generates a new individual by operations such as natural selection, crossing, and mutation on an individual group and searches for an individual that meets the purpose is called a “genetic algorithm”.

  Stereo matching based on an X-ray image is disclosed in Patent Document 1, general stereo matching is disclosed in Patent Documents 2 and 3, and genetic algorithms are disclosed in Patent Documents 4 to 6, for example.

  For example, instead of a general visible light image, an X-ray transmission image is used, and two stereo X-ray transmission images captured with parallax close to human eyes are presented by a stereo viewer or the like, and a stereoscopic image ( A device that can be viewed as an image whose depth can be perceived when viewed through is already disclosed in Patent Document 1.

JP-A 64-2628, “stereoscopic X-ray diagnostic apparatus” Japanese Unexamined Patent Publication No. 2000-99760, “3D Object Model Generation Method and Computer-Readable Recording Medium Recording a 3D Object Model Generation Program” JP 2010-44438 A, “Feature Extraction Device, Feature Extraction Method, Image Processing Device, and Program” JP-A-10-21385, “Image processing method and apparatus” JP-T-2004-038660, “Image processing method and image processing apparatus” JP 2005-346297 A, “Solid Object Recognition Device”

FIG. 1 is a principle diagram of the stereo measurement method.
As shown in this figure, a method for measuring the distance to the measurement point P of the object and the unevenness of the object based on the association between a plurality of (with parallax) images has been established as a stereo measurement method. In this case, it is necessary to obtain the corresponding point P _R in the right image of the measurement point P _L by the left image. This is called stereo matching.

FIG. 2 is a schematic diagram of a block matching method which is an example of stereo matching.
The block matching method is a typical means of stereo matching, and an image is cut out with a window of a certain size centered on the attention point (now P _L ) of the left image, and the template is used as the right image. By moving left and right above, a point (P _{R in} this case) that matches the pattern is obtained as a corresponding point, and this is performed for all points.
This method is simple, and it is relatively easy to determine the corresponding points for characteristic points such as the vertex portions in FIG. However, it is difficult to correctly determine corresponding points such as non-characteristic points and hidden points on the other hand. For this reason, there is a problem that a lot of noise is included in the result of distance measurement based on the block matching method.

FIG. 3 is a diagram illustrating a positional relationship between the object and the viewpoint.
In this figure, for example, when two images having a large parallax such as the viewpoint A and the viewpoint B are given, a technique for synthesizing an image at a viewpoint (for example, the viewpoint C or D) between them is mainly a multi-view TV. Has been studied as a method for.
However, the reliability of the image generated at this time decreases in proportion to the magnitude of the parallax between the viewpoints A and B. For this reason, if the parallax of the original images (viewpoints A and B) is too large, there is a problem that the conventional stereo matching fails due to excessive deformation of the figure and a good intermediate image cannot be created.

4A is a diagram illustrating an image when the parallax is large and the deformation is small, and FIG. 4B is a diagram illustrating an image when the parallax is large and the deformation is large.
As shown in this figure, even with the same “cylindrical can”, in the case of the arrangement shown in FIG. 4A, the deformation is small by chance, but in the arrangement shown in FIG.
Therefore, when the parallax is much larger than that of a normal stereo image, the object is often greatly deformed in the left and right images, and corresponding points are determined in the stereo matching based on the pattern as shown in FIG. I couldn't.

  Also, X-ray CT (Computed Tomography) is known as means for constructing a cross-sectional image of an object from an X-ray image. However, when X-ray CT is applied to X-ray baggage inspection etc., since the operating part (or rotating part) exists in the X-ray source and detector, the whole apparatus becomes large, processing becomes complicated, processing time becomes long, There was a problem that the device was expensive.

  The present invention has been developed to solve the above-described problems. That is, the object of the present invention is based on a plurality of X-ray images having large parallax, which makes conventional stereo matching difficult or substantially impossible, and is small and inexpensive, with simple processing and in a short time. It is an object of the present invention to provide an X-ray inspection apparatus and method that can inspect the above.

According to the present invention, an X-ray irradiation apparatus that irradiates an object with X-rays from a plurality of directions and captures a plurality of X-ray transmission images with parallax;
An image processing device for inspecting the object from the plurality of X-ray transmission images,
There is provided an X-ray inspection apparatus characterized in that an image processing apparatus determines correspondence between edges of an object, both ends of the edge of the object, and then three-dimensionally measures the position of the object.

According to the present invention, (A) an object is irradiated with X-rays from a plurality of directions, and a plurality of X-ray transmission images with parallax are captured,
(B) From the plurality of X-ray transmission images, the image processing device determines correspondence between the edges of the object and both ends of the edge of the object,
(C) Next, an X-ray inspection method is provided, in which the position of an object is three-dimensionally measured.

According to the apparatus and method of the present invention, from a plurality of X-ray images, for example, by an image processing apparatus using an evolutionary calculation method, (1) association of an edge of an object and (2) edge of an object Since both ends are determined, even when parallax is large and conventional stereo matching is difficult or practically impossible, it is easy to associate the edges of the object and determine both ends of the object edge. Processing can be done in a short time.
Therefore, the three-dimensional measurement of the position of the target object based on the result can determine the three-dimensional image of the target object, the distance measurement of points in the target object, the front-rear relationship of each target object, the distance from the floor surface, and the like. So the object can be inspected.

In addition, the apparatus of the present invention can be configured only by adding an X-ray source and a detector to the longitudinal direction of the apparatus (conveyor traveling direction) as compared to a conventional apparatus having one X-ray source and one detector. Compared to X-ray CT, it can be made smaller and less expensive.

It is a principle diagram of the stereo measurement method. It is a schematic diagram of the block matching method which is an example of stereo matching. It is a figure which shows the positional relationship of a target object and a viewpoint. It is a figure which shows an image when parallax is large and a deformation | transformation is small. It is a figure which shows an image in case a parallax is large and a deformation | transformation is large. It is a typical perspective view of the X-ray inspection apparatus by this invention. It is a layout view of the upstream X-ray source 16a and the detector 18a. It is the arrangement | positioning figure seen from the upstream of the X-ray source 16b and the detector 18b of the downstream side. It is a whole flowchart of the X-ray inspection method by this invention. It is a schematic diagram which shows the acquired left image (X-ray transmission image). It is a schematic diagram which shows the acquired right image (X-ray transmission image). It is the left image which cut out the image of the same height (line) from FIG. 7A. It is the right image which cut out the image of the same height (line) from FIG. 7B. It is an image figure of the generation change in (1 + λ) ES. It is a schematic diagram which shows the feature-value which an edge has. This coefficient is applied to the vicinity of the differential filter fx in the horizontal direction in order to detect the vertical edge. This coefficient is applied to the vicinity of the differential filter fy in the vertical direction in order to detect the edge in the horizontal direction. It is a figure which shows the edge of the both ends of the same target object.

  Hereinafter, preferred embodiments of the present invention will be described with reference to the drawings. In each figure, common portions are denoted by the same reference numerals, and redundant description is omitted.

5A is a schematic perspective view of the X-ray inspection apparatus according to the present invention, FIG. 5B is an arrangement diagram of the upstream X-ray source 16a and the detector 18a, and FIG. 5C is from the upstream side of the downstream X-ray source 16b and the detector 18b. FIG.
As shown in this figure, an X-ray inspection apparatus according to the present invention includes an X-ray irradiation apparatus 10 that irradiates an object 1 with X-rays from a plurality of directions and picks up a plurality of X-ray images with parallax, and a plurality of X-ray irradiation apparatuses 10. And an image processing device 20 for inspecting the object 1 from the X-ray image.
The X-ray irradiation apparatus 10 includes a conveyor device 12 that horizontally conveys the object 1, a shielding device 14 that surrounds a part of the conveyor device 12, and a pair of X-ray sources 16 a, 16 b and 1 provided inside the shielding device 14. A pair of detectors (X-ray detectors) 18a and 18b are provided.

In this example, the first X-ray source 16 a and the first detector 18 a are provided on the upstream side of the shielding device 14 and capture a first X-ray transmission image of the object 1. The second X-ray source 16b and the second detector 18b are provided on the downstream side of the shielding device 14 and capture a second X-ray transmission image having a large parallax with respect to the first X-ray transmission image of the object 1.
In this example, two sets of X-ray sources and detectors are provided, but three or more sets may be arranged.

Here, “large parallax” means parallax for which conventional stereo matching is difficult or substantially impossible. Conventional parallax that can be stereo-matched is, for example, parallax from two points separated by an interval corresponding to human eyes. On the other hand, “large parallax” in the present application means parallax having an angle with respect to an object of, for example, 30 ° or more and 90 ° or less.
Note that the present invention is particularly useful for such “large parallax”, but is not limited thereto, and may be parallax capable of conventional stereo matching.

  The above-described X-ray inspection apparatus of the present invention only adds an X-ray source and a detector to the longitudinal method of the apparatus (conveyor traveling direction) as compared with the conventional apparatus having one X-ray source and one detector. Due to the configuration, there is a feature that cost and space can be saved as compared with X-ray CT.

The image processing device 20 is, for example, a PC (computer) including a storage device, a calculation device, and a display device. From a plurality of X-ray transmission images, an object is associated with an edge of an object by an evolutionary calculation method. Both ends of the edge are determined, and then the position of the object is three-dimensionally measured.
The “evolution calculation method” means a method for obtaining a target optimum solution including an evolution strategy, evolutionary programming, and a genetic algorithm.
Details of processing in the evolutionary calculation method will be described later.

  FIG. 6 is an overall flowchart of the X-ray inspection method according to the present invention. In this figure, the X-ray inspection method of the present invention comprises steps (processes) S1 to S10.

In S <b> 1 of FIG. 6, the X-ray irradiation apparatus 10 described above is used to irradiate the target 1 with X-rays from a plurality of directions, thereby capturing a plurality of X-ray transmission images having parallax. Hereinafter, the X-ray transmission image is simply abbreviated as “image” unless particularly necessary.
A case where the plurality of images are two images on the left and right will be described.

  FIG. 7A is a schematic diagram showing the acquired left image (X-ray transmission image), and FIG. 7B is a schematic diagram showing the acquired right image (X-ray transmission image). In each figure, A to E indicate a plurality of objects 1.

  In S2 of FIG. 6, images on the same line are cut out from the acquired left and right images. Here, the “same line” is a common area on the left and right images, and means, for example, a rectangular area corresponding to the same height.

7C is a left image obtained by cutting out an image with the same height (line) from FIG. 7A, and FIG. 7D is a right image obtained by cutting out an image with the same height (line) from FIG. 7B.
In FIG. 7C and FIG. 7D, numerals (1 to 10) in the drawings indicate corresponding edges (contour lines) of the object 1 in each image.

  In S3 to S8 in FIG. 6, the evolution calculation method is used to determine the correspondence between the edges of the object 1 between the images and the both ends of the edge of the object 1 in each image (S3 to S7). The position of 1 is three-dimensionally measured (S8).

  Hereinafter, the evolution calculation method according to the present invention will be described.

1. Individual description in evolutionary computation (S3)
As shown in FIGS. 7C and 7D, L1 to Lm are edges on a certain line in the left image (FIG. 7C), and R1 to Rn are edges on the same line in the right image (FIG. 7D). Table 1 shows the correspondence between edges, and Table 2 shows the correspondence between edges in the left image.

Table 1 is an m × n bit table, and Table 2 is an m × (m−1) / 2 bit table, and these tables are the solutions to this problem, that is, chromosomes representing one individual in the evolutionary calculation method. Think.
That is, the chromosome is mxn + mx (m-1) / 2 = mx (2n + m-1) / 2 bits in total. For example, if m = n = 10, one solution is represented by 10 × (20 + 10−1) / 2 = 5 × 29 = 145 bits.

This chromosome is regarded as a solution only when all of the following constraint conditions (chromosome conditions 1 and 2) are satisfied.
(Chromosome condition 1)
In Table 1, the correspondence between the left and right is either 1: 1 or 1: 2. That is, an edge having no correspondence and a one-to-many correspondence of one to three or more are not allowed.
(Chromosome condition 2)
In Table 2, the correspondence on the right side of the edge is either one-to-one or one-to-two. That is, an edge having no correspondence and a one-to-many correspondence of one to three or more are not allowed.

2. Evolution method (generation change method, offspring generation method)
The evolution method used in the present invention is basically positioned as (1 + λ) ES (ES: Evolution Strategy), which is a kind of evolution calculation method.

FIG. 8 is an image diagram of generational change in (1 + λ) ES.
1 of (1 + λ) ES represents a parent individual, and λ represents a child individual created by changing a part of the parent individual using a random number. Then, λ child individuals are created from one parent individual, and one optimal individual among (1 + λ) individuals is selected to be the final next-generation individual.
Although λ is 4 in this example, the optimum value of λ is preferably determined by examining the statistical characteristics of the space to be searched.

A parent individual is generated in S3 of FIG. 6, and a child individual is generated in S4.
The specific method of individual generation is shown below. For the generation of offspring (S4), only mutation is used in the present invention.

(Individual generation procedure)
(1) A certain ratio (this ratio is referred to as “mutation rate”) among genes (0 or 1, in total m × (2n + m−1) / 2) in the chromosome of the original individual A (parent individual) The individual B (child individual) is generated by randomly selecting and inverting the bit (referred to as 0).
(2) It is examined whether the bit string in the chromosome of the individual B satisfies both (chromosome condition 1) and (chromosome condition 2) described above. If so, individual B is considered one of the official offspring. If either or both of the conditions are not satisfied, the individual B is regarded as “unanswered” and rejected, and the process returns to (1) to generate another individual B. (2) is repeatedly executed until an individual that can be a “solution” is generated.
(3) When a plurality of individuals are generated, the necessary number of child individuals are generated by repeating the processes (1) and (2) as many times as necessary (corresponding to λ, this time 4 times).
Even when one initial individual (first individual) is created, an individual that satisfies the above two conditions is randomly generated and used as the initial individual (parent individual).

3. Fitness Function In the present invention, the fitness function fitness is represented by a linear sum of _three evaluation values (f _{1 to} f ₃ ) as in the following equation (1).
fitness = f ₁ + f ₂ + f ₃ (1)

Here, f _{1 to} f ₃ are evaluation values based on the following evaluation criteria.
(F ₁ : Evaluation criteria 1) Evaluation values for Table 1. This is based on the validity of the correspondence between the left and right edges that are “corresponding” in Table 1.
(F ₂ : Evaluation Criteria 2) Evaluation values related to Table 2. This is based on the reliability of the edge pair in the left image that is “corresponding” in Table 2.
(F ₃ : Evaluation criteria 3) Based on the reliability of the edge pair in the right image determined from Tables 1 and 2.

In f ₁ , the corresponding edge is “A: is it valid as the left (or right) edge of the same object?”
In f _2, the corresponding edge: that the "B? reasonable or as left and right edge (= both ends) of the same object",
In f _3, the corresponding edge is: what to evaluate that "B reasonable or as left and right edge (= both ends) of the same object?".
A specific calculation method of the _three evaluation values (f _{1 to} f ₃ ) will be described below.

3.1 Calculation of “A: Valid for the left (or right) edge of the same object?” This is related to the validity of Table 1, with the left edge (m) and the right edge (n) Evaluate the validity of the correspondence.
Here, it is assumed that m = n is not always satisfied. This is because it is considered that there is a case where there is an omission of extraction at one of the edges for some reason, or there are cases where the ends of two objects are overlapped and the number of edges is seemingly small.
In this evaluation, either may be used as a reference, but here, the evaluation value of the degree of coincidence between the left and right edges is calculated from the expression (2) of Formula 1 with the left image as a reference.

Here, Li represents the i-th edge of the left image, and R (Li) represents the edge of the right image to which Li corresponds. S ₁ (Li, R (Li)) is the similarity when the edges Li and R (Li) are the corresponding edges, and takes a value from 0 to 1, and the closer to 1, both The reliability to which the edge corresponds is considered high.

Regarding the calculation method of S ₁ (Li, R (Li)), when there is only one right edge corresponding to the left edge (one-to-one case) and when there are two right edges corresponding to the left edge ( (1 to 2)

(1 to 1)
FIG. 9 is a schematic diagram illustrating the feature amount of the edge.
First, four feature amounts of the edge direction, intensity, and left and right tone values (X-ray transmission density values) shown in FIG. 9 are obtained.

FIG. 10 is a coefficient applied to the vicinity of the horizontal differential filter fx in order to detect the vertical edge, which is near 3 × 11, and the hatched portion is the target pixel.
FIG. 11 shows coefficients applied to the vicinity of the differential filter fy in the vertical direction in order to detect the edge in the horizontal direction, which is in the vicinity of 11 × 3, and the hatched portion is the target pixel.
If the values (absolute values) when using the filters shown in FIGS. 10 and 11 are Fx and Fy, respectively, the edge direction θ at the coordinates (x, y) of the image f is calculated by the following equation (3). The
θ (f, x, y) = arctan (Fy / Fx) (3)

Similarly, the edge strength E is expressed by the following equation (4).
E (f, x, y) = (Fx ² + Fy ² ) ^0.5 (4)

  In addition, the left gradation value Left in the coordinates (x, y) is expressed by Equation (A1) in Equation 2, and the right gradation value Right is expressed by Equation (B1). K is about 5.

By calculating the above calculation with respect to the target pixel g (x, y) of the right image that is supposed to correspond, four characteristics of the following formulas (3a) and (4a) and (A2) and (B2) of Formula 3 are obtained. Find the amount. Note that the values of the right image corresponding to fx, fy, Fx, and Fy are gx, gy, Gx, and Gy.
θ (g, x, y) = arctan (Gy / Gx) (3a)
E (g, x, y) = (Gx ² + Gy ² ) ^0.5 (4a)

Using the above feature quantities, S ₁ (Li, R (Li)) is calculated as shown in Equation (5). S ₁ (Li, R (Li)) takes a real value between 0 and ₁ .

Here, E _max is the maximum value of the edge strength.
By obtaining S ₁ (Li, R (Li)) while changing i, the matching degree f ₁ (equation (2)) of the left and right corresponding edges is calculated.

(1 to 2)
When there are two right edges (Rj, Rk) corresponding to a certain left edge Li, the degree of coincidence between Li and Rj and the degree of similarity between Li and Rk are calculated using the equation (5), respectively. The sum is divided by 2 to obtain an average value, which is regarded as the degree of coincidence of edges corresponding to Li. By obtaining this while changing i, the matching degree f ₁ (equation (2)) of the left and right corresponding edges is calculated.

3.2 “B: Is it valid as the left and right (= both ends) edges of the same object?”
Next, validity f ₂ for determining whether or not the two edges in the left image are the left and right edges of the same object 1 is calculated by Expression (6) of Formula 5.

Here, Li represents the left edge (determined by the x coordinate value) of the i-th edge pair in the left image, and Ri represents the right edge. S ₂ (Li, Ri) is an amount representing validity when the edges Li and Ri are edges at both ends of the same object, and takes a value from 0 to 1 and is closer to 1. We consider that both edges are both valid edges of the object.
Specifically, the calculation is performed in the same manner as in 3.1. However, in this example, the direction of the edge is not reliable, so the other three feature quantities, that is, the edge strength and the left and right tone values ( S ₂ (Li, Ri) is calculated from the X-ray transmission density value.

Now, let the edge Li be f (xL, yL) and Ri be f (xR, yR). At this time, the next quantity is calculated by the same method as described in 3.1.
Edge strength: E (f, xL, yL) and E (f, xR, yR) are calculated by the above-described equation (4).
The gradation value Left on the left side of the target edge is calculated by Expressions (A3) and (A4) of Equation 6, and the gradation value Right on the right side of the target edge is calculated by Expressions (B3) and (B4).

At this time, as shown in FIG. 12, in the case of the edges at both ends of the same object 1, (1) the probability that the gradation value on the right side of the left edge is similar to the gradation value on the left side of the right edge is high. And (2) the right gradation value of the left edge and the right edge of the left edge is considered to have little relation.
Therefore, it is possible to assume the similarity between Right (f, xL, yL) and Left (f, xR, yR) among the above feature quantities. From these, S ₂ (Li, Ri) is finally calculated from Equation (7) of Equation 7. In this equation, Left (f, xL, yL) and Right (f, xR, yR) are not used.

S ₂ (Li, Ri) takes a real value from 0 to 1, and the closer to 1, the higher the possibility that these two edges are the edges of both ends of the same object. By substituting this S ₂ (Li, Ri) into equation (6), the validity f ₂ of the edge pair in the left image is calculated.

When there is a one-to-two relationship in Table 2, that is, when an edge is considered to serve as the left end or the right end of two objects, S ₂ (Li, Ri) is obtained for each, and f ₂ It is reflected in the calculation. That is, S ₂ (Li, Ri) is obtained for all possible edge pairs.
Similarly, to calculate the validity f ₃ of the edge pair in the right image.

3.3 Method for calculating fitness The evaluation values f ₁ , f ₂ , and f ₃ are obtained by the method described in 3.2, and the final fitness fitness is expressed as a linear sum thereof (1) Calculate from

The present invention is not limited to the fitness function described above, and other fitness functions may be used.
However, “Validity of correspondence between left and right edges (f ₁ )”, “Validity of both edges of the same object in the left image (f ₂ )”, “Validity of both edges of the same object in the right image It is necessary to judge comprehensively from the three viewpoints of “sex (f ₃ )”.

  In addition, it is preferable to improve the reliability of edge correspondence of the entire image by considering other information, for example, “correlation with edge correspondence in the upper and lower lines”.

In S5 of FIG. 6, the above-mentioned applicability is calculated, the optimum individual is determined in S6, and it is determined in S7 whether the optimum individual is a practical solution.
Steps S3 to S7 are repeated until a practical solution is obtained, and when a practical solution is obtained, the practical solution is terminated as a solution.

  Next, in S8, the position of the object is triangulated from the correspondence between the edges of the object between the images and the both ends (edge pairs) of the edges of the object in each image, 3D measurement.

  By the evolution calculation method of S2 to S8 described above, it is possible to make stereo matching of left and right X-ray images with large parallax, which makes conventional stereo matching difficult or substantially impossible.

Next, in this example, an intermediate image is created in S9, and stereoscopic images from a plurality of points are displayed in S10. The object can be visually inspected by the stereoscopic image of the object.
That is, each object 1 can be extracted by the three-dimensional measurement (S8) of the object, so that it is displayed in a three-dimensional manner on the screen so that the object looks like a three-dimensional image when the parallax is adjusted. Can do.

In addition, since it is possible to correctly associate the edge portions representing the left and right edges of the left and right regions, in addition to stereoscopic image display, distance measurement of points in the object, the front-rear relationship of each object, the distance from the floor, Etc. can also be measured.
In addition, this allows you to selectively extract only a specific area and recognize what it is from features such as the shape of the area. Can do.
Furthermore, when displaying a stereoscopic image, it is possible to separate and display the objects one by one by coloring each object or emphasizing the distance direction from the actual depth distance.

  Further, a function of automatically recognizing the superposed target object 1 after separating it using distance information and a function of automatically extracting a target object (such as an electronic circuit) having a specific property can be added. .

According to the above-described apparatus and method of the present invention, (1) the association of the edge of the object 1 and (2) the object 1 from the plurality of X-ray images, for example, by the image processing apparatus using the evolution calculation method. Since the parallax is large and the conventional stereo matching is difficult or substantially impossible, the correspondence between the edges of the object 1 and the edges of the edges of the object 1 are determined. Decisions can be made in a short time with simple processing.
Therefore, based on the result of the three-dimensional measurement of the position of the target object 1, the three-dimensional image of the target object 1, the distance measurement of the points in the target object 1, the context of each target object 1, the distance from the floor, etc. Since it can obtain | require, the target object 1 can be test | inspected.

  In addition, the apparatus of the present invention can be configured only by adding an X-ray source and a detector to the longitudinal direction of the apparatus (conveyor traveling direction) as compared to a conventional apparatus having one X-ray source and one detector. Compared to X-ray CT, it can be made smaller and less expensive.

  In addition, this invention is not limited to embodiment mentioned above, Of course, it can change variously in the range which does not deviate from the summary of this invention.

1 object, 10 X-ray irradiation device,
12 conveyor devices, 14 shielding devices,
16a, 16b X-ray source,
18a, 18b detector (X-ray detector),
20 Image processing device

Claims (7)

An X-ray irradiation apparatus that irradiates an object with X-rays from a plurality of directions and captures a plurality of X-ray transmission images with parallax;
An image processing device for inspecting the object from the plurality of X-ray transmission images,
An X-ray inspection apparatus characterized in that, by an image processing apparatus, an edge of an object is associated, both ends of the edge of the object are determined, and then the position of the object is three-dimensionally measured.   The X-ray inspection apparatus according to claim 1, wherein an evolution calculation method is used for edge association and edge both end determination by the image processing apparatus. The X-ray irradiation device includes a conveyor device that horizontally conveys an object;
A shielding device surrounding a part of the conveyor device;
A first X-ray source and a first detector provided on the upstream side of the shielding device for capturing a first X-ray transmission image of the object;
A second X-ray source and a second detector, which are provided on the downstream side of the shielding device and capture a second X-ray transmission image having a large parallax with respect to the first X-ray transmission image of the object. Item 2. The X-ray inspection apparatus according to Item 1. (A) A plurality of X-ray transmission images with parallax are captured by irradiating an object with X-rays from a plurality of directions,
(B) From the plurality of X-ray transmission images, the image processing device determines correspondence between the edges of the object and both ends of the edge of the object,
(C) Next, an X-ray inspection method characterized in that the position of the object is three-dimensionally measured.   5. The X-ray inspection method according to claim 4, wherein an evolution calculation method is used for edge association and edge both end determination by the image processing apparatus. In (B) above,
(A) A parent individual is represented by a chromosome representing one individual in the evolutionary calculation method,
(B) changing a part of the chromosome of the parent individual to generate a child individual;
(C) calculating the fitness of the child individual to determine the optimal individual;
(D) determining whether the optimum individual is a desired practical solution;
5. The X-ray according to claim 4, wherein (e) when the solution is not a practical solution, the optimal individual is set as a parent individual and (a) to (d) are repeated, and when the solution is a practical solution, the optimal individual is set as a solution. Inspection method. In (c), the fitness is calculated by using an fitness function based on “the validity of the correspondence between the edges of the object” and “the reliability of both ends of the edges of the object”. The X-ray inspection method according to claim 6.

Images