WO2009015563A1 - Procédé et système d'identification de matériau par le biais d'images binoculaires de transmission d'énergie multiple - Google Patents

Procédé et système d'identification de matériau par le biais d'images binoculaires de transmission d'énergie multiple Download PDF

Info

Publication number
WO2009015563A1
WO2009015563A1 PCT/CN2008/001418 CN2008001418W WO2009015563A1 WO 2009015563 A1 WO2009015563 A1 WO 2009015563A1 CN 2008001418 W CN2008001418 W CN 2008001418W WO 2009015563 A1 WO2009015563 A1 WO 2009015563A1
Authority
WO
WIPO (PCT)
Prior art keywords
energy
image
grayscale
fault
tomogram
Prior art date
Application number
PCT/CN2008/001418
Other languages
English (en)
French (fr)
Inventor
Yali Xie
Qitian Miao
Hua Peng
Kejun Kang
Haifeng Hu
Zhiqiang Chen
Xueguang Cao
Chuanxiang Tang
Jianping Gu
Xuewu Wang
Hongsheng Wen
Bei He
Yaohong Liu
Shangmin Sun
Quanwei Song
Jin Lin
Xianli Ding
Original Assignee
Nuctech Company Limited
Tsinghua University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Priority claimed from CN2008100813251A external-priority patent/CN101358936B/zh
Application filed by Nuctech Company Limited, Tsinghua University filed Critical Nuctech Company Limited
Publication of WO2009015563A1 publication Critical patent/WO2009015563A1/zh

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/50—Depth or shape recovery
    • G06T7/55—Depth or shape recovery from multiple images
    • G06T7/593—Depth or shape recovery from multiple images from stereo images
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N13/00—Stereoscopic video systems; Multi-view video systems; Details thereof
    • H04N13/20—Image signal generators
    • H04N13/204—Image signal generators using stereoscopic image cameras
    • H04N13/239—Image signal generators using stereoscopic image cameras using two 2D image sensors having a relative position equal to or related to the interocular distance
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N23/00—Investigating or analysing materials by the use of wave or particle radiation, e.g. X-rays or neutrons, not covered by groups G01N3/00 – G01N17/00, G01N21/00 or G01N22/00
    • G01N23/02—Investigating or analysing materials by the use of wave or particle radiation, e.g. X-rays or neutrons, not covered by groups G01N3/00 – G01N17/00, G01N21/00 or G01N22/00 by transmitting the radiation through the material
    • G01N23/04—Investigating or analysing materials by the use of wave or particle radiation, e.g. X-rays or neutrons, not covered by groups G01N3/00 – G01N17/00, G01N21/00 or G01N22/00 by transmitting the radiation through the material and forming images of the material
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N23/00—Investigating or analysing materials by the use of wave or particle radiation, e.g. X-rays or neutrons, not covered by groups G01N3/00 – G01N17/00, G01N21/00 or G01N22/00
    • G01N23/02—Investigating or analysing materials by the use of wave or particle radiation, e.g. X-rays or neutrons, not covered by groups G01N3/00 – G01N17/00, G01N21/00 or G01N22/00 by transmitting the radiation through the material
    • G01N23/06—Investigating or analysing materials by the use of wave or particle radiation, e.g. X-rays or neutrons, not covered by groups G01N3/00 – G01N17/00, G01N21/00 or G01N22/00 by transmitting the radiation through the material and measuring the absorption
    • G01N23/083—Investigating or analysing materials by the use of wave or particle radiation, e.g. X-rays or neutrons, not covered by groups G01N3/00 – G01N17/00, G01N21/00 or G01N22/00 by transmitting the radiation through the material and measuring the absorption the radiation being X-rays
    • G01N23/087—Investigating or analysing materials by the use of wave or particle radiation, e.g. X-rays or neutrons, not covered by groups G01N3/00 – G01N17/00, G01N21/00 or G01N22/00 by transmitting the radiation through the material and measuring the absorption the radiation being X-rays using polyenergetic X-rays
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01V—GEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
    • G01V5/00—Prospecting or detecting by the use of ionising radiation, e.g. of natural or induced radioactivity
    • G01V5/20—Detecting prohibited goods, e.g. weapons, explosives, hazardous substances, contraband or smuggled objects
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G06T7/0004—Industrial image inspection
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N13/00—Stereoscopic video systems; Multi-view video systems; Details thereof
    • H04N13/20—Image signal generators
    • H04N13/204—Image signal generators using stereoscopic image cameras
    • H04N13/254—Image signal generators using stereoscopic image cameras in combination with electromagnetic radiation sources for illuminating objects
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10116—X-ray image
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30108—Industrial image inspection

Definitions

  • the present invention relates to the field of radiation imaging technology, and more particularly to a method for material identification using dual-view multiple-energy transmission images. Background technique
  • Radiation imaging technology enables the transmission of the object's transmission image in a non-contact state by means of the high-energy X-ray penetration capability.
  • the working principle of scanning radiation imaging uses X-rays emitted by the radiation source, and the X-rays are received by the detector after being received by the detector, and then converted into an electrical signal input image acquisition system, by image The acquisition system inputs an image signal into a computer display to display the detected image.
  • the transmission image formed by radiation imaging is a projection of all objects penetrated by the X-ray beam stream, and does not include information in the depth direction.
  • the scanned image is an image formed by superimposing all objects along the direction of the scanning beam, which is not conducive to checking the items hidden behind other objects, and also cannot identify the material of the transmitted image.
  • a more mature main body reconstruction technology uses computed tomography, but the disadvantages of this technology are: not only complicated equipment, high cost, but also unable to quickly check large objects. The pass rate is low, and again, the material of the transmissive object cannot be identified.
  • the dual-view radiation transmission image processing technique is a radiation imaging technique that can detach an object located at a different depth of the slice in the detection space from the image, thereby removing the occlusion. This technique is used to strip certain overlapping objects in the transmitted image so that the obscured object is more visible in the image, but the material properties cannot be identified.
  • the multi-energy transmission image recognition technology is a radiation transmission image processing technique that identifies properties of materials such as organic substances, mixtures, metals, etc., by utilizing the characteristics that different materials have different ray attenuation capabilities for different energies.
  • this technique can only identify the properties of the object that accounts for the main part of the attenuation. If the object to be identified absorbs only a small fraction of the total attenuation dose, it is not possible to use this technique to identify the properties of such an object.
  • the present invention provides a scanning radiation identification-imaging method for a large object radiation inspection system having a simple structure.
  • the method is a material identification method for combining a dual-view technique and a multi-energy transmission image technique for performing a transmission image.
  • the method firstly detects the space by the dual-view technique.
  • the tomographic template of the intermediate object along the depth direction is then reconstructed from the transmission image, and finally the multi-energy technique is used to identify the object with successful gray reconstruction in the fault.
  • a method for material identification using a dual-view multi-energy transmission image includes the steps of: passing two X-rays at a certain angle through an object to be inspected, thereby obtaining left and right transmission image data, and transmitting the left and right directions The image is segmented, and the results of the segmentation are matched;
  • Material identification is performed on objects with successful grayscale reconstruction in the fault.
  • the invention provides a method for material identification by using dual-view multi-energy transmission images, by which an object covering the main component of the ray absorption can be removed for an object overlapping in the direction of the beam, so that the original Objects that are not apparent due to the secondary component of the ray absorption become apparent and can be identified by their material properties: such as organic matter, mixtures, metals, etc.).
  • the method according to the invention makes it possible to identify the non-primary components in the direction of the radiation, which lays the foundation for the automatic identification of explosives, drugs and other harmful objects contained in the container.
  • FIG. 1 shows a schematic diagram of detecting a spatial fault according to the dual view technique of the present invention
  • FIG. 2 shows a specific flow chart of image segmentation in accordance with the present invention
  • Figure 3 shows an example of a template map of a matched fault according to the present invention.
  • Figure 5 shows an example of a left view obtained by the method according to the invention and its greyscale reconstruction
  • Figure 6 shows a flow of grayscale grayscale reconstruction including different energies in accordance with the present invention
  • Figure 7 shows a detailed flow of grayscale reconstruction for a single energy in accordance with the present invention.
  • Figure 8 is a flow chart showing the material identification of an object in any layer according to the present invention.
  • Figure 9 is a multi-energy recognition effect diagram for a plurality of materials without overlap;
  • Figure 10 shows a non-peeling recognition effect diagram
  • Figure 11 is a diagram showing the multi-energy material recognition effect of the double-view occlusion peeling process.
  • Figure 12 is a block diagram of a dual-view multi-energy scanning radiation imaging system in accordance with the present invention.
  • FIGS. 13A and 13B are top views of a schematic layout of a dual-view multi-energy scanning radiation imaging system in accordance with the present invention.
  • Figure 14 is a side elevational view showing the arrangement of a dual angle scanning multi-energy radiation imaging system in accordance with the present invention. detailed description
  • a specific implementation of the method of dual-view multi-energy transmission image material identification according to the present invention is described in three parts hereinafter. I. Using a dual-view processing technique for each energy double-view image to obtain a tomogram template corresponding to the transmission image of the energy, and combining the fault templates of different energies into a template of a set of tomographic images.
  • FIG. 1 is a schematic illustration of the detection of spatial faults in accordance with the dual viewing angle technique of the present invention.
  • the detection space is a three-dimensional space formed by a region scanned by a sector of a radiation source (coordinate origin 0) to a vertical array of detectors and ?: where 0 ⁇ Coordinate origin; S: source position; left detector array; p: right detector array; 0L left beam; OR: right beam; ⁇ : left and right beam angle.
  • the scan direction (vertical upward) is the positive direction of the f-axis
  • the coordinate value is the scan ordinal number
  • the direction of the detector arrangement (perpendicular to the outside of the paper) is the positive axis
  • the coordinate value is the detector ordinal number to the horizontal To the right is the Z-axis positive direction
  • the coordinate value is the fault ordinal number.
  • the space rectangular coordinate system is established with the position 0 of the ray source as the coordinate origin, and we call the space where the angular coordinate system is located as the detection space.
  • the fault is a series of spatial planes parallel to the X- 0-y plane, in the figure
  • the dotted line shows the projection of the plane of the fault on the plane, and the depth of each layer is the distance between the plane of the layer and the plane.
  • A is the projection of the main beam direction of the left beam on the - ⁇ z plane, /? is right
  • 0 is the angle formed by the projection of the left and right beams on the - plane.
  • a template map of the detected space object can be obtained by the image edge extraction technique. That is, several edges are obtained by detecting local discontinuities, and then they are connected.
  • the edge extraction method is reliable in X-ray transmission image segmentation due to the inherent characteristics of the X-ray transmission image when the objects overlap.
  • the present invention simultaneously uses Sobel and Canny edge detection operators to extract edges, and then combines the edges detected by the two detection operators.
  • the edge obtained by the edge is connected to form a closed area, thereby realizing the division of the left and right views.
  • Figure 2 shows a detailed flow chart of image segmentation in accordance with the present invention.
  • edges are extracted.
  • Sobel and Canny edge detection operators are simultaneously used to extract edges.
  • ⁇ represents the pixel gray value of the i-th column and the j-th row in the image
  • ⁇ ⁇ ' ⁇ is a set of all pixel points of the image
  • the present invention uses the Sobel edge detection operator for the number
  • Each pixel of the image [inspects the weighted difference of the gray levels of the upper, lower, left and right neighbors, the neighboring point weight is significant, and the next nearest neighboring point weight is small. Its definition is as follows:
  • , ⁇ ⁇ ⁇ 1 are the convolution operator
  • ⁇ is the convolution of the i-th column and the j-th row
  • the matrix definition of the convolution operator is Then select the threshold 73 ⁇ 4, and the point ( ) that meets the condition s(/, )>73 ⁇ 4 is the step edge point, which is the edge image.
  • the general steps of the Canny edge detection algorithm are: first smoothing the image with a Gaussian filter; then using the finite difference of the first-order partial derivative to calculate the amplitude and direction of the gradient; non-maximum suppression of the gradient amplitude; and finally using the double threshold
  • the algorithm detects and joins edges.
  • the Canny operator uses a double threshold algorithm to reduce false edges. That is, the non-maximum suppression image is binarized with two thresholds A and 7, ⁇ 2 ⁇ , to obtain two threshold edge images N, and N 2 (i,j)o N 2 (, ) is obtained using a high threshold. Contains few false edges, but with discontinuities. Then connect the edges into outlines in N 2 ( , _ ).
  • the algorithm finds the edges that can be joined at the endpoints of the edges in 7 '), and the algorithm continuously finds the edges in the corresponding 8 neighborhoods, so that the algorithm continuously collects the edges until ⁇ ('',') is connected to form the contour. .
  • a closed edge is obtained.
  • the edges detected by the Sobel and Canny edge detection operators are combined and edge-joined to close them.
  • the preliminary edge map of the present invention is the result of finding a logical "or" for the binary edge image obtained by the above two operators. Due to the influence of noise and the like, the edges obtained by the above methods are generally still discontinuous, so they need to be connected.
  • the invention connects the edge pixels according to their continuity in the gradient amplitude or gradient direction. For example, if the pixel ( ⁇ t) is in the neighborhood of the pixel (x, and their gradient magnitude and gradient direction are satisfied at a given threshold:
  • step 03 the segmentation of the view is performed according to the obtained closed edge. Since the closed edge divides the image into two areas, the morphological expansion-corrosion operation can be used to find points that belong to the interior of the area. Then, the region growth method can be used to fill the pixels in the region with the value "1", and the pixels outside the region are filled with the value "0", and the binary template of the inner region of each closed edge is obtained.
  • the size of the map is equal to the projection of the detection space on the plane, ie the number of scans (width) of the number of X detectors (height). This completes the image segmentation and gets the object template.
  • objects on the two template maps are matched according to a certain rule by a dual-view technique. That is, the connected area with each internal value of 1 in the left template picture is compared with each template in the right template picture one by one, and the corresponding template of the same object in the right view is found. Thus, each object that matches successfully corresponds to a template in the left and right views.
  • the difference in position between the two templates in the horizontal direction is the parallax ⁇ ".
  • ⁇ , and / ⁇ are the position of the center of gravity of the matching object in the tomographic template in the left and right views, and the parallax is proportional to the depth of each layer.
  • Figure 3 is a diagram showing an example of a template map of a matched fault in accordance with the present invention. See Figure 3, which shows the left and right view segmentation results.
  • Figure 3 (a) is a left view object template
  • Figure 3 (b) is a right view object template, the template of which can be seen as a rectangle.
  • the fault template of the transmission image obtained according to the dual-view tomography technique reflects the front-back position of the object in the depth direction in the detected space, and the geometry of the fault template reflects the shape contour of the object.
  • FIG. 4 is a flow chart showing a template diagram of each energy layer obtained in accordance with the present invention.
  • a set of transmission images containing the respective radiant energy is first established.
  • a double loop operation with nesting inside and outside is performed, and the inner and outer dashed boxes respectively represent the inner and outer double loops.
  • the inner loop is a template tomogram generation loop.
  • the system establishes a matching object set to match the object number as a loop variable, according to steps 01 to 03 described in (1), to create, match, and object the object in the dual-view transmission image of an energy. Parallax calculation.
  • a template of parallax approximation is merged into the same tomogram in a logical or logical manner to obtain a template tomogram set of matched objects in each fault under a certain radiant energy.
  • the outer loop is a loop for different radiant energies.
  • an inner loop is generated, a tomogram containing a template matching the good objects at each fault depth, and a transmission image generated under the next energy are selected.
  • the image repeats the above steps until the transmitted image of all energy is processed.
  • the template tomogram sets of different energies are layered in a logical or logical manner and are identified by a fault depth. Each layer contains a set of template tomograms of several object templates. 2. According to the fault template, the process of gray reconstruction of multiple energies is performed separately.
  • the fault template map obtained above only reflects the geometry of the object and its spatial position in the container. If material identification is to be performed, gray value reconstruction of multiple energies must also be performed. After gray value reconstruction, we can obtain the gray values of various energies of each segmented object. Thereby material identification of the object can be achieved.
  • grayscale reconstruction for each energy, we reconstruct the objects in each fault based on the two-view grayscale reconstruction method that is stripped from the outside to the inside. That is: first, the gray scale of the matched object in the X- 0-y plane is directly reconstructed (directly adjacent to the background area), and the background is still the original background value, and the object is curved in the outline of the object. The reconstructed grayscale image of the gray value to be obtained is scanned, and the object is stripped from the original image using the reconstructed grayscale. Then do the same for the outer layer of the object, repeat the above procedure until all the matched objects have been reconstructed.
  • Figure 5 (a) Taking the left view shown in Figure 5 (a) as an example, we perform grayscale reconstruction in combination with a template map similar to that shown in Figure 4.
  • three objects overlap each other, from the outside to the inside: a larger rectangle, a smaller rectangle, and a smaller ellipse.
  • Figures 5(b), 5(c) and 5(d) show the effect of grayscale reconstruction.
  • Figure 5 (b) is the outermost side
  • Figure 5 (c> is the middle
  • Figure 5 (d) is the innermost side.
  • Figure 5 (b) is the result of grayscale reconstruction of the outermost object, where the gray value of the light area is the background value in the original picture, and the gray value of the dark area is the gray area of the light area minus the outermost object reconstruction.
  • Gray value, dark area wheel The contour is the same as the contour in the template image of the object, and is a larger rectangle
  • Figure 5 (c) is the grayscale reconstruction result of the intermediate object, wherein the gray value of the light color region is the background value in the original image, and the dark region is The gray value is the gray area minus the intermediate object reconstructed gray value, and the dark area contour is the same as the contour of the object template, which is a smaller rectangle
  • Figure 5 (d) is the innermost object Gray-scale reconstruction results, wherein the gray value of the light-colored area is the background value in the original image, and the gray value of the dark-colored area is the gray-scale area minus the gray-scale area, and the dark area contour and the The template of the object has the same outline and is an elli
  • Figure 6 shows the flow of grayscale grayscale reconstruction including different energies.
  • the first weight is a fault set, which contains all template tomograms with the fault depth as the element distinguishing mark; the second weight is the object set, and the object serial number is used to distinguish the mark, including Each matched object at a specific fault depth; the third weight is an energy set, with different radiant energy as a distinguishing mark, and a transmission gradation reconstructed map at a specific energy of a specific matched object.
  • the above method of reconstructing the grayscale image of the object and completing the reconstruction of the grayscale image of an energy is to perform the grayscale reconstruction of the object by the method of peeling off from the outer to the inner and the grayscale.
  • the objects have all been completed.
  • the flow chart is shown in Figure 7.
  • FIG. 7 shows a detailed flow chart of grayscale reconstruction in accordance with the present invention. Referring to Figure 7, a detailed implementation process of the grayscale reconstruction according to the present invention will be described in detail.
  • the gray scale reconstruction of the object is performed by the method of sequentially peeling off from the outside to the inside and the gray scale.
  • step 01 a set of grayscale reconstruction candidate objects is created using the object obtained by the image segmentation process; in step 02, an attribute of an object is obtained.
  • step 03 it is confirmed whether the acquired object has an edge adjacent to the background.
  • step 04 the acquired object has a grayscale reconstructed object if it has an edge adjacent to the background; if the object has an object occlusion, the grayscale of the occlusion is reconstructed.
  • the object is erased in the image.
  • the gray scale of the reconstructed object be equal to the outer gray level of the edge minus the gray level of the edge side, that is,
  • gray values of various energies of the object represented by each template in each layer can be obtained. These gray values differ depending on the energy. By analyzing these differences, the material in any layer can be identified.
  • FIG. 8 shows the flow of the above identification process.
  • a "fault-object-energy" triple set obtained by the second part of the implementation process is introduced.
  • a double cycle of "fault-object” is then carried out.
  • the outer dotted line frame is a cycle for each fault depth.
  • the present invention creates a "object-energy" double set containing the matched objects of the fault and containing the grayscale reconstruction map for each energy of each object.
  • the object loop in the inner dashed box is performed, that is, the loop in the object with the object number is marked for all the objects in the matching graph in a certain fault.
  • Painted into a color tomogram in such a color tomogram: the identified object
  • the outline of the body is determined by the outline of the template, and the color of the fill in the outline is determined by the combination of the gray scale reconstruction result and the material recognition result. Determination of Color
  • the present invention will be described below in the related description of FIG. Then, create a new "object-energy" set for the next fault, and perform material identification and color map reconstruction until all faults have been processed the same.
  • all the color material recognition effect tomograms obtained above are integrated into a set of color material recognition effect tomograms of objects on different faults, which is the final result obtained by the dual-view and multi-energy image processing technology of the present invention. .
  • the multi-energy material identification method can identify the difference of the gray levels of different energy transmission maps for objects without overlap.
  • Orange is the identification color of organic or light materials
  • green is the identification color of light metals or mixtures
  • blue is the metal identification color.
  • the recognition effect diagram is as shown in Fig. 9.
  • FIG. 9 there is shown a multi-energy recognition effect diagram of graphite, aluminum, iron, lead, and polyethylene rectangular at a mass thickness of 30 g.
  • the left-to-right targets are arranged in the order of graphite, aluminum, iron, lead, and polyethylene.
  • the recognition order is: orange, green, blue, blue, and orange. The material is identified correctly.
  • Fig. 10 shows a non-peeling recognition effect diagram.
  • FIG. 10 there is a large rectangular steel plate obstruction, in the middle is a blocked cylinder containing petroleum liquefied gas, the left is a carton-mounted disc, and the right is a carton containing cigarettes. It can be seen that the identification of the petroleum liquefied gas in the cylinder and the cigarette in the right side of the carton basically recognizes the error, and a partial recognition error occurs in the carton loaded on the left side of the carton. If the double-view layering method is used to peel off the obstruction and then recognize it, the effect is as shown in Fig. 11.
  • Figure 11 for the dual-view occluded multi-energy material recognition effect diagram: where Figure 11 (a) is the occlusion; Figure 11 (b) is the identified object.
  • Figure 11 shows that after using the dual-view occlusion stripping multi-energy material identification technology, the steel sheet covering in Figure 11 (a) is recognized as blue, which is metal, and the carton-mounted disc in Figure 11 (b) is liquefied. Both LPG and carton cigarettes are identified as orange, organic, and the recognition results are correct.
  • Figure 12 is a block diagram of a dual viewing angle multi-energy scanning radiation imaging system in accordance with the present invention. As shown in Fig. 12, the dual-view multi-energy scanning radiation imaging system of the present invention includes the following devices:
  • Multi-energy radiation source 1 X-ray generator, capable of generating X-ray beams of different energies.
  • the beam current controller receives the X-rays emitted by the radiation source 1 and emits two symmetrical or asymmetrical X-rays at an angle.
  • the left detector array 4 receives X-rays of different energies emitted by the multi-energy radiation source and converts them into electrical signals for input to the left image acquisition system 6.
  • the right detector array 5 receives X-rays of different energies emitted by the multi-energy radiation source and converts them into electrical signals for input to the right image acquisition system 7.
  • the left image acquisition system 6 receives an electrical signal from the left detector array 4 and acquires left image data therefrom.
  • the right image acquisition system 7 receives the electrical signal from the right detector array 5 and obtains the right image data therefrom.
  • the computer processing system 8 receives the left and right image data from the left image acquisition system 6 and the right image acquisition system 7 and processes them, respectively displaying the image of the measured object in the computer display, and also displaying the different depths reconstructed by the two. Tomographic image of the fault.
  • the radiation source 1 emits two symmetrical or asymmetrical X-rays at an angle through the beam controller 2, and each X-ray beam passes through the object to be inspected 3 by the left detector array 4 and
  • the right detector array 5 receives; then converts the electrical signals into the left image acquisition system 6 and the right image acquisition system 7, respectively, and the image data in the left and right image acquisition systems 6 and 7 can be processed by the computer processing system 8 respectively.
  • the image of the measured object is displayed on the computer display, and the tomographic image of the depth at different depths is also displayed.
  • the dual-view multi-energy scanning radiation imaging system can respectively obtain a tomographic template map of a transmission image corresponding to the energy of the double-view image of each energy by using a dual-view processing technique, and merge the tomographic templates of different energies.
  • a template of a set of tomographic images according to the tomographic template, respectively, a gray reconstruction process of multiple energies is performed, and the reconstructed tomographic image is subjected to material identification of any layer.
  • DETAILED DESCRIPTION OF THE INVENTION The description of the dual-view multi-energy scanning radiation imaging method of the present invention is not repeated here.
  • a preferred embodiment of the invention utilizes a double-slit collimator as a beam current controller for beam current control of the radiation emitted by the source.
  • FIG. 13 and 14 are a plan view and a side view, respectively, showing a schematic view of the arrangement of the apparatus embodying the present invention.
  • 13A is a case where the beam ray is symmetrical
  • FIG. 13B is a case where the beam ray is asymmetrical.
  • the beam controller is provided with two collimating slits to form a symmetric or asymmetric X-ray of the radiation source with a certain angle of the beam, left and right detector arrays.
  • 4 and 5 respectively scan the object to be measured at a symmetrical angle to the beam fan defined by the collimating slit of the double-slit collimator, and transmit the respective electrical signals to the corresponding left and right images.
  • the system then performs image processing by computer processing system 8 to derive a tomographic image containing depth information.
  • this method can not only enter the main components of the transmitted image in the ray direction.
  • the traditional multi-energy material identification method can only identify the main components in the ray direction. For example, in the direction of the ray, a thicker steel plate overlaps with a bag of smaller drugs. If the traditional multi-energy material identification method is used, in the ray direction, only the steel plate can be recognized, and the drug cannot be identified. .
  • the material identification can be performed separately in each fault, and not only can It is also recognized that steel maple (the main component in the direction of the ray attenuates the ray is large), and it is also possible to identify the drug (the non-primary component in the ray direction is less attenuated by the ray).
  • This method is particularly useful for material identification of container transmission images. For the transmission scanning image of the container, because the container has a large thickness and a long radiation penetration distance, harmful objects such as explosives and drugs are often non-main components in the ray direction. Therefore, the method lays a foundation for automatically identifying harmful substances such as explosives and drugs in a container transmission scanned image.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Theoretical Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Multimedia (AREA)
  • Signal Processing (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Chemical & Material Sciences (AREA)
  • Analytical Chemistry (AREA)
  • Biochemistry (AREA)
  • General Health & Medical Sciences (AREA)
  • Immunology (AREA)
  • Pathology (AREA)
  • Quality & Reliability (AREA)
  • Toxicology (AREA)
  • Electromagnetism (AREA)
  • High Energy & Nuclear Physics (AREA)
  • General Life Sciences & Earth Sciences (AREA)
  • Geophysics (AREA)
  • Analysing Materials By The Use Of Radiation (AREA)

Description

一种利用双视角多能量透射图像进行材料识别的方法及系统 技术领域
本发明涉及辐射成像技术领域， 尤其涉及利用双视角多能量透射图像进行材料 识别的方法。 背景技术
辐射成像技术借助于高能 X射线的穿透能力， 可以在非接触情况下对物体的内部 进行透射， 从而得到物体的透射图像。 对于大型物体的检査， 现有技术中， 扫描辐射 成像的工作原理均采用由辐射源发射 X射线， X射线穿过被检测物体由探测器接收后 转换成电信号输入图像获取系统， 由图像获取系统将图像信号输入计算机显示器显示 所检测图像。 一般情况下， 辐射成像所形成的透射图像是 X射线束流所穿透的全部物 体的投影， 不包含深度方向的信息， 如果多个物体正好沿着射线入射的方向放置， 则 在所形成的扫描图像是沿扫描束流方向全部物体叠加形成的图像， 这样不利于对隐藏 于其它物体背后的物品进行检査， 同时也无法对透射图像的材料进行识别。 为了解决 上述问题， 在辐射成像领域中， 一种比较成熟的主体重建技术是采用计算机断层扫描 技术， 但这种技术的缺点是： 不仅设备复杂、 成本高， 而且无法对大型物体进行快速 检査、 通过率低， 再有， 就是不能识别透射物体的材料。
相对地， 双视角辐射透射图像处理技术是一种可以将位于在探测空间中不同断层 深度的物体从图像中剥离， 从而去除遮挡物的辐射成像技术。 利用这一技术剥离透射 图像中的某些重叠物体，使被遮物体在图像中更为明显，但不能对材料属性进行识别。 多能量透射图像识别技术是一利用特定物质对于不同能量的射线衰减能力不同这一 特性， 对材料的属性， 如对有机物、 混合物、 金属等进行识别的辐射透射图像处理技 术。然而， 当物体有重叠时，这一技术只能将占衰减主要部分的物体的属性识别出来。 如果所要识别的物体只吸收了总衰减剂量中的很少一部分， 则无法只利用该技术识别 这样物体的属性。
发明内容
针对上述现有技术中存在的不足， 本发明提供一种结构简单的大型物体辐射检査 系统的扫描辐射识别 -成像方法。 该方法是一种将双视角技术和多能量透射图像技术 相结合， 用来进行透射图像的材料识别方法。 该方法首先通过双视角技术得到探测空 间中物体沿深度方向的断层模板图， 然后由透射图像重构出断层灰度图， 最后利用多 能量技术对断层中灰度重构成功的物体进行材料识别。
根据本发明的利用双视角多能量透射图像进行材料识别的方法， 包括如下步骤： 使两条成一定夹角的 X射线经过被检物体， 从而获得左右透射图像数据， 对所 述左、 右透射图像进行分割， 对所述分割的结果进行匹配；
建成透射图在深度方向上的断层模板图；
由透射图像重构出断层灰度图，
对任意一种能量的透射图像重复上述过程， 得到任意一种能量在各层的断层模板 图；
将处于同一位置的不同能量的断层模板图合并， 得到预定种能量中所有能量的各 层的断层模板图；
对断层中灰度重构成功的物体进行材料识别。
本发明提供了一种利用双视角多能量透射图像进行材料识别的方法， 通过这种 方法， 可以实现对于在射线束方向上重叠的物体， 剥离其中占射线吸收的主要成分的 遮挡物， 使原本因占射线吸收的次要成分所以不明显的物体变得明显， 并且可以识别 其材料属性： 如有机物、 混合物、 金属等）。
根据本发明的方法可以对射线方向上非主要成份进行材料识别， 这为自动识别 集装箱内藏有的爆炸物品、 毒品等其它有害物体奠定了基础。 附图描述
图 1示出了根据本发明的双视角技术探测空间断层示意图；
图 2示出了根据本发明的图像分割的具体流程图；
图 3示出了根据本发明的匹配好的断层的模板图的实例
图 4示出了根据本发明的得到各能量各层模板图的流程；
图 5示出了根据本发明的方法得到的左视图及其灰度重建的示例；
图 6示出了根据本发明的包括不同能量的灰度图灰度重建的流程；
图 7示出了根据本发明的就单一能量的灰度重建的详细流程；
图 8示出了根据本发明的对任一层内的物体进行材料识别的流程图； 图 9示出了对无重叠的多种材料的多能量识别效果图；
图 10示出了无剥离识别效果图； 图 11示出了经双视角遮挡物剥离处理的多能量材料识别效果图。
图 12为根据本发明的双视角多能量扫描辐射成像系统架构图；
图 13A和图 13B为根据本发明的双视角多能量扫描辐射成像系统布置示意图的俯 视图；
图 14为根据本发明的双视角扫描多能量辐射成像系统布置示意图的侧视图。 具体实施方式
下面结合说明书附图对本发明进行详细描述。
根据本发明的双视角多能量透射图像材料识别的方法的具体实施过程在下文中 分三部分进行描述。 一. 分别对每一种能量的双视角图像运用双视角处理技术得到对应这一种能量 的透射图像的断层模板图， 并将不同能量的断层模板合并为一组断层图像的模板。
图 1为根据本发明的双视角技术探测空间断层示意图。 如图 1所示， 如图 1所 示， 探测空间为由射线源 （坐标原点 0) 至探测器 和 ?处的垂直阵列） 所成的扇 面扫描经过的区域形成的三维空间： 其中 0·· 坐标原点； S: 放射源位置； 左探测 器阵列； p: 右探测器阵列； 0L 左射线束； OR: 右射线束； Θ : 左、 右射线束夹角。
在图 1 中， 以扫描方向 （垂直向上） 为 f轴正向， 坐标值为扫描序数， 以探测 器排列方向 （垂直于纸面向外） 为 轴正向， 坐标值为探测器序数， 以水平向右为 Z 轴正向， 坐标值为断层序数。 以射线源 5"所在位置 0为坐标原点， 建立空间直角坐标 系， 我们称这一直角坐标系所在的空间为探测空间。 断层为一系列与 X- 0-y平面平行 的空间平面， 图中虚线所示为断层所在平面在 平面上的投影， 各层深度即为该 层所在平面与 平面的距离。图中 A为左射线束主束方向在 -^z平面上的投影， /？为右射线束主束方向在 ^平面上的投影。 0为左、 右射线束在 - 平面上投 影所成夹角。
由图像边缘提取技术可得到探测空间物体的模板图。 即通过检测局部不连续性得 到若干边缘， 然后将它们连接。 由于 X射线透射图像在物体重叠时的固有特性， 边缘 提取方法在 X射线透射图像分割中是可靠的。 本发明同时采用 Sobel和 Canny边缘检 测算子来提取边缘， 然后将两个检测算子检测到的边缘进行综合。 对最后得到的边缘 进行边缘连接处理， 形成封闭的区域， 从而实现左右视图的分割。 图 2示出了根据本发明的图像分割的具体流程图。
参考图 2, 根据本发明的图像分割实现过程详细描述如下- 首先， 在步骤 01， 提取边缘。 本发明中同时采用 Sobel和 Canny边缘检测算子来 提取边缘。 对于数字图像 i^' S,其中 ^ 代表图像中第 i列、 第 j行的像素灰 度值， ί ^'， ί即为图像所有像素点的集合，本发明采用 Sobel边缘检测算子对数字图 像 [ 的每个像素考察上、 下、 左、 右邻点灰度的加权差， 近邻点权重大， 次近邻 点权重小。 其定义如下：
=|(/(/- 1， — 1)+2/('·— 1,/)
+ (/-i,;+i))-(/ -+i,y-i)
+ 2f(i + l,j)+f(i + l,j + l)]
+|(/( -i, -i)+2/(, -i)
+ / +i,y-i))-(/( -i,y + i)
+ 2f(i,j + \)+f(i + l,j + \))
上式中 l /|、 ΙΔ^1分别为卷积算子 、 △ 在第 i列、 第 j行的卷积和） 卷积算子的矩阵定义式为
Figure imgf000006_0001
然后选取阈值 7¾， 符合条件 s(/, )>7¾的点 ( )即为阶跃状边缘点， 为 边缘图像。
Canny边缘检测算法的一般步骤为： 先利用高斯滤波器平滑图像； 然后用一阶偏 导的有限差分来计算梯度的幅值和方向； 对梯度幅值进行非极大值抑制； 最后利用双 阈值算法检测和连接边缘。 Canny算子使用双阈值算法减少假边缘。 即对非极大值抑 制图像用两个阈值 A和 7 二值化， Ά2Ώι^Τ ， 得到两个阈值边缘图像 N, 和 N2(i,j)o N2(, )使用高阈值得到， 含有很少的假边缘， 但有间断。然后要在 N2 ( ,_ ) 中把边缘连接成轮廓。 算法在 7')中边缘的端点， 在 ， 图的相应 8邻域位置 寻找可以连接的边缘， 这样， 算法不断地在 中收集边缘， 直到将 ^(''，')连接 起来， 形成轮廓为止。
在步骤 02， 获得封闭边缘。 综合 Sobel和 Canny边缘检测算子检测到的边缘， 并 进行边缘连接， 使其封闭。 本发明初步的边缘图为对上述两算子所得二值边缘图像求逻辑 "或"的结果。 由 于噪声等影响的存在， 由上述方法获得的边缘一般仍是不连续的， 所以还需要对它们 进行连接。 本发明根据边缘像素在梯度幅度或梯度方向上的连续性将其连接。 例如， 如像素 (^t)在像素 (x， 的邻域且它们的梯度幅度与梯度方向在给定的阈值下满足：
ΆΐοΐΆη (θχ/σ}, ) , Γ是幅度阈值， 是
Figure imgf000007_0001
角度阈值。 这样， 对所有的边缘像素都进行上述的判断和连接就可以得到一个闭合的 边界。
在步骤 03， 依据得到的封闭边缘进行视图的分割。 因为闭合的边缘将图像分为内 外两个区域， 所以可以采用形态学膨胀-腐蚀操作找到属于区域内部的点。 然后可以 以此点为起点采用区域生长法将区域内的像素用数值 " 1 "进行填充， 而区域外的像 素用数值 "0 "填充,得到每个封闭边缘内部区域的二值模扳， 模板图的大小等于探测 空间在 平面上的投影， 即扫描次数 （宽） X探测器个数 （高）。 至此完成图像 分割， 得到对象模板。
根据本发明， 通过双视角技术按一定的规则将两模板图上物体进行匹配。 即， 将 左模板图中的每一内部值为 1的连通区域与右模板图中的各模板逐一比较， 找到同一 物体在右视图中的对应模板。 这样， 每一个匹配成功的物体就对应左、 右视图中的一 个模板， 两模板在水平方向上位置差就是视差^"。 根据双视角理论， 各断层深度 z与视差^"的关系为： tan (6>/2) = pr / z。
所有匹配好的对象将在相应深度的断层模板中绘出， 其深度为- z = pr/tan (θ/2) = [μχ ί - μχ ί )/tan- 1 (θ/2)
其中 Α，,和/^ 是断层模板中匹配成功的对象在左右视图中的重心横坐标位置， 视差与各层深度成正比。
图 3所示是根据本发明的匹配好的断层的模板图实例图。参见图 3， 其中所示是 左右视图分割结果。 图 3 (a) 为左视图对象模板； 图 3 (b)为右视图对象模板， 由图可 见这一对象的模板均为一矩形。
根据双视角断层技术得到的透射图像的断层模板， 其所属的层数反映了物体在 被检测空间中沿深度方向所处的前后位置， 而断层模板的几何形状则反映了物体的形 状轮廓。
对任意一种能量的透射图像重复这一处理过程， 就可以得到任意一种能量在各 层的断层模板图。 将处于同一位置的不同能量的断层模板图以逻辑或方式合并， 就可 以得到所有能量的各层的断层模板图。
图 4所示是根据本发明的获到各能量各层模板图的流程图。 参见图 4， 首先建 立包含由各个辐射能量产生的透射图像的集合。 之后进行一个内外嵌套的二重循环操 作， 内外两虚线框分别代表内、 外两重循环。 内循环是模板断层图生成循环。 在这一 循环中系统建立匹配对象集， 以匹配对象序号为循环变量按照 （1 ) 中所述步骤 01至 03 ) 对某一能量的双视角透射图像中的物体进行模板的创建、 匹配与物体视差计算。 将视差近似的模板以逻辑或的方式合并于同一断层图中， 从而取得某一辐射能量下各 断层中的匹配好的物体的模板断层图集合。 外循环是针对不同辐射能量的循环， 在这 一循环顺序实施内循环、 生成本能量下包含各个断层深度上匹配好物体的模板的断层 图、 选择下一能量下生成的透射图像， 对这一图像重复以上步骤直至所有能量的透射 图像被处理完。 外循环结束后， 将不同能量的模板断层图集合以逻辑或的方式同层合 并为一以断层深度为区分标志的， 每层包含若干物体模板的模板断层图集合。 二. 根据断层模板， 分别进行多种能量的灰度重建的过程。
上面得到的断层模板图仅反映物体的几何形状以及其在集装箱中的空间位置。 如果要进行材料识别， 还必须进行多种能量的灰度值重建。 经过灰度值重建， 我们可 以得到每个分割出来的物体的多种能量的灰度值。 从而可以实现物体的材料识别。
在灰度重建时， 针对每一种能量， 我们依据从外向内逐次剥离的双视角灰度重 建方法对各个断层中的物体进行重建。 即： 先对 X- 0- y平面中最外侧 （与背景区域直 接相邻） 的匹配好的对象灰度进行重建， 构成一幅背景仍为原背景值， 物体轮廓内为 弧立地对此物体扫描应得的灰度值的重建灰度图， 并从原图上利用重建灰度剥去这一 物体。 再对次外层的对象进行同样处理， 重复以上手续直至所有匹配好的对象都已完 成重建。
我们以图 5 (a)所示的左视图为例，结合类似图 4所示的模板图，进行灰度重建。 参见图 5 (a)， 其中有三个物体互相重叠， 由外向内依次是： 较大的矩形， 较小 的矩形， 小椭圆。 图 5 (b) , 图 5 (c)以及图 5 (d)示出了灰度重建的效果图。 图 5 (b) 为最外侧， 图 5 (c〉为中间以及图 5 (d)为最里侧。
图 5 (b)为最外侧的物体的灰度重建结果， 其中浅色区域灰度值为原始图中的背 景值， 深色区域的灰度值为浅色区域灰度减最外侧物体重构灰度所得值， 深色区域轮 廓与该物体的模板图中轮廓相同， 为一较大矩形； 图 5 (c)为中间物体的灰度重建结 果， 其中浅色区域灰度值为原始图中的背景值， 深色区域的灰度值为浅色区域灰度减 中间物体重建灰度所得值， 深色区域轮廓与该物体的模板图中轮廓相同， 为一较小矩 形； 图 5 (d)为最里侧的物体的灰度重建结果， 其中浅色区域灰度值为原始图中的背 景值， 深色区域的灰度值为浅色区域灰度减最里侧物体重建灰度所得值， 深色区域轮 廓与该物体的模板图中轮廓相同， 为一椭圆形。
图 6示出了包括不同能量的灰度图灰度重建的流程。
参见图 6， 其中虚线框代表对各种不同能量的循环。 首先， 创建包含各种不同辐 射能量且以能量为区分标志的透射图像以及由实施过程之第一部分得到的以断层深 度为区分标志的， 每层包含若干物体模板的模板断层图集的集合。 在一次循环中对一 种能量的双视角透射图， 依据第一部分中图像分割的结果， 即第一部分所得的模板断 层图集， 经灰度重建、 原图物体消除、 生成物体重构灰度图， 完成一个能量的灰度图 重建。 关于灰度重建的详细过程将在下文参照图 7进行描述。 然后不断选择新能量所 对应的透射图像， 仍依据第一部分中图像分割的结果， 重复以上步骤。 最后将所有能 量的物体重建灰度图综合为一个集合， 以备材料识别使用。这个集合可以（但不限于） 是一个三重集合： 第一重为断层集， 包含以断层深度为元素区分标志的所有模板断层 图； 第二重为对象集， 以对象序号为元素区分标志， 包含特定断层深度上的各个匹配 好的对象； 第三重为能量集， 以不同辐射能量为区分标志， 包含特定的匹配好的对象 的在各个能量下的透射灰度重建图。
以上生成物体重构灰度图， 完成一个能量的灰度图重建的方法是采用由外向内、 灰度逐次剥离的方法进行物体的灰度重构。 先对最外层 （与背景区域直接相邻） 的匹 配好的对象灰度进行重构， 并从原图上剥去， 再对次外层的对象进行同样处理， 重复 以上过程直至所有匹配好的对象都已完成。 其流程图如图 7所示。
图 7示出了根据本发明的灰度重建的详细流程图。参考图 7， 对根据本发明的灰 度重建具体实现过程进行详细描述。
本发明在重构对象灰度时采用由外向内、 灰度逐次剥离的方法进行物体的灰度 重构。 先对最外层 （与背景区域直接相邻） 的匹配好的对象灰度进行重构， 并从原图 上剥去， 再对次外层的对象进行同样处理， 重复以上过程直至所有匹配好的对象都已 完成。
该流程具体内容如下： 在步骤 01 , 利用由图像分割过程得到的对象建立灰度重构备选对象集； 在步骤 02， 获取一对象的属性。
在步骤 03， 确认获取的对象是否有与背景相邻边缘。
在步骤 04， 获取的对象如有与背景相邻边缘， 重构对象灰度； 如果对象有物体遮 挡， 那么要重构遮挡物灰度。
在步骤 05， 在图像中消去此对象。
重复 02至 05步骤针对对象集中的对象进行灰度重建直至所有对象的灰度重构完 成。
经过处理的每一个对象，重构灰度时只会有两种区域：一种是与背景相邻的部分， 一种是被其他物体遮挡住的部分。 注意， 对于原本无法与背景相邻的对象， 在经过足 够多次的遮挡物剥离后， 遮挡物原本存在的区域必有变为背景值的部分， 那么这部分 区域也可视为新的背景区域， 所以最终所有原本被完全遮挡的对象也将可以与背景值 相邻。 设重构出的物体灰度等于边缘外侧灰度减边缘侧灰度， 即
Sobi = (^out - Sn )。 三. 对重建好的断层图像进行任一层的材料识别
经过上述的灰度值重建后， 可以得到各层内每个模板所代表的物体的多种能量的 灰度值。 这些灰度值由于能量的不同而有差异， 分析这些差异就可以对任一层内的物 体进行材料识别。
图示出了上述识别过程的流程。 参见图 8， 首先， 引入由实施过程之第二部分获 得的 "断层-对象-能量"三重集合。 随后进行一个 "断层-对象"的二重循环。）其中， 外虚线框内为对各个断层深度的循环。 在外循环中， 对于每一个断层深度， 本发明创 建一个包含本断层各个匹配好的物体的， 并且包含每个物体各个能量下灰度重建图的 "对象-能量"二重集合。 然后就进行内虚线框内的对象循环， 即对于某一断层所有 匹配好的图中物体进行以物体序号为标志的循环。 我们对本断层中每一个左右视图匹 配好且各能量灰度重构好的物体， 顺序进行灰度差异分析、 材料识别、 彩色图生成的 处理 （此处理过程在本申请人提出的公开号为 CN 1995993 的现有技术中详细描述， 在此不再赘述）。 待此断层所有材料属性识别完毕， 本发明将在以此断层深度为标志 的断层图中依据由第一部分得到的模板断层图、 由第二部分得到的灰度重建图以及上 面获得的物体识别结果绘成一彩色断层图， 在这样一幅彩色断层图中： 被识别出的物 体的轮廓由模板轮廓决定， 轮廓内填充的颜色由灰度重建结果和材料识别结果共同决 定。 颜色的决定方式本发明将在下面对图 9的相关说明中表述。 然后， 对于下一个断 层再次创建新的 "对象-能量"集合， 进行材料识别和彩色图重建， 直至所有断层都 己经过同样处理。 再按断层顺序将上面得到的所有彩色材料识别效果断层图综合为一 不同断层上各物体的彩色材料识别效果断层图的集合， 这就是本发明由双视角、 多能 量图像处理技术获得的最终结果。
多能量材料识别方法可以对于无重叠的物体， 根据不同能量透射图灰度的差异 进行识别。 我们定义但不限于此定义： 橙色为有机物或轻材料的识别色， 绿色为轻金 属或混合物的识别色， 蓝色为金属识别色。 我们还设定， 各种颜色的深浅决定于重构 灰度值的大小。 这样， 识别效果图如图 9所示。
参见图 9， 其中示了质量厚度为 30g时的石墨、 铝、 铁、 铅、 聚乙烯矩型的多能 量识别效果图。
在图 9 中， 从左至右的目标排列顺序为， 石墨、 铝、 铁、 铅、 聚乙烯， 识别色 排列顺序为： 橙、 绿、 蓝、 蓝、 橙， 材料识别无误。
而当物体有相互遮挡时， 如果不作遮挡物剥离处理， 这种别就有可能不正确， 图 10示出了无剥离识别效果图。
如图 10中为较大的矩型钢板遮挡物， 中间为被挡住的装有石油液化气的钢瓶， 左为纸箱装光盘， 右为纸箱装香烟。 可见识别钢瓶内的石油液化气和右边的纸箱装香 烟基本上都识别错误， 对左边的纸箱装光盘也发生了部分识别错误。 如果运用双视角 分层方法剥离遮挡物后再识别， 效果即如图 11所示。
参见图 11 , 其为双视角遮挡物剥离多能量材料识别效果图： 其中图 11 (a)为遮 挡物； 图 11 (b)被识别物。
图 11显示， 运用双视角遮挡物剥离多能量材料识别技术后， 图 11 (a)中的钢板 遮挡物被识别为蓝色， 是金属， 而图 11 (b)图中的纸箱装光盘、 液化石油气、 纸箱装 香烟均被识别为橙色， 是有机物， 识别结果正确。
图 12为根据本发明的双视角多能量扫描辐射成像系统架构图。 如图 12所示， 本 发明的双视角多能量扫描辐射成像系统包括以下装置：
多能量辐射源 1， X射线的发生器， 能够产生不同能量的 X射线束。
束流控制器， 接收辐射源 1发出的 X射线， 并发出两条对称的或非对称的、 成一 定夹角的 X射线。 左探测器阵列 4, 接收由多能量辐射源发出的不同能量的 X射线， 并将其转换成 电信号输入到左图像获取系统 6。
右探测器阵列 5， 接收由多能量辐射源发出的不同能量的 X射线， 并将其转换成 电信号输入到右图像获取系统 7。
左图像获取系统 6， 接收左探测器阵列 4发出的电信号， 并从中获取左图像数据。 右图像获取系统 7， 接收右探测器阵列 5发出的电信号， 并从中获取右图像数据。 计算机处理系统 8， 接收来自左图像获取系统 6和右图像获取系统 7的左、 右图 像数据并进行处理， 分别在计算机显示器中显示所测物体图像， 也可以显示由两者重 建出不同深度的断层透视图像。
本发明中， 辐射源 1经束流控制器 2发出两条对称的或非对称的、 成一定夹角的 X射线，各条 X射线束穿过被检物体 3分别由左探测器阵列 4和右探测器阵列 5接收； 然后转换成电信号分别输入至左图像获取系统 6和右图像获取系统 7， 左、 右图像获 取系统 6和 7中的图像数据经计算机处理系统 8处理后可以分别在计算机显示器中显 示所测物体图像， 也可以显示由两者重建出不同深度的断层透视图像。
根据本发明的双视角多能量扫描辐射成像系统可以分别对每一种能量的双视角 图像运用双视角处理技术得到对应这一种能量的透射图像的断层模板图， 并将不同能 量的断层模板合并为一组断层图像的模板； 再根据断层模板， 分别进行多种能量的灰 度重建的过程， 对重建好的断层图像进行任一层的材料识别。 其具体工作方法和过程 如上文中对本发明的双视角多能量扫描辐射成像方法的描述， 在此不再重复。
本发明一个优选的实施方式是采用双缝准直器作为束流控制器来对射线源发出 的射线进行束流控制。
图 13和图 14分别为实施本发明的装置布置示意图的俯视图和侧视图。 其中图 13A为束流射线对称的情形， 图 13B为束流射线非对称的情形。 如图 13A和图 13B所 示， 束流控制器上置有两条准直缝使辐射源发出的 X射线形成对称或非对称的、 有一 定夹角的射线束流， 左、 右探测器阵列 4和 5分别正对由双缝准直器的准直缝限定的 束流扇面， 以对称的角度分别对被测物体进行扫描检查， 并将各自的电信号传输给对 应的左、 右图像获取系统， 然后由计算机处理系统 8进行图像处理得出含有深度信息 的断层透视图像。 从上面示例可以看出， 该方法不但可以对透射图像在射线方向上的主要成分进 行材料识别， 同时也可以通过剥离主要成分， 对非主要成分进行材料识别。 而传统的 多能量材料识别方法， 在射线方向只能对主要成分进行材料识别。 例如， 在射线方向 上，一块较厚的钢板与一袋较小的毒品相互重叠，若采用传统的多能量材料识别方法， 则在射线方向上， 一般只能识别出钢板， 而不能识别出毒品。若采用本发明所述方法， 由于先利用双视角技术将钢板和毒品分成不同的两个断层， 各层分别进行多能量的灰 度重建后， 则可以在各断层分别进行材料识别， 从而不但可以识别出钢枫 （在射线方 向上的主要成分对射线衰减较大）， 也可以识别出毒品 （在射线方向上的非主要成分 对射线衰减较小）。 该方法对集装箱透射图像的材料识别特别有用。 对于集装箱的透 射扫描图像来说， 由于集装箱厚度较大， 射线穿透距离长， 所以在射线方向上， 爆炸 物、 毒品等有害物体往往都是非主要成份。 因此， 该方法为在集装箱透射扫描图像内 自动识别爆炸物、 毒品等有害物质奠定了基础。

Claims

权利要求
K 一种利用双视角多能量透射图像进行材料识别及成像方法， 包括如下步骤：
1) 使两条成一定夹角的 X射线经过被检物体， 从而获得左右透射图像数据， 对所述左、 右透射图像进行分割， 对所述分割的结果进行匹配；
2) 建成透射图在深度方向上的断层图；
3) 对任意一种能量的透射图像重复上述过程，得到任意一种能量在各层的断 层图；
4) 将处于同一位置的不同能量的断层图合并，得到预定种能量中所有能量的 各层的断层图；
5) 对断层中灰度重构成功的物体进行材料识别。
2、 根据权利要求 1所述的方法， 其特征在于,所述步骤 1中是采用边缘提取算法 对左、 右透射图像进行分割。
3、 根据权利要求 1或 2所述的方法， 其特征在于,所述对左、 右透射图像进行分 割进一步包括以下步骤-
1) 提取图像的边缘；
2) 获得图像的封闭边缘；
3) 依据得到的封闭边缘进行视图的分割。
4、 根据权利要求 1所述的方法， 其特征在于,所述步骤 1中对分割的结果进行匹 配是根据左、 右分割结果的几何特征进行匹配。
5、 根据权利要求 1所述的方法， 其特征在于,所述步骤 1中对分割的结果迸行匹 配包括以下歩骤：
1) 建立分割结果的对象；
2) 为以上对象属性集分配相应的权重。
6、 根据权利要求 1所述的方法， 其特征在于，所述步骤 2中建成透射图在深度方 向上的断层图是根据左、 右视图匹配的结果以及对应的绝对视差值实现的。
7根据权利要求 1所述的方法， 其特征在于,所述步骤 3) 中重构断层图灰度值采 用由外向内、 灰度逐次剥离的方法进行物体的灰度重建， 其中， 先对与背景区域直接 相邻的最外层的匹配好的对象灰度进行重构， 并从原图上剥去， 再对次外层的对象进 行同样处理， 重复以上过程直至所有匹配好的对象都已完成。
8、 根据权利要求 1所述的方法， 其特征在于,所述步骤 3中重构断层图灰度值包 括以下步骤：
1 ) 利用所述对象建立灰度重构备选对象集；
2) 获取一对象的属性；
3 ) 确认获取的对象是否有与背景相邻边缘；
4 ) 获取的对象如有与背景相邻边缘， 重构对象灰度； 如果对象有物体遮挡， 那么要重构遮挡物灰度；
5) 在图像中消去此对象；
重复所述步骤 2) 至步骤 5) 对对象集中的对象进行灰度重建直至所有对象的灰 度重构完成。
9、 根据权利要求 3所述的方法， 其特征在于,所述提取图像的边缘的步骤是同时 采用 Sobel和 Canny边缘检测算子来提取边缘。
10、 根据权利要求 9所述的方法， 其特征在于， 所述 Sobel边缘检测算子对数字 图像 的每个像素考察上、 下、 左、 右邻点灰度的加权差， 近邻点权重大， 次近 邻点权重小。
11、 根据权利要求 9所述的方法， 其特征在于， 使用所述 Canny边缘检测算子提 取边缘的步骤如下：
1 ) 利用高斯滤波器平滑图像；
2) 然后用一阶偏导的有限差分来计算梯度的幅值和方向；
3) 对梯度幅值进行非极大值抑制；
4 ) 利用双阈值算法检测和连接边缘。
12、 根据权利要求 4 所述的双视角扫描辐射成像方法， 其特征在于,所述步骤 2 中获得图像的封闭边缘是根据边缘像素在梯度幅度或梯度方向上的相似性将其连接 的。
13、 根据权利要求 8所述的方法， 其特征在于,所述步骤 3中获取的对象如有与 背景相邻边缘， 重构对象灰度， 如有遮挡物， 同时重构遮挡物灰度； 在对象集和图像 中消去此对象。
14、 根据权利要求 1所述的方法， 其中， 还包括：
对一种能量的双视角透射图， 进行灰度重建、 原图物体消除、 生成物体重构灰度 图， 完成一个能量的灰度图重建； 然后，选择下一能量重复以上步骤，最后将所有能量的物体重建灰度图综合为一 个集合。
15、 根据权利要求 1所述的方法， 其中，
所述步骤 6 ) 中的识别还包括：
对每一个左右视图匹配好且各能量灰度重构好的物体， 顺序进行灰度差异分析、 材料识别、 彩色图生成的处理；
所有物体处理完毕后， 再按断层顺序综合为一组不同断层上各物体的识别效果 图。
16、 一种双视角多能量扫描辐射成像系统， 包括辐射源（1 )， 左探测器阵列（4 )， 右探测器阵列 （5 ) , 左图像获取系统 （6 )， 右图像获取系统 （7 ) 和计算机处理系统
(8 )， 其特征在于，
所述辐射源 （1 ) 为多能量 X射线发生器， 能够产生不同能量的 X射线束； 所述左探测器阵列 （4)， 接收不同能量的 X射线， 并将其转换成电信号输入到左 图像获取系统 （6);
所述右探测器阵列 （5)， 接收不同能量的 X射线， 并将其转换成电信号输入到右 图像获取系统 （7 );
所述左图像获取系统 （6 )， 接收左探测器阵列 （4) 发出的电信号， 并从中获取 左图像数据；
所述右图像获取系统 （7 )， 接收右探测器阵列 （5 ) 发出的电信号， 并从中获取 右图像数据。
所述计算机处理系统（8), 接收来自左图像获取系统（6)和右图像获取系统（7 ) 的左、 右图像数据， 并对所述不同能量下左、 右图像数据进行处理， 实现对某一特定 断层上物体的材料识别。
17、 根据权利要求 16 所述的扫描辐射成像系统， 其特征在于， 其中所述计算机 处理系统对所述不同能量下左、 右图象数据进行分割， 对所述分割的结果进行匹配， 从而建成不同能量透视图在深度方向上的断层图， 并重构断层图的灰度值， 得到任意 一种能量在各层的断层图， 将处于同一位置的不同能量的断层图合并， 得到预定种能 量中所有能量的各层的断层图， 并对断层中匹配的物体进行材料识别。
18、 根据权利要求 16所述的扫描辐射成像系统， 其特征在于， 还包括： 束流控制器 （2)， 接收辐射源 （1 ) 发出的 X 射线， 并发出两条对称的或非对称 的、 成一定夹角的 X射线。
19、 根据权利要求 18 所述的扫描辐射成像系统， 其特征在于， 其中所述束流控 制器 (2)为双缝准直器。
20、 根据权利要求 19所述的扫描辐射成像系统， 其特征在于， 其中所述双缝准 直器上置有两条准直缝使辐射源发出的射线形成对称或者非对称的、 成一定夹角的束 流扇面。
21、 根据权利要求 16所述的扫描辐射成像系统， 其特征在于所述探测器阵列为 "I"型探测器》
22、 根据权利要求 16所述的扫描辐射成像系统， 其特征在于：
所述计算机处理系统 （8)， 还可以基于左、 右图像数据分别在计算机显示器中显 示所测物体图像。
PCT/CN2008/001418 2007-08-02 2008-08-04 Procédé et système d'identification de matériau par le biais d'images binoculaires de transmission d'énergie multiple WO2009015563A1 (fr)

Applications Claiming Priority (4)

Application Number Priority Date Filing Date Title
CN200710119871.5 2007-08-02
CN200710119871 2007-08-02
CN200810081325.1 2008-02-25
CN2008100813251A CN101358936B (zh) 2007-08-02 2008-02-25 一种利用双视角多能量透射图像进行材料识别的方法及系统

Publications (1)

Publication Number Publication Date
WO2009015563A1 true WO2009015563A1 (fr) 2009-02-05

Family

ID=39767414

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2008/001418 WO2009015563A1 (fr) 2007-08-02 2008-08-04 Procédé et système d'identification de matériau par le biais d'images binoculaires de transmission d'énergie multiple

Country Status (4)

Country Link
DE (1) DE102008040952A1 (zh)
FR (1) FR2919780B1 (zh)
GB (1) GB2454047B (zh)
WO (1) WO2009015563A1 (zh)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104567758B (zh) 2013-10-29 2017-11-17 同方威视技术股份有限公司 立体成像系统及其方法

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP0610084A2 (en) * 1993-02-05 1994-08-10 Max Robinson The visual presentation of information derived from a 3D image system
US20030081720A1 (en) * 2001-10-31 2003-05-01 Swift David C. 3D stereoscopic X-ray system
GB2390005A (en) * 2002-06-17 2003-12-24 Royal Holloway University Of L Screening Apparatus
JP2007017304A (ja) * 2005-07-08 2007-01-25 Matsushita Electric Ind Co Ltd Ｘ線検査装置およびｘ線検査方法
CN101210895A (zh) * 2006-12-28 2008-07-02 清华同方威视技术股份有限公司 一种双视角扫描辐射成像的方法及系统

Family Cites Families (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6018562A (en) * 1995-11-13 2000-01-25 The United States Of America As Represented By The Secretary Of The Army Apparatus and method for automatic recognition of concealed objects using multiple energy computed tomography
GB9720658D0 (en) * 1997-09-29 1997-11-26 Univ Nottingham Trent Detecting, improving and charecterising material in 3-D space
US7031425B2 (en) * 2002-11-27 2006-04-18 Ge Medical Systems Global Technology Company, Llc Methods and apparatus for generating CT scout images
CA2574402A1 (en) * 2004-07-20 2006-01-26 William Awad System and method for detecting the presence of a threat in a package
CN1995993B (zh) 2005-12-31 2010-07-14 清华大学 一种利用多种能量辐射扫描物质的方法及其装置

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP0610084A2 (en) * 1993-02-05 1994-08-10 Max Robinson The visual presentation of information derived from a 3D image system
US20030081720A1 (en) * 2001-10-31 2003-05-01 Swift David C. 3D stereoscopic X-ray system
GB2390005A (en) * 2002-06-17 2003-12-24 Royal Holloway University Of L Screening Apparatus
JP2007017304A (ja) * 2005-07-08 2007-01-25 Matsushita Electric Ind Co Ltd Ｘ線検査装置およびｘ線検査方法
CN101210895A (zh) * 2006-12-28 2008-07-02 清华同方威视技术股份有限公司 一种双视角扫描辐射成像的方法及系统

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
YUE W. ET AL.: "Automated 2-D Cephalometric Analysis on X-ray Images by a Model Based Approach", IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, vol. 53, no. 8, August 2006 (2006-08-01), pages 1615 - 1623 *

Also Published As

Publication number Publication date
FR2919780A1 (fr) 2009-02-06
FR2919780B1 (fr) 2017-09-08
GB2454047A (en) 2009-04-29
GB0814150D0 (en) 2008-09-10
DE102008040952A1 (de) 2009-02-19
GB2454047B (en) 2011-02-23

Similar Documents

Publication Publication Date Title
US8194953B2 (en) Method and system of material identification using binocular steroscopic and multi-energy transmission images
US7545906B2 (en) Method and system for binocular steroscopic scanning radiographic imaging
AU2002307053B2 (en) X-ray inspection system
EP2856430B1 (en) Determination of z-effective value for set of voxels using ct density image and sparse multi-energy data
US8917927B2 (en) Portable backscatter advanced imaging technology scanner with automated target recognition
US20090245463A1 (en) Automatic material discrimination by using computer tomography
US8180139B2 (en) Method and system for inspection of containers
US8670522B2 (en) Stereo X-ray inspection apparatus and method for forming three-dimensional image through volume reconstruction of image acquired from the same
JP6487703B2 (ja) Ｘ線検査装置及びｘ線検査方法
US9414788B2 (en) X-ray imaging apparatus and method of generating stereoscopic image
JP2005530153A (ja) 三次元座標を取得するための立体ｘ線撮像装置
US20160335756A1 (en) Inspection system
EP2869094A1 (en) Stereoscopic imaging systems and methods
EP3460531B1 (en) Scanning imaging system for security inspection of an object and imaging method thereof
US8542893B2 (en) System and method of pulmonary emboli detection for computed tomography
CN207263664U (zh) 辐射检查系统
EP1918738A2 (en) Methods for determining a position and shape of a bag placed in a baggage handling container using x-ray image analysis
WO2009015563A1 (fr) Procédé et système d'identification de matériau par le biais d'images binoculaires de transmission d'énergie multiple
Frosio et al. Optimized acquisition geometry for X-ray inspection
JP2015524061A (ja) 放射線撮影装置で生じた映像で板型物体を分離する方法と装置
CN107292888A (zh) 待测物中屏蔽区域的获取方法、装置以及检测系统
CN106910165A (zh) 修复原始ct投影数据的方法及装置、ct成像系统

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 08783607

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 08783607

Country of ref document: EP

Kind code of ref document: A1