An explosive recognition system and method

Abstract

The present invention relates to an explosive identification system and method, belonging to the technical field of security inspection, and solves the problem that the prior art cannot separately identify explosives in different states. The explosive identification system includes: an artificial intelligence identification container module, a three-dimensional object segmentation module, and an explosive identification module. The explosive identification module includes a container judgment unit, an explosive matching unit, a liquid and powdery explosive identification library, and a solid explosive identification library. This system can separately identify liquid and powder explosives from solid explosives, improving the accuracy of explosive identification.

Description

Technical Field

[0001] The present invention relates to the technical field of security inspection, and in particular, to an explosive recognition system and method. Background Art

[0002] With the increasing demand for public safety, security inspections are required in crowded public areas such as subways and airports. There are many types of explosives, including liquid explosives, powdered explosives, and solid explosives according to their types. Currently, the main technical methods for detecting explosives are divided into two categories: microtrace explosive detection and bulk explosive detection. Among them, microtrace explosive detection refers to the technology of sampling and analyzing trace or microtrace explosive residues. The basic principle is that residues in the form of gas or solid particles will always be left during the processing of explosives. These residues are collected and analyzed using relevant detection technologies to determine whether there are explosives. Common microtrace explosive detection technologies include electrochemical technology and laser Raman spectroscopy technology.

[0003] Bulk explosive detection refers to detecting visible amounts of explosives. It usually includes X-ray, γ-ray imaging technology, and nuclear-based technology. Both X-ray and γ-ray detect characteristic quantities such as the density and atomic number of substances through high-energy electromagnetic waves. Nuclear technology mainly includes nuclear quadrupole resonance technology and neutron technology.

[0004] Dual-energy CT technology reconstructs tomographic images of the scanned object containing density and atomic number from the projection data of X-rays. By analyzing these characteristic data, hidden objects can be effectively identified, and dangerous items in the scanned object can be recognized. However, density and atomic number are not the only attributes of explosives. Some common safety items may also have density and atomic number similar to those of explosives. If the state differences of explosives are not considered, the false alarm rate will increase significantly. Summary of the Invention

[0005] In view of the above analysis, embodiments of the present invention aim to provide an explosive recognition system and method to solve the problem that existing systems and methods cannot separately recognize explosives in different states.

[0006] On the one hand, an embodiment of the present invention provides an explosive recognition system, which includes:

[0007] An artificial intelligence recognition container module for recognizing the bounding box of a container;

[0008] A three-dimensional object segmentation module for segmenting three-dimensional objects in three-dimensional volume data and obtaining the bounding box of each three-dimensional object;

[0009] An explosive identification module, which includes a container judgment unit, an explosive matching unit, a liquid and powdery explosive identification library, and a solid explosive identification library. The container judgment unit is used to match the bounding box of each three-dimensional object with the bounding box of the container identified by the artificial intelligence, and judge whether the three-dimensional object is wrapped by a container. The explosive matching unit determines whether the three-dimensional object enters the liquid and powdery explosive identification library or the solid explosive identification library for matching according to the judgment result of the container judgment unit.

[0010] Preferably, the artificial intelligence container identification module includes a CT detection unit and an artificial intelligence model identification unit. The CT detection unit is used for real-time online detection to obtain three-dimensional orthographic DR images at different angles. The artificial intelligence model identification unit is used to identify the DR images obtained by the CT detection unit, and perform an intersection operation based on the identification results at different angles to obtain the bounding box of the three-dimensional container identification.

[0011] Preferably, the artificial intelligence container identification module further includes an artificial intelligence model training unit, which is used to perform artificial intelligence model training before real-time online detection to select an accurate artificial intelligence model.

[0012] Preferably, the artificial intelligence model training unit includes a data acquisition unit and a model training unit.

[0013] The data acquisition unit is used to acquire the density and atomic number data of CT reconstruction, perform orthographic projection at different angles to obtain high- and low-energy projection data, and obtain a colored DR image after color assignment.

[0014] The model training unit is used to pre-collect the CT reconstruction data of the container, use the DR images obtained by the data acquisition unit as the training set, perform prediction using the artificial intelligence model to obtain the spatial position of the container, calculate the loss function. If the loss function meets the requirements, the artificial intelligence model in the current state can be used for container identification. If the loss function does not meet the requirements, the artificial intelligence model is corrected and re-identified until the conditions are met.

[0015] Preferably, the three-dimensional object segmentation module includes a two-dimensional tomography acquisition unit, a two-dimensional object segmentation unit, a connectivity judgment unit, and a three-dimensional segmentation result unit.

[0016] The two-dimensional tomography acquisition unit is used to reconstruct the projection data at different angles of the scanned piece obtained by dual-energy CT scanning to obtain multiple two-dimensional tomographies, and sequentially number the two-dimensional tomographies.

[0017] The two-dimensional object segmentation unit is used to segment the two-dimensional objects in each two-dimensional slice, obtain the features of each two-dimensional object in each two-dimensional slice, and number the two-dimensional objects in each two-dimensional slice in sequence;

[0018] The connectivity judgment unit is used to analyze the three-dimensional body connectivity and obtain the connectivity result;

[0019] The three-dimensional segmentation result unit is used to define the start position and end position of the slice of each three-dimensional object as Z1 and Z2 of the bounding box; the union of the minimum rectangular boxes of each slice is taken, and the obtained result is defined as [X1, Y1, X2, Y2]; the bounding box of the final three-dimensional object is [X1, Y1, Z1, X2, Y2, Z2]; where X1, Y1, Z1, X2, Y2, Z2 are the x, y, z coordinates of the start point and end point of the minimum circumscribed cube of the three-dimensional object respectively.

[0020] Preferably, the three-dimensional object segmentation module further includes an image preprocessing unit, which is used to preprocess the two-dimensional slice image obtained by the two-dimensional slice acquisition unit to improve the signal-to-noise ratio of the image.

[0021] Preferably, the connectivity judgment unit includes a preliminary judgment unit, a connected region determination unit, a return control unit, and a repeat control unit;

[0022] The preliminary judgment unit is used to use the kth two-dimensional object M in the ith two-dimensional slice ik as a basis to traverse each two-dimensional object in the (i - 1)th two-dimensional slice, and judge whether the differences in the central position, area, average density, and average atomic number between each two-dimensional object in the (i - 1)th two-dimensional slice and the two-dimensional object M ik meet the threshold requirements; the connected region determination unit is used to, when the preliminary judgment unit determines that there is a two-dimensional object in the (i - 1)th two-dimensional slice that meets the threshold requirements, according to the differences in central position, area, average density, and average atomic number, comprehensively score the two-dimensional objects that meet the threshold requirements. The smaller the difference, the higher the score, and the two-dimensional object with the highest comprehensive score is determined as the two-dimensional object in the (i - 1)th two-dimensional slice that is connected to the two-dimensional object M ikConnected regions with connectivity; the return control unit is used to return the program to the preliminary judgment unit when the preliminary judgment unit determines that there is no two-dimensional object in the (i - 1)-th two-dimensional slice that meets the threshold requirement, so that the preliminary judgment unit traverses each two-dimensional object in the (i - j)-th two-dimensional slice and repeats the preliminary judgment; where i ≥ 2, k ≥ 1, i ≥ j ≥ 2; j takes values from small to large, and only when there is no two-dimensional object in the currently taken two-dimensional slice that meets the threshold requirement, j further takes a value 1 greater than the current value; the repetition control unit is used to control the preliminary judgment unit, the connected region determination unit, and the return control unit to repeat until the connectivity analysis of each two-dimensional object in each two-dimensional slice is completed. Preferably, the explosive identification module further includes an explosive judgment unit, and the explosive judgment unit is used to judge whether a three-dimensional object is an explosive.

[0023] Preferably, the explosive identification system further includes a result output module, and the result output module outputs corresponding results according to the judgment result of the explosive judgment unit.

[0024] On the other hand, the present invention also provides an explosive identification method, using the above explosive identification system, and the explosive identification method includes:

[0025] Before real-time online detection, the artificial intelligence model training unit conducts artificial intelligence model training to select an accurate artificial intelligence model, including: the data acquisition unit acquires the density and atomic number data of CT reconstruction, performs forward projection at different angles to obtain high- and low-energy projection data, and obtains a colored DR image after color assignment; the model training unit pre-collects the CT reconstruction data of the container, uses the DR image obtained by the data acquisition unit as the training set, performs prediction using an artificial intelligence model to obtain the spatial position of the container; calculates the loss function, and if the loss function meets the requirements, then the artificial intelligence model in the current state can be used for container identification; if the loss function does not meet the requirements, the artificial intelligence model is corrected and re-identified until the conditions are met;

[0026] The CT detection unit performs real-time online detection to obtain three-dimensional forward projection DR images at different angles; the artificial intelligence model identifies the DR images and performs intersection operation according to the identification results at different angles to obtain the bounding box of the three-dimensional identification of the container.

[0027] The three-dimensional object segmentation module performs connectivity analysis based on the parameter difference between the basic two-dimensional objects in the basic two-dimensional tomogram and the two-dimensional objects in the two-dimensional tomograms above the basic two-dimensional tomogram, and realizes the segmentation of three-dimensional objects in the three-dimensional volume data; the three-dimensional segmentation result unit returns to the control unit and defines the start position and end position of the tomogram of each three-dimensional object as Z1 and Z2 of the bounding box; the union of the minimum rectangular boxes of each tomogram is taken, and the obtained result is defined as [X1, Y1, X2, Y2]; the bounding box of the final three-dimensional object is [X1, Y1, Z1, X2, Y2, Z2]; where X1, Y1, Z1, X2, Y2, Z2 are the x, y, z coordinates of the start point and end point of the minimum circumscribed cube of the three-dimensional object respectively.

[0028] The container judgment unit in the explosive identification module matches the bounding box of each three-dimensional object with the bounding box of the container identified by the artificial intelligence to judge whether the three-dimensional object is wrapped by a container; the explosive matching unit determines that the three-dimensional object enters the liquid and powdery explosive identification library or the solid explosive identification library for matching according to the judgment result of the container judgment unit, and the explosive judgment unit judges whether the three-dimensional object is an explosive.

[0029] According to the judgment result of the explosive judgment unit, the result output module outputs the corresponding result.

[0030] Compared with the prior art, the present invention can at least achieve one of the following beneficial effects:

[0031] 1. For the existing security inspection CT equipment, the artificial intelligence identification container module in the present invention identifies the container and outputs the spatial position where the container is located. The three-dimensional object segmentation module segments the three-dimensional objects in the three-dimensional volume data and obtains the bounding box of each three-dimensional object, and matches the position of the three-dimensional object with the position of the container identified by the artificial intelligence; judges whether each object is wrapped by a container. If so, it is a liquid or powdery object, and liquid and powdery explosive matching is performed; otherwise, it is a solid, and solid explosive matching is performed. This method can quickly identify liquid explosives and powdery explosives separately from solid explosives, improving the accuracy of explosive identification.

[0032] 2. The three-dimensional object segmentation module of the present invention performs connectivity judgment with each two-dimensional object in the two-dimensional tomogram as the connectivity judgment object, that is, the present invention is two-dimensional region growing. Compared with the three-dimensional region growing method in the prior art, the present invention reduces the range of region growing from 26 pixels in three dimensions to 8 pixels in two dimensions. Therefore, the three-dimensional object segmentation module of the present invention is faster, and thus the explosive can be identified faster.

[0033] 3. Compared with the threshold method of the prior art, due to the presence of artifacts in CT images, the threshold method cannot accurately segment all objects. However, the present invention is a two-dimensional region growing method that can accommodate differences and reduce the influence of artifacts. Therefore, the three-dimensional object segmentation module of the present invention has a higher accuracy in three-dimensional object segmentation, and thus can more accurately identify explosives.

[0034] 4. Compared with the boundary method of the prior art, since the boundary method requires first-order or second-order differential calculations and iterative operations, while the three-dimensional object segmentation module of the present invention does not require any iteration for three-dimensional object segmentation. Therefore, the three-dimensional object segmentation module of the present invention is faster, and thus can identify explosives faster.

[0035] In the present invention, the above technical solutions can also be combined with each other to achieve more preferred combination schemes. Other features and advantages of the present invention will be described in the subsequent specification. Moreover, some advantages can be made obvious from the specification, or can be understood by implementing the present invention. The objectives and other advantages of the present invention can be realized and obtained from the content specifically pointed out in the specification and the drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] The drawings are only for the purpose of showing specific embodiments, and are not considered as a limitation of the present invention. Throughout the drawings, the same reference signs represent the same components.

[0037] Figure 1 is the principle flowchart of the explosive identification system of the present invention;

[0038] Figure 2 is the principle flowchart of the three-dimensional object segmentation module of the present invention;

[0039] Figure 3 is the schematic diagram before and after the number change of each two-dimensional object after connectivity analysis using the three-dimensional object segmentation module of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0040] The following will specifically describe the preferred embodiments of the present invention with reference to the drawings. The drawings form a part of this application and are used together with the embodiments of the present invention to explain the principle of the present invention, but are not used to limit the scope of the present invention.

[0041] In the field of security inspection, the technology of identifying explosives through dual-energy CT technology calculates the average electron density and atomic number corresponding to each item through dual-energy CT technology, and then compares them with the known explosive information to determine whether it is an explosive. However, there are many types of explosives. If the state of the object is not considered, there will be many false alarms for explosives. Therefore, identifying the corresponding state of explosives is the key.

[0042] Moreover, during the security inspection process, after the CT scan of the package is completed, it is conveyed to the user through a conveyor belt, and explosives need to be identified during this short process. Therefore, rapid three-dimensional segmentation of the items and identification of explosives are necessary.

[0043] Thus, the present invention provides an explosive identification system, as Figure 1 shown, the explosive identification system includes:

[0044] 1. An artificial intelligence identification container module for identifying the bounding box of the container.

[0045] The artificial intelligence identification container module includes an artificial intelligence model training unit, a CT detection unit, and an artificial intelligence model identification unit.

[0046] M1.1: Artificial intelligence model training unit

[0047] The artificial intelligence model training unit is used to perform artificial intelligence model training before real-time online CT detection to select an accurate artificial intelligence model for identifying the position of the container.

[0048] Specifically, the artificial intelligence model training unit includes a data acquisition unit and a model training unit;

[0049] The data acquisition unit is used to acquire the density and atomic number data of the CT reconstruction, perform forward projections at different angles (such as 0 degrees, 45 degrees, 90 degrees, 135 degrees, 180 degrees), obtain the projection data of high and low energies, and obtain a colored DR image after color assignment.

[0050] Among them, different colors represent different materials. Exemplarily, organic substances such as food and plastic are shown in orange; inorganic substances such as books and ceramics are shown in green; mixtures also mostly appear green, and stainless steel products and metals are shown in blue.

[0051] The model training unit is used to pre-collect the CT reconstruction data of the container, use the DR image obtained by the data acquisition unit as the training set, perform prediction using an artificial intelligence model (such as an RCNN neural network model) to obtain the spatial position of the container; compare the obtained spatial position of the container with the pre-collected CT reconstruction position data of the container, calculate the loss function, and if the loss function meets the requirements, then the artificial intelligence model in the current state can be used for container identification; if the loss function does not meet the requirements, then correct the artificial intelligence model and re-identify until the conditions are met. Finally, output the bounding box of the container identification result.

[0052] M1.2: CT detection unit

[0053] The CT detection unit is used for real-time online CT detection, obtaining density and atomic number data reconstructed by CT, and performing forward projections at different angles (such as 0 degrees, 45 degrees, 90 degrees, 135 degrees, 180 degrees) to obtain projection data of high and low energies, and obtaining a colored DR image after color assignment.

[0054] Among them, different colors represent different materials. Exemplarily, organic substances such as food and plastics are shown as orange; inorganic substances such as books and ceramics are shown as green; mixtures also mostly appear green, such as stainless steel products; metals are shown as blue.

[0055] M1.3: Artificial intelligence model recognition unit

[0056] The artificial intelligence model recognition unit is used to recognize the DR image obtained by the CT detection unit, perform intersection operations based on the recognition results at different angles, and obtain the bounding box for three-dimensional recognition of the container. The artificial intelligence model is the artificial intelligence model selected from the artificial intelligence model training unit.

[0057] It should be noted that the role of the artificial intelligence model training unit is to obtain an accurate artificial intelligence model for recognizing containers to ensure the accuracy of the artificial intelligence recognition container module during online automatic recognition. During the actual security inspection process, the artificial intelligence model training unit runs during equipment debugging, and when performing real-time online security inspection, the artificial intelligence recognition container module runs directly.

[0058] 2. The three-dimensional object segmentation module is used to segment three-dimensional objects in the three-dimensional volume data and obtain the bounding box of each three-dimensional object.

[0059] The three-dimensional object segmentation module includes a two-dimensional tomography acquisition unit, an image preprocessing unit, a two-dimensional object segmentation unit, a connectivity judgment unit, and a three-dimensional segmentation result unit.

[0060] M2.1: Two-dimensional tomography acquisition unit, which is used to reconstruct and invert the projection data of different angles (such as 0 to 360 degrees, with an angle set every 1 degree) of the scanned piece obtained by dual-energy CT scanning to obtain high-energy, low-energy, and density and atomic number data images, and then obtain multiple two-dimensional tomographies and number the two-dimensional tomographies in sequence.

[0061] M2.2: Image preprocessing unit, which is used to preprocess the two-dimensional tomography images obtained by the two-dimensional tomography acquisition unit to improve the signal-to-noise ratio of the images and make the segmentation results more accurate.

[0062] Specifically, the image preprocessing includes image smoothing, image enhancement, image region segmentation, and dilation and erosion.

[0063] M2.3: 2D object segmentation unit, which is used to segment 2D objects in each 2D slice, obtain the features of each 2D object in each 2D slice, and sequentially number the 2D objects in each 2D slice.

[0064] Specifically, the 2D object segmentation unit is a region growing method segmentation unit or a boundary method segmentation unit.

[0065] Exemplarily, the region growing method segmentation unit is used to traverse each 2D slice and select seed points in each 2D slice, then traverse the neighborhood of the seed points in the 2D slice. In the neighborhood, points with a difference less than the threshold from the seed points are used as growth points for neighborhood growth, so as to complete the segmentation of 2D objects in each 2D slice.

[0066] Exemplarily, the difference threshold for density image segmentation seed points is 0.1.

[0067] The features of the 2D object include density mean, density mean square deviation, atomic number mean, atomic number mean square deviation, object area, object perimeter, abscissa of regional center of gravity, ordinate of regional center of gravity, and minimum circumscribed rectangle.

[0068] M2.4: Connectivity judgment unit, which is used to analyze 3D body connectivity. It includes a preliminary judgment unit, a connected region determination unit, a return control unit, a number modification unit, and a repetition control unit.

[0069] M2.4.1: Preliminary judgment unit

[0070] The preliminary judgment unit is used to take the kth 2D object M in the ith 2D slice as a basis, traverse each 2D object in the (i - 1)th 2D slice, and judge whether the differences in the central position, area, average density, and average atomic number between each 2D object in the (i - 1)th 2D slice and the 2D object M meet the threshold requirements. ik For the basis, traverse each 2D object in the (i - 1)th 2D slice, and judge whether the differences in the central position, area, average density, and average atomic number between each 2D object in the (i - 1)th 2D slice and the 2D object M meet the threshold requirements. ik meet the threshold requirements.

[0071] M2.4.2: Connected region determination unit

[0072] The connected region determination unit is used to, when the preliminary judgment unit determines that there are 2D objects meeting the threshold requirements in the (i - 1)th 2D slice, perform a comprehensive scoring on the 2D objects meeting the threshold requirements according to the differences in central position, area, average density, and average atomic number. The smaller the difference, the higher the score. The 2D object with the highest comprehensive score is determined as the connected region in the (i - 1)th 2D slice that has connectivity with the 2D object M. ik has connectivity.

[0073] Exemplarily, the difference thresholds for the central position, area, average density, and average atomic number are 14, 50, 0.1, and 1 respectively.

[0074] Furthermore, a comprehensive scoring module is provided in the connected region determination unit, and the comprehensive scoring module calculates the comprehensive score according to Equation (I):

[0075]

[0076] where s0 is the area of the two-dimensional object M ik and ρ0 is the average density of the two-dimensional object M ik and Z0 is the average atomic number of the two-dimensional object M ik and x0 and y0 are the abscissa and ordinate of the center position of the two-dimensional object M respectively ik ; s1 is the area of the two-dimensional object that meets the threshold requirements, ρ1 is the average density of the two-dimensional object that meets the threshold requirements, Z1 is the average atomic number of the two-dimensional object that meets the threshold requirements, x1 and y1 are the abscissa and ordinate of the center position of the two-dimensional object that meets the threshold requirements respectively, and a, b, c, and d are parameters greater than 0.

[0077] Among them, the values of a, b, c, and d are weighting ratios, which are determined according to the difference thresholds of the corresponding features. For example, a is 14 / 4, b is 50 / 4, c is 0.1 / 4, and d is 1 / 4, indicating that the weights of the four are the same under the above difference thresholds.

[0078] M2.4.3: Return control unit

[0079] The return control unit is used to return the program to the preliminary judgment unit when the preliminary judgment unit determines that there is no two-dimensional object that meets the threshold requirements in the (i - 1)-th two-dimensional tomogram, so that the preliminary judgment unit traverses each two-dimensional object in the (i - j)-th two-dimensional tomogram and repeats the preliminary judgment; where i≥2, k≥1, i≥j≥2; j takes values from small to large, and only when there is no two-dimensional object that meets the threshold requirements in the currently taken two-dimensional tomogram, j further takes a value 1 greater than the current value.

[0080] It should be noted that i≥2, that is, the preliminary judgment unit does not analyze the connectivity of the first two-dimensional tomogram, directly takes it as the existing connected region, and directly starts analyzing the connectivity based on the two-dimensional objects in the second two-dimensional tomogram.

[0081] In the present invention, the two-dimensional tomogram without two-dimensional objects that meet the threshold requirements is defined as an error tomogram. An error tomogram limit module is provided in the return control unit, and the number of layers A of the error tomogram allowed to be traversed is set in the error tomogram limit module; when j is equal to A, if there is still no two-dimensional object that meets the threshold requirements in the (i - A)-th two-dimensional tomogram, the error tomogram limit module restricts the return control unit from returning the program to the preliminary judgment unit; where A≥1 (for example, A is 1, 2, or 3).

[0082] That is to say, when the preliminary judgment unit determines that there is no two-dimensional object meeting the threshold requirement in the (i - 1)-th two-dimensional slice, the return control unit returns the program to the preliminary judgment unit, enabling the preliminary judgment unit to traverse each two-dimensional object in the (i - j)-th two-dimensional slice, where i ≥ j ≥ 2 and j takes values from small to large. The preliminary judgment is repeated. Only when there is no two-dimensional object meeting the threshold requirement in the currently selected two-dimensional slice, will j further take a value that is 1 greater than the current value, until a two-dimensional object (connected region) meeting the threshold requirement is found or j = A; if there is a two-dimensional object meeting the threshold requirement in the two-dimensional slice corresponding to the currently selected j, this traversal ends. If a two-dimensional object meeting the threshold requirement is never found, this traversal ends.

[0083] M2.4.4: Number Modification Unit

[0084] The number modification unit is used to modify the number of the two-dimensional object M ik to be the same as the number of the two-dimensional object with the highest comprehensive score determined by the connected region determination unit. Or, when the preliminary judgment unit never finds a two-dimensional object meeting the threshold requirement, the number modification unit modifies the number of the two-dimensional object M ik to be a number that is 1 greater than the maximum number of the connected region that has completed connectivity analysis.

[0085] It should be noted that each two-dimensional object in the first slice can be regarded as a connected region that has completed connectivity analysis.

[0086] M2.4.5: Repetition Control Unit

[0087] The repetition control unit is used to control the preliminary judgment unit, the connected region determination unit, and the return control unit to repeat the process until the connectivity analysis of each two-dimensional object in each two-dimensional slice is completed. The repetition control unit includes a two-dimensional object repetition control unit and a two-dimensional slice repetition control unit.

[0088] The two-dimensional object repetition control unit includes a two-dimensional object repetition control unit and a two-dimensional slice repetition control unit.

[0089] The two-dimensional object repetition control unit is used to control the preliminary judgment unit, the connected region determination unit, and the return control unit to perform the connectivity analysis of each two-dimensional object in the i-th two-dimensional slice.

[0090] The two-dimensional slice repetition control unit is used to control the preliminary judgment unit, the connected region determination unit, and the return control unit to perform the connectivity analysis of the two-dimensional objects in each two-dimensional slice.

[0091] M2.5: 3D object feature acquisition unit, which is used to combine 2D objects with connectivity to form 3D objects and acquire the features of each 3D object.

[0092] The features of the 3D object include spatial position, volume, average density, average atomic number, density mean square deviation, and atomic number mean square deviation. The corresponding features of the 3D object are as follows:

[0093]

[0094]

[0095]

[0096]

[0097]

[0098] Among them, S i represents the area corresponding to the i-th pixel of the 3D object, Voxel represents the pixel size corresponding to the pixel, ρ i and Z i respectively represent the density value and atomic number value corresponding to the pixel; V, ρ mean , Z mean , σ ρ , σ Z respectively represent the volume, average density, average atomic number, density mean square deviation, and atomic number mean square deviation corresponding to the 3D object.

[0099] In order to quickly obtain the mean square deviation feature, the density mean square deviation and the atomic number mean square deviation can be abbreviated as σ ρsimple and σ Zsimple :

[0100]

[0101]

[0102] Among them, S k,i , ρ k,i , Z k,i respectively represent the area, density, and atomic number corresponding to the i-th pixel in the k-th (k = 1 - K) layer slice.

[0103] M2.6: 3D segmentation result unit

[0104] The three-dimensional segmentation result unit is used to define the starting position and ending position of the fault of each three-dimensional object as Z1 and Z2 of the bounding box; the minimum rectangular box of each fault is taken as the union, and the result is defined as [X1, Y1, X2, Y2]; the final bounding box of the three-dimensional object is [X1, Y1, Z1, X2, Y2, Z2]; wherein X1, Y1, Z1, X2, Y2, Z2 are the x, y, z coordinates of the starting point and end point of the minimum circumscribed cube of the three-dimensional object, respectively.

[0105] 3. Explosive identification module.

[0106] The explosive identification module includes a container judgment unit, an explosive matching unit, an explosive judgment unit, a liquid and powder explosive identification library, and a solid explosive identification library.

[0107] M3.1: Container judgment unit

[0108] The container determination unit is used to match the bounding box of each three-dimensional object with the bounding box of the container identified by the artificial intelligence to determine whether the three-dimensional object is enclosed by a container.

[0109] M3.2: Explosives Matching Unit

[0110] The explosive matching unit determines that the three-dimensional object enters the liquid and powder explosive identification library or the solid explosive identification library for matching according to the judgment result of the container judgment unit.

[0111] Specifically, if the judgment result of the container judgment unit is that the three-dimensional object is wrapped in a container, the explosive matching unit will enter the three-dimensional object into the liquid and powdered explosive identification library for matching; if the judgment result of the container judgment unit is that the three-dimensional object is not wrapped in a container, the explosive matching unit will enter the three-dimensional object into the solid explosive identification library for matching.

[0112] It should be noted that the term "enclosed" means that the bounding box of the three-dimensional object is smaller than and is close to the bounding box of the container. Since liquid and powdered substances are fluid, they must be contained in a container and in contact with the inner wall of the container. Therefore, when the bounding box of the three-dimensional object is smaller than and is close to the bounding box of the container, it can be determined that the three-dimensional object is liquid or powdered, and it is entered into the liquid and powdered explosives recognition library for matching; otherwise, the three-dimensional object is solid, and it is entered into the solid explosives recognition library for matching.

[0113] It is worth noting that the "solid" of the present invention refers to a block solid that does not have fluidity and cannot completely contact the inner wall of the container in the container. The "container" of the present invention refers to a container that can hold liquid or powder substances, such as a bottle, a can, etc.

[0114] M3.3: Explosive Judgment Unit

[0115] The explosive judgment unit is used to judge whether a three-dimensional object is an explosive. That is, the explosive judgment unit is used to judge whether the three-dimensional object successfully matches the explosives in the explosive database it enters.

[0116] The way for the explosive judgment unit to judge explosives can be an autonomous judgment criterion or an artificial intelligence method.

[0117] Exemplarily, the autonomous judgment criterion includes: judging autonomously according to features, with the density deviation less than a certain number, the atomic number deviation less than a certain number, etc., then it is determined that the match is successful; otherwise, the match is unsuccessful.

[0118] Exemplarily, the method of using artificial intelligence to judge includes: judging according to features in an artificial intelligence way, specifically as follows:

[0119] According to the feature values of the three-dimensional object obtained by the connectivity judgment unit, perform feature classification of explosive data:

[0120] (1) Input: The selected feature variables and the corresponding explosive attributes.

[0121] (2) According to the features and attributes, use different classification methods, such as SVM, K-NN, to obtain a classification model.

[0122] (3) Calculate the result. If the result meets the requirements, then this model can use the current artificial intelligence model to perform substance feature recognition; otherwise, correct the model and perform recognition again until the conditions are met.

[0123] (4) Output: The attributes of the corresponding object.

[0124] This matching method automatically compares through an artificial intelligence method, and finally compares each object with the explosives in the database to obtain the matching results of each object.

[0125] 4. Result Output Module

[0126] The result output module outputs corresponding results according to the judgment results of the explosive judgment unit.

[0127] Specifically, if the judgment result of the explosive judgment unit is an explosive, then the result output module outputs the location of the explosive and information such as the name, weight, volume, density of the explosive. If the judgment result of the explosive judgment unit is that none of the objects in a package are explosives, then the result is directly output.

[0128] Among them, the location of the explosive can be directly output according to the location result of the corresponding object obtained by the three-dimensional segmentation result unit.

[0129] On the other hand, the present invention also provides an explosive identification method, which uses the above explosive identification system. The explosive identification method includes:

[0130] Before real-time online detection, the artificial intelligence model training unit trains an artificial intelligence model to select an accurate artificial intelligence model, including: the data acquisition unit acquires density and atomic number data reconstructed by CT and performs forward projection at different angles to obtain projection data of high and low energies, and after color assignment, a colored DR image is obtained; the model training unit pre-collects CT reconstruction data of the container, uses the DR image obtained by the data acquisition unit as a training set, performs prediction using an artificial intelligence model to obtain the spatial position of the container; calculates the loss function. If the loss function meets the requirements, then the artificial intelligence model in the current state can be used for container identification; if the loss function does not meet the requirements, the artificial intelligence model is corrected and identification is performed again until the conditions are met;

[0131] The CT detection unit performs real-time online detection to obtain three-dimensional forward projection DR images at different angles; the artificial intelligence model identifies the DR images, and performs intersection operation based on the identification results at different angles to obtain the bounding box for three-dimensional identification of the container;

[0132] The two-dimensional tomography acquisition unit reconstructs the projection data of different angles of the scanned piece obtained by dual-energy CT scanning to obtain a plurality of two-dimensional tomographies, and numbers the two-dimensional tomographies in sequence;

[0133] The image preprocessing unit performs image smoothing, image enhancement, image region segmentation, and dilation and erosion processing on the two-dimensional tomography images obtained by the two-dimensional tomography acquisition unit;

[0134] The two-dimensional object segmentation unit segments the two-dimensional objects in each two-dimensional tomography, obtains the features of each two-dimensional object in each two-dimensional tomography, and numbers the two-dimensional objects in each two-dimensional tomography in sequence;

[0135] Based on the k-th two-dimensional object M in the i-th two-dimensional tomography ik as the basis, the preliminary judgment unit traverses each two-dimensional object in the (i - 1)-th two-dimensional tomography and judges each two-dimensional object in the (i - 1)-th two-dimensional tomography respectively with the two-dimensional object M ikWhether the differences in the central position, area, average density, and average atomic number meet the threshold requirements; if the preliminary judgment unit determines that there is a two-dimensional object in the (i - 1)-th two-dimensional tomogram that meets the threshold requirements, the connected region determination unit comprehensively scores the two-dimensional object that meets the threshold requirements based on the differences in central position, area, average density, and average atomic number. The smaller the difference, the higher the score. The two-dimensional object with the highest comprehensive score is determined as the connected region in the (i - 1)-th two-dimensional tomogram that has connectivity with the two-dimensional object M ik A connected region with connectivity; if the preliminary judgment unit determines that there is no two-dimensional object in the (i - 1)-th two-dimensional tomogram that meets the threshold requirements, the return control unit returns the program to the preliminary judgment unit, causing the preliminary judgment unit to traverse each two-dimensional object in the (i - j)-th two-dimensional tomogram and repeat the preliminary judgment; where i ≥ 2, k ≥ 1, i ≥ j ≥ 2; j takes values from small to large. Only when there is no two-dimensional object in the currently selected two-dimensional tomogram that meets the threshold requirements, j further takes a value that is 1 greater than the current value; when j is equal to A, if there is still no two-dimensional object in the (i - A)-th two-dimensional tomogram that meets the threshold requirements, the error tomogram limitation module restricts the return control unit from returning the program to the preliminary judgment unit; the two-dimensional object repetition control unit controls the preliminary judgment unit, the connected region determination unit, and the return control unit to perform connectivity analysis on each two-dimensional object in the i-th two-dimensional tomogram; the two-dimensional tomogram repetition control unit controls the preliminary judgment unit, the connected region determination unit, and the return control unit to perform connectivity analysis on the two-dimensional objects in each two-dimensional tomogram

[0136] The number modification unit modifies the number of the two-dimensional object M ik To the same number as the two-dimensional object with the highest comprehensive score determined by the connected region determination unit, or, when the preliminary judgment unit never finds a two-dimensional object that meets the threshold requirements, the number modification unit modifies the number of the two-dimensional object M ik To a number that is 1 greater than the largest number in the connected region where connectivity analysis has been completed

[0137] The three-dimensional object feature acquisition unit combines the two-dimensional objects with connectivity to form a three-dimensional object and acquires the features of each three-dimensional object

[0138] The three-dimensional segmentation result unit defines the start position and end position of each three-dimensional object's tomogram as Z1 and Z2 of the bounding box; the union of the smallest rectangular boxes of each tomogram is taken, and the resulting result is defined as [X1, Y1, X2, Y2]; the bounding box of the final three-dimensional object is [X1, Y1, Z1, X2, Y2, Z2]; where X1, Y1, Z1, X2, Y2, Z2 are the x, y, z coordinates of the starting point and ending point of the smallest circumscribed cube of the three-dimensional object

[0139] The container judgment unit in the explosive identification module matches the bounding box of each three-dimensional object with the bounding box of the container identified by the artificial intelligence to determine whether the three-dimensional object is wrapped by a container; the explosive matching unit determines that the three-dimensional object enters the liquid and powdery explosive identification library or the solid explosive identification library for matching according to the judgment result of the container judgment unit, and the explosive judgment unit judges whether the three-dimensional object is an explosive;

[0140] According to the judgment result of the explosive judgment unit, the corresponding result is output through the result output module.

[0141] To further clearly illustrate the analysis process of connectivity by the two-dimensional tomography acquisition unit, the image preprocessing unit, the two-dimensional object segmentation unit, and the connectivity judgment unit, take Figure 3 as an example for illustration.

[0142] S1: The two-dimensional tomography acquisition unit reconstructs the projection data of different angles of the scanned piece obtained by dual-energy CT scanning to obtain 4 two-dimensional tomographies ( Figure 3 Only 4 two-dimensional tomographies are shown for the sake of illustration), and the image preprocessing unit performs image smoothing, image enhancement, image region segmentation, and dilation and erosion processing on the two-dimensional tomography images obtained by the two-dimensional tomography acquisition unit, and numbers the 4 two-dimensional tomographies in order from top to bottom, which are 1, 2, 3, and 4 respectively;

[0143] S2: The two-dimensional object segmentation unit segments the two-dimensional objects in each two-dimensional tomography, obtains the features of each two-dimensional object in each two-dimensional tomography, numbers the two-dimensional objects in each two-dimensional tomography in order, numbers the two-dimensional objects in the first two-dimensional tomography as 1, 2, 3, 4, numbers the two-dimensional objects in the second two-dimensional tomography as 1, 2, 3, numbers the two-dimensional objects in the third two-dimensional tomography as 1, 2, and numbers the two-dimensional objects in the fourth two-dimensional tomography as 1, 2, 3;

[0144] S3: The connectivity judgment unit analyzes and judges the connectivity of the three-dimensional body:

[0145] S3.1: First, analyze the connectivity of the second two-dimensional tomography. Based on the first two-dimensional object M 21 in the second two-dimensional tomography, the preliminary judgment unit traverses each two-dimensional object in the first two-dimensional tomography. The first and second two-dimensional objects in the first two-dimensional tomography meet the threshold requirements, and the connectivity judgment unit judges that the first two-dimensional object has the highest comprehensive score. Then, the first two-dimensional object in the first two-dimensional tomography is the connectivity region of the two-dimensional object M 21 and the number modification unit changes the number of the two-dimensional object M 21The number is changed to be the same as the number of the first two-dimensional object in the first two-dimensional slice (the original numbers are both 1, and no change is needed here); the two-dimensional object repetition control unit controls the preliminary judgment unit to use the second two-dimensional object M in the second two-dimensional slice 22 As a basis, traverse each two-dimensional object in the first two-dimensional slice. The second and third two-dimensional objects in the first two-dimensional slice meet the threshold requirements, and the connectivity judgment unit determines that the third two-dimensional object has the highest comprehensive score. Then, the third two-dimensional object in the first two-dimensional slice is the two-dimensional object M 22 In the connected region, the number modification unit changes the number of the two-dimensional object M 22 To be the same as the number of the third two-dimensional object in the first two-dimensional slice, that is, to change the number of the two-dimensional object M 22 To 3; the two-dimensional object repetition control unit controls the preliminary judgment unit to use the third two-dimensional object M in the second two-dimensional slice 23 As a basis, traverse each two-dimensional object in the first two-dimensional slice. There is no two-dimensional object in the first two-dimensional slice that meets the threshold requirements, and the maximum number of the two-dimensional objects in the first two-dimensional slice is 4. The number modification unit changes the number of the two-dimensional object M 23 To 5; the connectivity analysis of the second two-dimensional slice is completed.

[0146] The two-dimensional slice repetition control unit controls the preliminary judgment unit, the connected region determination unit, and the return control unit to analyze the connectivity of the third two-dimensional slice. Using the first two-dimensional object M in the third two-dimensional slice 31 As a basis, the preliminary judgment unit traverses each two-dimensional object in the second two-dimensional slice. There is no two-dimensional object in the second two-dimensional slice that meets the threshold requirements. The return control unit controls the program to return to the preliminary judgment unit. The preliminary judgment unit traverses each two-dimensional object in the first two-dimensional slice. The second and third two-dimensional objects in the first two-dimensional slice meet the threshold requirements, and the connectivity judgment unit determines that the second two-dimensional object has the highest comprehensive score. That is, the second two-dimensional object in the first two-dimensional slice is the two-dimensional object M 31 In the connected region, the number modification unit changes the number of the two-dimensional object M 31 To be the same as the number of the second two-dimensional object in the first two-dimensional slice, that is, to change the number of the two-dimensional object M 31 To 2; the two-dimensional object repetition control unit controls the preliminary judgment unit to use the second two-dimensional object M in the third two-dimensional slice 32 As a basis, traverse each two-dimensional object in the second two-dimensional slice. The two-dimensional objects numbered 1 and 5 in the second two-dimensional slice meet the threshold requirements. The connectivity judgment unit determines that the two-dimensional object numbered 5 has the highest comprehensive score. That is, the two-dimensional object numbered 5 in the second two-dimensional slice is the two-dimensional object M 32 In the connected region, the number modification unit changes the number of the two-dimensional object M32 The number is changed to 5; the connectivity analysis of the 3rd two-dimensional fault is completed.

[0147] The two-dimensional fault repetition control unit controls the preliminary judgment unit, the connected area determination unit, and the return control unit to analyze the connectivity of the 4th two-dimensional fault, using the 1st two-dimensional object M of the 4th two-dimensional fault 41 as the basis. The preliminary judgment unit traverses each two-dimensional object in the 3rd two-dimensional fault. There is no two-dimensional object in the 3rd two-dimensional fault that meets the threshold requirement. The return control unit controls the program to return to the preliminary judgment unit. The preliminary judgment unit traverses each two-dimensional object in the 2nd two-dimensional fault. There is no two-dimensional object in the 2nd two-dimensional fault that meets the threshold requirement. The return control unit controls the program to return to the preliminary judgment unit. The preliminary judgment unit traverses each two-dimensional object in the 1st two-dimensional fault. There is no two-dimensional object in the 1st two-dimensional fault that meets the threshold requirement. It is determined that no two-dimensional object that meets the threshold requirement is found. The number modification unit, based on the maximum number of the connected area found previously being 5, changes the number of the two-dimensional object M 41 to 6; the two-dimensional object repetition control unit controls the preliminary judgment unit to use the 2nd two-dimensional object M of the 4th two-dimensional fault 42 as the basis, traverses each two-dimensional object in the 3rd two-dimensional fault. There is no two-dimensional object in the 3rd two-dimensional fault that meets the threshold requirement. The return control unit controls the program to return to the preliminary judgment unit. The preliminary judgment unit traverses each two-dimensional object in the 2nd two-dimensional fault. There is no two-dimensional object in the 2nd two-dimensional fault that meets the threshold requirement. The return control unit controls the program to return to the preliminary judgment unit. The preliminary judgment unit traverses each two-dimensional object in the 1st two-dimensional fault. There is no two-dimensional object in the 1st two-dimensional fault that meets the threshold requirement. It is determined that no two-dimensional object that meets the threshold requirement is found. The number modification unit, based on the maximum number of the connected area found previously being 6, changes the number of the two-dimensional object M 42 to 7; the two-dimensional object repetition control unit controls the preliminary judgment unit to use the 3rd two-dimensional object M of the 4th two-dimensional fault 43 as the basis, traverses each two-dimensional object in the 3rd two-dimensional fault. The two-dimensional objects numbered 5 and 3 in the 3rd two-dimensional fault meet the threshold requirement. The connectivity judgment unit determines that the two-dimensional object numbered 5 has the highest comprehensive score. The number modification unit changes the number of the two-dimensional object M 43 to 5. That is, the connectivity analysis of all two-dimensional objects in all two-dimensional faults is completed.

[0148] Compared with the prior art, the present invention can achieve at least the following beneficial effects: For the existing security inspection CT equipment, the present invention uses artificial intelligence to identify containers and output the spatial positions where the containers are located; uses fast three-dimensional segmentation technology to output features, and matches the position features of three-dimensional objects with the positions of containers identified by artificial intelligence; determines whether each object is wrapped by a container. If so, it is a liquid or powdery object, and liquid and powdery explosives are matched; otherwise, it is a solid, and solid explosives are matched. This system and method can quickly identify liquid explosives and powdery explosives separately from solid explosives, improving the accuracy of explosive identification. The three-dimensional object segmentation module of the present invention performs connectivity judgment on each two-dimensional object in the two-dimensional tomogram, that is, the present invention is two-dimensional region growing. Compared with the three-dimensional region growing method in the prior art, the present invention reduces the range of region growing from 26 pixels in three dimensions to 8 pixels in two dimensions. Therefore, the three-dimensional object segmentation module of the present invention is faster, and thus faster in identifying explosives. Compared with the threshold method in the prior art, due to the presence of artifact phenomena in CT images, the threshold method cannot accurately segment all objects, while the present invention is two-dimensional region growing, which can accommodate differences and reduce the influence of artifacts. Therefore, the three-dimensional object segmentation module of the present invention has higher accuracy in segmenting three-dimensional objects, and thus more accurately identifies explosives. Compared with the boundary method in the prior art, since the boundary method requires first-order or second-order differential calculations and iterative operations, while the three-dimensional object segmentation module of the present invention does not require any iteration in segmenting three-dimensional objects, the three-dimensional object segmentation module of the present invention is faster, and thus faster in identifying explosives.

[0149] The explosive identification system and method of the present invention are further described below through specific embodiments.

[0150] Embodiment 1

[0151] An explosive identification system includes: an artificial intelligence container identification module, a three-dimensional object segmentation module, an explosive identification module, and a result output module;

[0152] The artificial intelligence container identification module includes an artificial intelligence model training unit, a CT detection unit, and an artificial intelligence model identification unit. The artificial intelligence model training unit includes a data acquisition unit and a model training unit.

[0153] The three-dimensional object segmentation module includes a two-dimensional tomogram acquisition unit, an image preprocessing unit, a two-dimensional object segmentation unit, a connectivity judgment unit, a three-dimensional object feature acquisition unit, and a three-dimensional segmentation result unit; the connectivity judgment unit includes a preliminary judgment unit, a connected region determination unit, a return control unit, a number modification unit, and a repetition control unit.

[0154] The explosive recognition module includes a container judgment unit, an explosive matching unit, an explosive judgment unit, a liquid and powdery explosive recognition library, and a solid explosive recognition library.

[0155] Embodiment 2

[0156] Using the system of Embodiment 1 for explosive recognition, including:

[0157] Before real-time online detection, the artificial intelligence model training unit conducts artificial intelligence model training to select an accurate artificial intelligence model, including: the data acquisition unit acquires the density and atomic number data of CT reconstruction, performs forward projections at different angles to obtain high- and low-energy projection data, and obtains a colored DR image after color assignment; the model training unit pre-collects the CT reconstruction data of the container, uses the DR image obtained by the data acquisition unit as the training set, performs prediction using an artificial intelligence model to obtain the spatial position of the container; calculates the loss function, and if the loss function meets the requirements, then the artificial intelligence model in the current state can be applied for container recognition; if the loss function does not meet the requirements, the artificial intelligence model is corrected and recognition is performed again until the conditions are met;

[0158] The CT detection unit performs real-time online detection to obtain three-dimensional forward projection DR images at different angles; identifies the DR images through an artificial intelligence model, and performs intersection operation based on the recognition results at different angles to obtain the bounding box for three-dimensional recognition of the container;

[0159] The two-dimensional tomography acquisition unit reconstructs the projection data of the scanned piece at different angles obtained by dual-energy CT scanning to obtain a plurality of two-dimensional tomographies, and numbers the two-dimensional tomographies in sequence;

[0160] The image preprocessing unit performs image smoothing, image enhancement, image region segmentation, and dilation and erosion processing on the two-dimensional tomography images obtained by the two-dimensional tomography acquisition unit;

[0161] The two-dimensional object segmentation unit segments the two-dimensional objects in each two-dimensional tomography, obtains the features of each two-dimensional object in each two-dimensional tomography, and numbers the two-dimensional objects in each two-dimensional tomography in sequence;

[0162] Based on the kth two-dimensional object M in the ith two-dimensional tomography ik as a basis, the preliminary judgment unit traverses each two-dimensional object in the (i - 1)th two-dimensional tomography, and judges each two-dimensional object in the (i - 1)th two-dimensional tomography respectively with the two-dimensional object M ikWhether the differences in the central position, area, average density, and average atomic number meet the threshold requirements; if the preliminary judgment unit determines that there is a two-dimensional object in the (i-1)-th two-dimensional tomogram that meets the threshold requirements, the connected region determination unit comprehensively scores the two-dimensional object that meets the threshold requirements based on the differences in central position, area, average density, and average atomic number. The smaller the difference, the higher the score. The two-dimensional object with the highest comprehensive score is determined as the connected region in the (i-1)-th two-dimensional tomogram that has connectivity with the two-dimensional object M ik If the preliminary judgment unit determines that there is no two-dimensional object in the (i-1)-th two-dimensional tomogram that meets the threshold requirements, the return control unit returns the program to the preliminary judgment unit, causing the preliminary judgment unit to traverse each two-dimensional object in the (i-j)-th two-dimensional tomogram and repeat the preliminary judgment; where i≥2, k≥1, i≥j≥2; j takes values from small to large. Only when there is no two-dimensional object in the currently selected two-dimensional tomogram that meets the threshold requirements, j further takes a value that is 1 greater than the current value; when j is equal to A, if there is still no two-dimensional object in the (i-A)-th two-dimensional tomogram that meets the threshold requirements, the error tomogram limiting module limits the return control unit to return the program to the preliminary judgment unit; the two-dimensional object repetition control unit controls the preliminary judgment unit, the connected region determination unit, and the return control unit to perform connectivity analysis on each two-dimensional object in the i-th two-dimensional tomogram; the two-dimensional tomogram repetition control unit controls the preliminary judgment unit, the connected region determination unit, and the return control unit to perform connectivity analysis on the two-dimensional objects in each two-dimensional tomogram;

[0163] The number modification unit modifies the number of the two-dimensional object M ik to the same number as the two-dimensional object with the highest comprehensive score determined by the connected region determination unit. Or, when the preliminary judgment unit never finds a two-dimensional object that meets the threshold requirements, the number modification unit modifies the number of the two-dimensional object M ik to a number that is 1 greater than the maximum number of the connected region that has completed connectivity analysis;

[0164] The three-dimensional object feature acquisition unit combines the two-dimensional objects with connectivity to form a three-dimensional object and acquires the features of each three-dimensional object;

[0165] The three-dimensional segmentation result unit defines the start position and end position of each three-dimensional object's tomogram as Z1 and Z2 of the bounding box; the union of the minimum rectangular boxes of each tomogram is taken, and the resulting result is defined as [X1, Y1, X2, Y2]; the bounding box of the final three-dimensional object is [X1, Y1, Z1, X2, Y2, Z2]; where X1, Y1, Z1, X2, Y2, Z2 are the x, y, z coordinates of the starting point and ending point of the minimum circumscribed cube of the three-dimensional object respectively;

[0166] The container judgment unit in the explosive recognition module matches the bounding box of each three-dimensional object with the bounding box of the container recognized by the artificial intelligence to determine whether the three-dimensional object is wrapped by a container; the explosive matching unit determines that the three-dimensional object enters the liquid and powdery explosive recognition library or the solid explosive recognition library for matching according to the judgment result of the container judgment unit, and the explosive judgment unit judges whether the three-dimensional object is an explosive;

[0167] According to the judgment result of the explosive judgment unit, the result output module outputs the corresponding result.

[0168] In Embodiment 2, on a 2.4GHz computer, three-dimensional object segmentation is performed on a three-dimensional volume data of 512*512*343. The completion time of the three-dimensional object segmentation is 5s, and 11 types of three-dimensional objects are segmented.

[0169] Comparative Example 1

[0170] The same scan as in Embodiment 2 is scanned, and the three-dimensional objects in the three-dimensional volume data of the scan are segmented by using the three-dimensional region growing method in the prior art. On a 2.4GHz computer, for a three-dimensional volume data of 512*512*343, the region growing takes about 50s to complete.

[0171] Comparative Example 2

[0172] The same scan as in Embodiment 2 is scanned, and the three-dimensional objects in the three-dimensional volume data of the scan are segmented by using the threshold method in the prior art. On a 2.4GHz computer, for a three-dimensional volume data of 512*512*343, the threshold method can only segment 6 types of three-dimensional objects.

[0173] Comparative Example 3

[0174] The same scan as in Embodiment 2 is scanned, and the three-dimensional objects in the three-dimensional volume data of the scan are segmented by using the boundary method in the prior art. On a 2.4GHz computer, for a three-dimensional volume data of 512*512*343, the boundary method takes 45s.

[0175] It can be seen from the result comparison between Embodiment 2 and Comparative Examples 1 and 3 that, compared with the existing three-dimensional region growing method and boundary method for three-dimensional object segmentation in the three-dimensional volume data by using the three-dimensional object segmentation module of the present invention, the present invention can complete the three-dimensional object segmentation more quickly; it can be seen from the result comparison between Embodiment 2 and Comparative Example 2 that, compared with the existing threshold method, the present invention has higher accuracy in segmenting three-dimensional objects.

[0176] Those skilled in the art can understand that all or part of the processes of implementing the methods of the above embodiments can be completed by instructing relevant hardware through a computer program, and the program can be stored in a computer-readable storage medium. Among them, the computer-readable storage medium is a disk, an optical disc, a read-only memory or a random access memory, etc.

[0177] As described above, the above are only the preferred specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by those skilled in the art within the technical scope disclosed by the present invention should be covered within the protection scope of the present invention.

Claims

1. An explosive recognition system, characterized in that, The explosive identification system includes: An artificial intelligence identification container module, which is used to identify the bounding box of the container; A three-dimensional object segmentation module, which is used to segment the three-dimensional objects in the three-dimensional volume data and obtain the bounding box of each three-dimensional object; An explosive identification module, which includes a container judgment unit, an explosive matching unit, a liquid and powdery explosive identification library, and a solid explosive identification library. The container judgment unit is used to match the bounding box of each three-dimensional object with the bounding box of the container identified by the artificial intelligence identification container module to determine whether the three-dimensional object is wrapped by a container; the explosive matching unit determines that the three-dimensional object enters the liquid and powdery explosive identification library or the solid explosive identification library for matching according to the judgment result of the container judgment unit; The artificial intelligence identification container module includes a CT detection unit and an artificial intelligence model identification unit. The CT detection unit is used for real-time online detection to obtain three-dimensional orthographic DR images at different angles; the artificial intelligence model identification unit is used to identify the DR images obtained by the CT detection unit and perform intersection operation according to the identification results at different angles to obtain the bounding box of the three-dimensional identification of the container; The artificial intelligence identification container module further includes an artificial intelligence model training unit, which is used to perform artificial intelligence model training before real-time online detection to select an accurate artificial intelligence model; the artificial intelligence model training unit includes a data acquisition unit and a model training unit; The data acquisition unit is used to acquire the density and atomic number data of CT reconstruction, perform orthographic projection at different angles to obtain high- and low-energy projection data, and obtain a colored DR image after color assignment; The model training unit is used to pre-collect the CT reconstruction data of the container, use the DR images obtained by the data acquisition unit as the training set, perform prediction using an artificial intelligence model to obtain the spatial position of the container; calculate the loss function. If the loss function meets the requirements, the artificial intelligence model in the current state can be used for container identification; if the loss function does not meet the requirements, the artificial intelligence model is corrected and re-identified until the conditions are met.

2. The explosive identification system according to claim 1, characterized in that, The three-dimensional object segmentation module includes a two-dimensional tomography acquisition unit, a two-dimensional object segmentation unit, a connectivity judgment unit, and a three-dimensional segmentation result unit; The two-dimensional tomography acquisition unit is used to reconstruct the projection data at different angles of the scanned part obtained by dual-energy CT scanning to obtain a plurality of two-dimensional tomographies and number the two-dimensional tomographies in sequence; The two-dimensional object segmentation unit is used to segment the two-dimensional objects in each two-dimensional tomography, obtain the features of each two-dimensional object in each two-dimensional tomography, and number the two-dimensional objects in each two-dimensional tomography in sequence; The connectivity judgment unit is used to analyze the three-dimensional body connectivity to obtain a connectivity result; The three-dimensional segmentation result unit is used to define the starting position and ending position of each slice of the three-dimensional object as Z1 and Z2 of the bounding box; the union of the minimum rectangular boxes of each slice is taken, and the resulting result is defined as [X1, Y1, X2, Y2]; the bounding box of the final three-dimensional object is [X1, Y1, Z1, X2, Y2, Z2]; where X1, Y1, Z1, X2, Y2, and Z2 are the x, y, and z coordinates of the starting point and ending point of the minimum circumscribed cube of the three-dimensional object, respectively.

3. The explosive identification system according to claim 2, wherein The three-dimensional object segmentation module further includes an image preprocessing unit, which is used to preprocess the two-dimensional slice images obtained by the two-dimensional slice acquisition unit to improve the signal-to-noise ratio of the images.

4. The explosive identification system according to claim 2, characterized in that The connectivity judgment unit includes a preliminary judgment unit, a connected region determination unit, a return control unit, and a repetition control unit; The preliminary judgment unit is used to use the kth two-dimensional object M in the ith two-dimensional slice ik as a basis to traverse each two-dimensional object in the (i - 1)th two-dimensional slice, and judge whether the differences in the central position, area, average density, and average atomic number between each two-dimensional object in the (i - 1)th two-dimensional slice and the two-dimensional object M ik meet the threshold requirements; the connected region determination unit is used to, when the preliminary judgment unit determines that there is a two-dimensional object meeting the threshold requirements in the (i - 1)th two-dimensional slice, perform a comprehensive scoring on the two-dimensional object meeting the threshold requirements according to the differences in central position, area, average density, and average atomic number. The smaller the difference, the higher the score. The two-dimensional object with the highest comprehensive score is determined as the connected region in the (i - 1)th two-dimensional slice that has connectivity with the two-dimensional object M ik ; the return control unit is used to, when the preliminary judgment unit determines that there is no two-dimensional object meeting the threshold requirements in the (i - 1)th two-dimensional slice, return the program to the preliminary judgment unit, so that the preliminary judgment unit traverses each two-dimensional object in the (i - j)th two-dimensional slice and repeats the preliminary judgment; where i≥2, k≥1, i≥j≥2; j takes values from small to large. Only when there is no two-dimensional object meeting the threshold requirements in the currently selected two-dimensional slice, j will further take a value 1 greater than the current value; the repetition control unit is used to control the preliminary judgment unit, the connected region determination unit, and the return control unit to repeat until the connectivity analysis of each two-dimensional object in each two-dimensional slice is completed.

5. The explosive identification system according to claim 1, wherein The explosive identification module further includes an explosive judgment unit, and the explosive judgment unit is used to judge whether the three-dimensional object is an explosive.

6. The explosive identification system according to claim 1, wherein The explosive identification system further includes a result output module, and the result output module outputs corresponding results according to the judgment results of the explosive judgment unit.

Images