CN104165896A - Liquid goods safety inspection method and device - Google Patents
Liquid goods safety inspection method and device Download PDFInfo
- Publication number
- CN104165896A CN104165896A CN201410406756.6A CN201410406756A CN104165896A CN 104165896 A CN104165896 A CN 104165896A CN 201410406756 A CN201410406756 A CN 201410406756A CN 104165896 A CN104165896 A CN 104165896A
- Authority
- CN
- China
- Prior art keywords
- container
- liquid
- section
- article
- pallet
- Prior art date
- Legal status (The legal status 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 status listed.)
- Granted
Links
Landscapes
- Analysing Materials By The Use Of Radiation (AREA)
Abstract
The invention discloses a liquid goods safety inspection method and device. Liquid goods placed in a special inspection tray are subjected to double-visual-angle X-ray imaging, the contents of a vessel projection image is analyzed to reconstruct the shape of the cross section of a vessel with double-visual-angle X-ray projection information only, and a classifier method is utilized to carry out automatic inspection on the risk of the detected liquid goods in a high-efficiency manner. The liquid goods safety inspection device comprises two visual angles, each visual angle is composed of an X-ray source and a detector, and each X-ray source and each detector form one detection unit. The method and the device have the advantages of high inspection precision, high imaging quality, high inspection speed and high performance-cost ratio, and can be used for inspecting the safety of a plurality of liquid goods one time.
Description
Technical field
The present invention relates to a kind of method and apparatus of liquid article safety inspection, belong to liquid article safety inspection technical field.
Background technology
At present, in existing liquid article safety inspection technology, based on the lossless detection method of X ray transmission imaging technology, possesses check result accuracy with it high simultaneously, low to container material quality susceptibility, the features such as simple operation, be subject to the attention of more and more safety check manufacturer, this wherein, more representational is that publication number is two sections of patents of invention of CN101140247A and CN101629916A, their common ground is all based on CT layer scanning technology, its method can be sketched and be: send X ray transmission by radiographic source and be examined liquid article, utilize detector to accept the beam transmitted through liquid article, and form hundreds of multi-angle projection data, by these hundreds of multi-angle projection data are carried out to inversion calculation, calculate the gamma ray absorption coefficient that is examined liquid article, wherein, the patent of invention of CN101629916A is owing to having utilized dual-energy x-ray, more can obtain the density and the material information that are examined liquid article simultaneously, finally, by liquid article absorption coefficient or liquid article density, material information and default database are compared, complete the inspection to tested liquid article.The liquid article safety detection method of this class based on CT layer scanning technology, its sharpest edges are to check that precision is high, because it has obtained hundreds of multi-angle projection data in checking process, the various backprojection reconstruction technology that recycling is relatively ripe, can obtain desirable tomography cross-section data.
Although above-mentioned CT layer scanning technology detection performance is high, also exist significantly not enough: 1) checking efficiency is on the low side, once conventionally can only check a conventional volume size liquid article, and a checking process is consuming time longer; 2) check that object coverage rate is restricted, except being not suitable for detecting the liquid of super large superelevation, and can only check specially that liquid article can not compatiblely check parcel; These 2 deficiencies are all by limiting the application in the larger occasion of flow of the people of CT type liquid article safety detection method and equipment, as places such as airport, railway station, big assemblies.
Therefore, for common liquid article safety inspection, be necessary to study and develop a kind of method and apparatus of liquid article safety inspection, can ensure higher inspection precision, can realize quick inspection to multiple liquid articles again, can also compatiblely carry parcel to passenger and check.
Summary of the invention
The object of the present invention is to provide a kind of method and apparatus of the liquid article safety inspection that can overcome above-mentioned technical matters, the present invention is by carrying out Double-visual angle x-ray imaging to the liquid article being positioned in special inspection pallet, can not damage in former wrapped situation, only utilize the X ray projection information at 2 visual angles, once multiple liquid articles in pallet are checked fast, realize efficiently the dangerous automatic detection that is examined liquid article.The order that the method for liquid article safety inspection of the present invention is not blocked liquid mutually by front and back is placed in pallet, and in detection process, the alternately appearance of parcel and pallet, the interior liquid article of pallet and the alternately appearance of on-liquid article, multiple liquid article several situations from beginning to end all can not affect detectivity; The stereoscopic image of method of the present invention based on liquid article, the attribute of analyzing container, carries out container section reconstruction, and then obtains the various features information of liquid in container, judges that accordingly whether liquid article is dangerous.
The method of liquid article safety inspection of the present invention can be distinguished liquid article and on-liquid article in parcel and pallet, pallet, cut apart multiple liquid articles from beginning to end, then selects the detection slice position of suitable liquid container; In each section, first the attribute of analyzing container, carry out the reconstruction of Double-visual angle container section based on container attribute again, the various features such as material, density of liquid is tried to achieve in the cross section that utilizes reconstruction to obtain, and judges that by for example support vector machine of suitable sorter whether liquid is dangerous; Method of the present invention can be applicable to the various channel type X-ray safety inspection equipments that shine visual angle that are no less than 2 visual angles and have side.
The method of liquid article safety inspection of the present invention only utilizes Double-visual angle radioscopic image to carry out container section reconstruction, and first analyzing container attribute comprises the information such as the fundamental type of container material quality, container section; Again nonlinear least square method is combined with method of conjugate gradient, realize the reconstruction of the container section shape under Double-visual angle condition; The method that described container section is rebuild can be applicable to the various channel type X-ray safety inspection equipments that shine visual angle that are no less than 2 visual angles and have side.
It is parcel or pallet that the method for liquid article safety inspection of the present invention is distinguished according to radioscopic image content, and the method for described difference parcel and pallet can be applicable to comprise single-view, various visual angles equipment at interior various conventional channels formula X-ray safety inspection equipment.
The method of liquid article safety inspection of the present invention is liquid article or on-liquid article according to each article in the shape in article region in pallet, gray scale, material information difference pallet, and the method for described difference liquid article and on-liquid article can be applicable to comprise single-view, various visual angles equipment at interior various conventional channels formula X-ray safety inspection equipment.
The method of liquid article safety inspection of the present invention separates multiple liquid articles from beginning to end according to the shape in liquid article region, gray scale, material information, finally on each liquid container, surveys slice position and selects to use for subsequent probe; The system of selection of the dividing method of described liquid article from beginning to end and detection slice position can be applicable to comprise single-view, various visual angles equipment at interior various conventional channels formula X-ray safety inspection equipment.
After obtaining described Double-visual angle liquid article image, liquid detecting method of the present invention specifically comprises the following steps:
1) Images Classification: the safety check image collecting is divided into two classes, one class is parcel image, one class is pallet image, wherein in pallet, may put the article that the various needs such as liquid article, waistband, wallet, mobile phone, coat, Ms's handbag carry out independent safety inspection;
2) image segmentation: the present invention allows that on-liquid article and liquid are mixed to be put in same pallet and to carry out safety inspection, but require the imaging that do not overlap of liquid article and other article in any visual angle, the image at 2 visual angles of 1 pallet is divided into multistage, every section of image is inner or be liquid article entirely, be called container image sections, or be non-liquid article entirely, be called on-liquid image sections;
3) detecting location is selected: the multiple liquid articles in pallet in same container image sections are likely from beginning to end, according to features such as changes in material properties, volume, local shapes, to may separate by multiple containers from beginning to end, then in each container, find the slice position of several local material stability of characteristics, for subsequent probe;
4) container attributes estimation: whether the container that first judges liquid is high density container, if high density container, also will judge that the basic configuration of container section is sorted out, if low-density container, also want analyzing container to be projected in local morphological feature, obtain in a word the information of container of various necessity;
5) low-density container section is rebuild: for low-density containers such as plastic bottles, by container section modelling, propose the algorithm of a kind of nonlinear least square method in conjunction with method of conjugate gradient, in iterative estimation cross section, whether each grid belongs to container area, finally obtains cross sectional shape;
6) high density container section is rebuild: for high density containers such as vials, by container section modelling, the algorithm of a kind of nonlinear least square method in conjunction with method of conjugate gradient proposed, in iterative estimation cross section, whether each grid belongs to container, liquid or air, finally obtain cross sectional shape, in this process, utilized step 4) the container attribute information that obtains;
7) decision-making: based on step 5) or 6) the container section shape that obtains, calculate the various features such as material, density of liquid, use support vector machine as sorter, provide the whether dangerous judgement of liquid.
Finally, as long as 1 container has 1 section by step 7) being judged as danger, system is just reported to the police to this danger container.
Method of the present invention allows liquid and on-liquid article to be alternately put in pallet, and does not need deliberately to keep at a distance between liquid, in addition, before rebuilding container section, obtains abundanter container attribute information by X ray projected image.
Liquid article safety inspection device of the present invention adopts brand-new Double-visual angle topological design, two described visual angles are made up of 2 groups of x-ray sources and detector, every group of x-ray source and detector are called a probe unit, wherein, described probe unit be at the bottom of the middle part being installed in transfer passage according to unit of view V1, side according to unit of view V2; At the bottom of described middle part, comprise that according to unit of view V1 the end, middle part is according to x-ray source and the first detector; Described side comprises that according to unit of view V2 side is according to x-ray source and the second detector; The image that is detected article that two described visual angles gather, processes judgement by above-mentioned 7 the liquid detection steps of the present invention; Two visual angles described in liquid article safety inspection device of the present invention are for really orthogonal.
Advantage of the present invention is the cross section by rebuilding liquid article, then calculate the material of liquid, the various features information such as density, except checking the liquid in low-density (as plastics) container, can also check the liquid in high density (as glass) container, there is higher inspection precision and image quality, in addition, be compared to prior art, the present invention can disposablely carry out safety inspection to multiple liquid articles, inspection speed is fast, cost performance is high, be applicable to occasion that flow of the people is larger as airport, railway station, the liquid article security inspection applications in the places such as big assembly.
In addition, the present invention also has the function of parcel safety inspection simultaneously, makes safety inspector observe parcel by two orthogonal angles, and between object, blocks that probability is lower, imaging distortion is less.Realize maximum functions with relatively minimum visual angle in a word.
Brief description of the drawings
Fig. 1 is the process flow diagram of liquid article safety detection method of the present invention;
Fig. 2 is the structural representation of liquid article safety inspection device of the present invention;
Fig. 3 is the schematic top plan view of the plastic pallet of load bearing fluid of the present invention, and wherein liquid is put groove at pallet middle part;
Fig. 4 is the front elevational schematic of transfer passage and x-ray source position in liquid article safety inspection device of the present invention, and wherein ABCD represents the external polygon of checked property body section;
Fig. 5 is that side is shone in the V2 of visual angle, the latter half schematic diagram of the drop shadow curve of the section of bulge (Fig. 5 (a)) and square container (Fig. 5 (b));
Fig. 6 is the schematic diagram of container section original shape, i.e. grey color part, and wherein ABCD is the external polygon in object cross section in Fig. 4;
Fig. 7 is the schematic diagram of multiple containers image from beginning to end;
Fig. 8 is high density container section schematic diagram;
Fig. 9 is that the side of low-density container section is according to drop shadow curve's schematic diagram of visual angle V2.
Embodiment
Describe the present invention below in conjunction with drawings and Examples.
Fig. 1 has shown the process flow diagram of a kind of liquid article safety detection method of the present invention; First, liquid lies against in the lining groove in the plastic pallet of certain size, before and after each article, puts in order, carries out X-ray scanning; Then, described device, according to the pallet image at two obtained visual angles, is analyzed liquid article wherein, realizes liquid article safety inspection.Survey flow process and mainly comprise following module: Images Classification module 1, image segmentation module 2, detecting location select module 3, container attributes estimation module 4, low-density container section to rebuild module 5, high density container section reconstruction module 6, decision-making module 7, finally liquid result of detection is returned to the control program of safety check system, as find easily to fire the situations such as dangerous material, program is reported to the police.
Fig. 2 is the structural representation of liquid article safety inspection device of the present invention, wherein, 11, 12 is the plumbous door curtain of channel outlet, 13 for side is according to the detector set of visual angle V2, 14 for side is according to the collimating apparatus of visual angle V2, 15 for side is according to the radiographic source of visual angle V2, 16 is the detector set according to visual angle V1 at the bottom of middle part, 17 is the passage of safety inspection device, 18, 19 is the plumbous door curtain of feeder connection, 20 is plastic pallet, 21 is channel exit optical inductor, 22 is the radiographic source according to visual angle V1 at the bottom of middle part, 23 is the collimating apparatus according to visual angle V1 at the bottom of middle part, 24 is the optical inductor at feeder connection place, 25 is conveyor, 26 is liquid article, 27 is system control and signal processing circuit unit, 28 is overall treatment computing machine.Liquid article 26 will lie in plastic pallet 20 32 li of grooves as shown in Figure 3 in accordance with regulations.
When the plastic pallet 20 of carrying liquid article 26 enters from the entrance on passage the right, irradiate successively the scanning of line source 22, V2 side irradiation line source 15 through the V1 end, middle part, system control and signal processing circuit unit 27 obtain respectively the image at two visual angles, and issue computing machine 28 and carry out analyzing and processing.Wherein, please refer to Fig. 3 about the form of plastic pallet, there is 1 road lining groove in pallet central authorities for putting liquid article.
For flow process demonstrated in Figure 1, introduce one by one function and the implementation method of the each module in Fig. 1 below;
In Images Classification module 1, the safety check image collecting need to be divided into two classes, a class is parcel image, a class is pallet image.In pallet, may put the article that the various needs such as liquid article, waistband, wallet, mobile phone, coat, Ms's handbag carry out independent safety inspection.And requirement, liquid article and other on-liquid article can overlapping imagings in two visual angles, and Security Inspection Equipments of the present invention does not allow on-liquid article placed side by side in pallet along throughput direction with liquid article in the time of application.
In the time that the current checked property of difference is a parcel or pallet, first to determine according to picture material the size of checked property.By rim detection and Hough conversion, find four limits of checked property.Because the size of pallet is known, according to the size of above-mentioned four limit boundary information deducibility checked properties, thereby can identify not pallet of parcel greatly.Then, can determine whether pallet according to the feature of following several respects for remaining image: 1) on the border, front and back at the four Shang HuoV2 visual angles, limit at V1 visual angle, if there is the pixel of more metal material, for wrapping up but not plastic pallet; 2) ensemble average gray scale, conventionally in parcel, the average gray of each pixel is lower, and average gray is low to a certain extent, must be parcel; 3) no matter be in V1 visual angle or V2 visual angle, there is a gray threshold T
traycorresponding empty pallet background, if current visual angle image exists the gray scale of a certain amount of pixel higher than T in four limits
tray, think pallet.In fact, if there is liquid in pallet, because the laying for goods rule described in shape and the preceding paragraph of liquid is limit, it can not all block near pallet background position, always have some pallet background pixels to be exposed, this is just easy to be judged as pallet.Just in case do not have liquid in pallet and be covered with by various foreign material, if be mistaken for parcel, the detectivity of equipment is used as parcel and carries out follow-up explosive detection and also do not damage by this pallet; Otherwise, be almost extremely individually sky and the big or small parcel that is similar to pallet, be likely mistaken for pallet, this problem will be illustrated (referring to the description to image segmentation module 2) in subsequent step.
In image segmentation module 2, each row of pallet image are analyzed as unit of analysis, judge image is from which row to which row, and object is therebetween liquid, and this section of image is called a container image sections; And which which row of image be listed as to, object is on-liquid (personal belongings) therebetween, and this section of image is called an on-liquid image sections; Wherein, each row pixel of image, the value of X ray covering of the fan each detector in the time of single pass of corresponding visual angle V1 or V2.List in each of image, cut apart taking pallet background as threshold value: from then on the 1st pixel of row advances backward, when running into the pixel of gray scale lower than pallet background, think the border that runs into object in pallet; From then on last 1 pixel of row is pushed ahead, when running into the pixel of gray scale lower than pallet background, think the border that runs into the object other end; Have object after the border of these row of image, can carry out background gray scale and reject processing.Then, use the object pixel between this border, two ends, according to following feature identify when prostatitis whether be a row pixel of liquid article:
1) the position whether spacing on border, two ends (useful for V1 visual angle especially), and these pixels puts groove near liquid.
2) average gray; The average gray of between border, two ends and center 1/3 part, these two kinds of average gray are too high to be thought and not to be liquid.
3) P (liquid|G of each pixel p between border, two ends
p, M
p) mean value, wherein P (liquid|G
p, M
p) represent when the gray scale of p be G
pand material value is M
ptime, some p belongs to the probability of liquid article; Probability tables P (liquid|G
p, M
p) coming from the statistics to liquid article image in advance, material value comes from dual intensity half-tone information.
4) mean value/variance of gray scale partial gradient; Be included in and on column direction and line direction, ask respectively gradient.
5) between border, two ends, run into the continuous high gray-scale pixels that approaches pallet background gray scale.
6), between border, two ends, run into continuously the pixel (this class pixel color is mazarine in safety check field) of the corresponding metal of material value.
Above-mentioned 6 row that all meet fluid characteristics, and continuous columns is abundant, think that these row represent liquid article shared row in image, by continuous liquid row and the blank column being mingled with, be considered as a container image sections, so-called blank column is the row that could not find any object pixel; Then,, by continuous on-liquid row and the blank column being mingled with, be considered as an on-liquid image sections.For on-liquid image sections, carry out common parcel explosive detection; For container image sections, survey flow process by carrying out follow-up liquid.
In Images Classification module 1, there are a kind of special circumstances, be almost extremely individually the parcel of sky and the approximate pallet of size, likely be mistaken for pallet, the attribute of this parcel and pallet are approximate in fact, and itself or entirety form an on-liquid image sections, thereby carry out common parcel explosive detection; Or in parcel, by chance there is liquid and can separate liquid image section, carrying out follow-up liquid and survey flow process.
Select in module 3 at detecting location, can, separately according to the liquid article image at V1 or V2 visual angle, by separated each liquid container in a container image sections, form independent one by one container, for subsequent analysis.Earlier figures as segmentation module 2 in, each row of image have all found the up-and-down boundary of object.If multiple liquid articles are from beginning to end in a pallet, will form one by being much listed as continuously the large liquid regions that form.Need to be separated each container in this large region, the Main Basis of cutting apart is following several criterion: 1) space; 2) the bottle end of vial; 3) bottleneck; 4) sudden change of material behavior; 5) sudden change of liquid article shape.
Space is that article are cut apart the simplest foundation, as shown in 71 of Fig. 7.Especially, most of container bottom is not flat, and for example plastics Coke bottle, can form certain interval time from beginning to end, and the pixel grey scale in gap and liquid pixel are totally different, easily identification.In addition, the bottle end of vial, has outstanding gray feature, its gray scale is obviously lower, and on material value away from scope (even the main part of vial of fluid organic material, also be taking the material behavior of fluid organic material as main), and form one and approached the short and thick lines perpendicular to pallet working direction, as shown in 73 of Fig. 7; Utilize common image-recognizing method, can judge so short broad-brush existence.
Bottleneck/bottle cap is conventionally than body thin many (bottlenecks 72 as shown in Figure 7), the direction of advancing taking pallet as axle, with image respectively the unit of classifying as analyze.First calculate the width of liquid regions between the up-and-down boundary of every row; If in continuous some row, the width of each row enough narrow and width is the trend smoothly attenuating, and these row can be considered bottleneck/bottle cap position; Especially, in the time there is significantly sudden change in the width of the close row of liquid regions, at the bottom of this generally means and has run into place's bottle.In addition, for side, according to image, the projected position of bottleneck should be higher than the groove in pallet, and this is also a distinct characteristic.By the each region between gap, the bottle end, bottleneck/bottle cap everywhere in container image sections, be first considered as 1 container.
Certainly the above criterion connected situation of divided ownership liquid article completely, for example working as some does not have obvious bottleneck, the similar liquid article of width, from beginning to end and while luckily can seamless unoccupied place docking.A kind of material value for this reason defining in the upper liquid regions of each row i of image is expressed model
the pixel that container area is listed at this is divided into 6 equal portions,
represent the average material value of pixel in the scope of first part herein, by that analogy.Then calculate MD (i)=| M
i-M
i+ Δ i|, represent that close two are listed as the distributional difference of material value between i and i+ Δ i.There is enough positions of large sudden change in MD (i), is considered as running into a new container.Finally, to each internal tank splitting, gather multiple detection sections, they are row of corresponding V1/V2 visual angle image respectively.Require to try one's best equidistantly between these sections, and before and after each section, each MD value being listed as is very little, thereby ensures near the stability of the liquid attribute in location to be detected.
So-called section, is radiogenic ray covering of the fan while container being carried out to single pass perpendicular to throughput direction, the container section that ray covering of the fan passes; The same cross section of container or section can successively be scanned by the radiographic source at two visual angles, and in two visual angle images, produce respectively the gray-scale value of a row pixel, wherein the projection gray level value of the corresponding ray of each pixel; The pixel value of same section in the respective column of two visual angle images, has formed respectively the drop shadow curve at these two visual angles of section.In the present invention by taking the section of container be the drop shadow curve at cross section and corresponding two visual angles thereof as foundation, judge the danger of liquid in container.
Next for each detection section of selecting, carry out container attributes estimation module 4; In container attributes estimation module 4, first need to determine whether high density container, method is to calculate the material value of this container and the R component value of RGB pseudo color image, obtains a container material quality proper vector by the method for adding up, and this vector comprises 4 characteristic variables.The first two characteristic variable is to cut into slices as benchmark taking this detection, belonging to the poor of Calculating material value average and this average and the onesize neighborhood material value of this container zone line average within the scope of container edge neighborhood, latter two characteristic variable is the poor of average in the middle of the R component edge average in kind calculated and this average and container.This eigenmatrix of normalization, makes different characteristic quantized value have comparability.By a large amount of training, obtain priori and can distinguish plastic containers and the high density container differentiation threshold value at this eigenmatrix, this threshold value table is shown an interval, what be greater than this interval upper limit is high density container, what be less than this interval lower limit is plastic containers, and the gray scale geometric properties that will cut into slices by this detection again in interval is further screened container material quality.Concrete grammar is, high density container has gray scale valley point in the drop shadow curve that surveys section on the position near both sides of the edge, and this is because according to projection relation, the path length of relevant ray on high density container outer wall, absorb greatly, gray-scale value is low, and plastic containers do not possess this feature.
After being judged as high density container, and then to judge the type of container section shape; The type of container section, is mainly divided into 4 large types: circular, oval, square, rectangle, notice that this is classification roughly, and for example hexagon or octagon, can be similar to and be considered as circle or ellipse.As shown in Figure 4, the rectangle at figure center represents a cross section of container; From tangent four rays of two radiographic sources and cross section, form the external polygon ABCD in cross section.Can first will utilize the external polygon of container section, by finding this polygonal incircle, judge whether the length breadth ratio of container is approximately 1:1, if be circular or square, otherwise be ellipse or rectangle.And then utilize the latter half of the drop shadow curve at the V2 visual angle of section to carry out pattern-recognition.Fig. 5 is side is surveyed the drop shadow curve of section the latter half schematic diagram according to container in the V2 of visual angle, the corresponding bulge of Fig. 5 (a), and the corresponding square container of Fig. 5 (b), both have notable difference; In drop shadow curve, the left direction of gray scale axle represents gray scale high (X ray damping capacity is few), otherwise right direction represents gray scale low (X ray damping capacity is many).Can utilize suitable mode identification method, the mode of for example sample coupling, distinguishes and more approaches bulge when the drop shadow curve of starting section, or square container; Not obviously the container of 1:1 for length breadth ratio, use equally similar mode identification method, judge elliptical vessel or rectangular tank.
Why will judge the classification of container section, be in order to rebuild in the work of module 6 in ensuing high density container section, to obtain sufficient priori.The space plane at container section place, is divided into a large amount of grids of suitable yardstick.As a rule, rebuild container section from the external polygon (as shown in Figure 4) in cross section, progressively external polygon planted agent is belonged to the grid of air, be designated as air, the final remaining grid that belongs to container forms the shape of container section.Method of the present invention has been introduced new information here---and container section is sorted out.For example, in the time that container section is classified as ellipse, utilize external polygonal four edges, form an oval-shaped formula by data fitting.This oval appropriateness is amplified, and with external polygon stack, the grid that both are common, the original shapes of composition container section, this has greatly reduced the number that need to be designated as the grid of air, has reduced the parameter of follow-up method, has increased the reliability that cross section rebuilds.As shown in Figure 6, suppose that container section has classified as ellipse, the net point of grey color part in figure, is cross section initial configuration, starts to analyze than the external polygon ABCD of direct use, reduce a lot of parameters.If section type is judged as rectangle or circle, also can carry out similar processing.
Next paper high density container section is rebuild module 6, and object is the original shape by above-mentioned cross section, obtains the true form of container section.Introduce a kind of new iterative optimization method in this present invention---based on the optimization method of non-linear least square and FR method of conjugate gradient.The general objective function of optimization method is shown in formula (1):
Wherein M is the number of two all rays in visual angle in current slice, A
ifor the damping capacity (this obtains according to ray projection gray scale) of ray i,
for ray i length through container outer wall part in cross section,
for ray i length through liquid in container part in cross section, μ
wallfor the X ray attenuation coefficient of outer wall, μ
liqfor the X ray attenuation coefficient of liquid.In addition the liquid level S of liquid, container left side wall thickness d
1, container upside wall thickness d
2, container right side wall thickness d
3with container downside wall thickness d
4, be the parameter in Optimized model, though directly do not appear in formula (1), obviously, they take turns the variation in iteration at each, will affect the form of container section.Specifically can be with reference to figure 8, figure Oxford gray part is the outer wall of high density container, and light grey part represents liquid, and 81 represent d
1, 82 represent d
2, 83 represent d
3, 84 represent d
4, 85 represent liquid level, liquid level is S.Why will design 4 wall thickness parameters, be mainly that bottle wall is made as the not diffusive that can reduce net result of uniform thickness everywhere because the wall thickness of glass container is not uniformly in fact.Represent below the round of Optimized Iterative with t.
In the time of t=1, to above-mentioned 7 parameter μ
wall, μ
liq, S, d
1, d
2, d
3and d
4revise, according to non-linear least square method, ask the matrix J shown in formula (2),
Although note S, d
1, d
2, d
3and d
4do not appear in the computing formula of each R (i), but can design these 5 variablees step delta separately, in current cross sectional shape, investigate as S, d
1, d
2, d
3and d
4increase independently Δ and reduce when Δ, the variation of each R (i) value, thus be able to every partial derivative at matrix rear portion in calculating formula (2).Want in addition shown in calculating formula (3)
.
Obviously, the calculating of formula (3) can be carried out according to the content of formula (2), for example
can be by
combination calculates.Then calculate the correction of each parameter
wherein s is 7 dimensional vectors, and its 7 vector elements are representing respectively the correction of 7 parameters in t wheel, and adjusts the identity of the inner each grid in cross section according to the new value of these parameters.
In t+1 wheel, will, according to FR method of conjugate gradient, revise the profile of each container section, the identity of each grid in container section, comprises air, outer wall, liquid.By outermost cross section container grid, and be close to " air " grid of these container grids outside one decks, carry out the correction of identity, establish this two kinds of grid p
1～p
nn altogether.Calculate
similarly, although p
1～p
ndo not appear in the computing formula of F, but suppose each grid p
jthere is identity to timing, for example container=> air or air=> container, F value will change, thereby be able to every partial derivative in calculating formula.For
enough large p
j, it is carried out to identity and exchanges.Once the profile of container section is adjusted, because now will keep S, d
1, d
2, d
3and d
4each parameter constant, the identity of the inner each grid in cross section also needs to occur the correction of a new round thereupon, for example, along with some container grid identity that belongs to outer wall becomes after air, inner some the corresponding identity in this position belongs to the grid of liquid, and its identity will become container outer wall thereupon.
Alternately carry out two kinds of processing that t wheel and t+1 take turns, first correction model parameter, then revise the profile in cross section, until the improvement of objective function F is less than certain threshold value.
Rebuild module 5 for low-density container section, adopt similar principle, but process is simpler, because there is not outer wall factor.Simply utilize the starting point of external polygon as cross sectional shape, then optimize the objective function shown in formula (4).
In the time of t=1, to above-mentioned 1 parameter μ
liqrevise, correspondingly J matrix becomes the capable simple form of 1 row M.In t+1 wheel, still according to FR method of conjugate gradient, revise the profile of container section---by outermost cross section container grid, and the air-grid of these container point outside one decks, carry out the correction of identity, method is similar, repeats no more.Although it is emphasized that and do not need to estimate cross section type, also want analyzing container to be projected in local morphological feature.For instance, if there is liquid level, the part of the corresponding container upside of V2 visual angle drop shadow curve, can present drop shadow effect as shown in Figure 9; Fig. 9 is that the side of container section is according to drop shadow curve's schematic diagram of visual angle V2, in the drop shadow curve of Fig. 9, the left direction of gray scale axle represents gray scale high (X ray damping capacity is few), gray scale axle right direction represents gray scale low (X ray damping capacity is many), corresponding those rays through liquid level of drop shadow curve's section that upper right corner small circle marks, easily identification; Because the ray of the part correspondence of drop shadow curve top side is through liquid level, be positioned at the outer layer container grid p on these rays
j, will be in every wheel
basis on increase larger weight so that p
jthe identity of such grid can be become rapidly air, thereby can form as early as possible the shape of liquid level.
Obtaining after the cross sectional shape of container section, in decision-making module 7, can utilize the existing method of the art to calculate the various features such as the material of liquid, density, volume, container wall thickness, so the section of each container can obtain 1 multidimensional characteristic vectors.In the training stage, fire the container section of liquid for loading, the proper vector obtaining is added into set T
p; For other Safety liquid vessel sections, the proper vector obtaining is added into set T
n.According to pattern recognition theory, can utilize suitable sorter belonging to T
pvector sum belong to T
nvector separate to cog region.The present invention selects support vector machine, distinguishes to load to fire liquid and container safe liquid; Other ripe sorter in area of pattern recognition, such as decision tree or neural network etc. can play the effect of classification equally.
Finally, if according to above-mentioned various features, be judged as dangerous liquid container for there being section to be classified device, by overall treatment computing machine 28 will be according to aforesaid result of detection and container segmentation result as shown in Figure 2, position corresponding on safety inspector's screen draws alert box.
In sum, Images Classification module 1, makes the method and apparatus of liquid article safety inspection of the present invention, can be compatible to the hocket actual demand of detection of parcel, personal belongings and liquid article; Image segmentation module 2, make safety inspector in inspection pallet, place liquid and on-liquid article neatly, the liquid article negligible amounts particularly carrying as passenger or fall when single, can also place other personal belongings in same pallet, thereby improve the efficiency of safety inspection flow process; Detecting location is selected module 3, can cut apart liquid article from beginning to end, makes safety inspector in inspection pallet, place liquid article still less restrictedly, need between liquid article, deliberately not keep spacing; Container attributes estimation module 4, has obtained the abundant priori about container; Low-density container section rebuilds module 5 and high density container section is rebuild module 6, can obtain exactly the cross sectional shape of container in the situation that only having 2 perspective data, and wherein the priori of container plays vital effect; In decision-making module 7 for manifold extraction of container and the use of sorter, make system in the time carrying out decision-making, no longer be confined to contrast intuitively the material of liquid, the value of density, but from large scale training data, excavate abstract classifying rules, obtain the recognition capability of more excellent dangerous liquid.
A kind of liquid article safety inspection device of the present invention is with reference to shown in Fig. 2, and this device shines unit of view V1 at the bottom of comprising overall treatment computing machine 28, system control and signal processing circuit unit 27, conveyor 25, transfer passage 17, side photograph unit of view V2, middle part; Wherein, described side comprises that according to unit of view V2 side is according to x-ray source 15 and the second detector 13; At the bottom of described middle part, comprise that according to unit of view V1 the end, middle part is according to x-ray source 22 and the first detector 16.
As shown in Figure 4, at the bottom of above-mentioned side photograph x-ray source 15, middle part, lay respectively at the different azimuth of transfer passage 17 according to x-ray source 22,, in the positive apparent direction of described transfer passage 17, described side is arranged at the sidepiece of transfer passage 17 according to x-ray source 15, at the bottom of described middle part, be arranged at the middle part below of transfer passage 17 according to x-ray source 22, thereby two angles form an orthogonal Double-visual angle layout pattern below the sidepiece of transfer passage 17, middle part.
Equally, corresponding at the bottom of above-mentioned middle part according to x-ray source 22, side according to x-ray source 15, first detector 16, second detector 13 corresponding with it, be also attached to respectively the diverse location of described transfer passage 17.Described first, second detector, is dual intensity (being high low energy) detector, thereby provides the material properties information that is examined liquid for liquid probe algorithm.
Described the first detector 16 is door type detector, and described the second detector 13 is L-type detector.
The above; be only the specific embodiment of the present invention, but protection scope of the present invention is not limited to this, is anyly familiar with those skilled in the art in scope disclosed by the invention; the variation that can expect easily or replacement, all should be encompassed in the protection domain of the claims in the present invention.
Claims (8)
1. the method for a liquid article safety inspection, it is characterized in that, first described method is distinguished liquid article and on-liquid article in parcel and pallet, pallet, is cut apart multiple liquid articles from beginning to end, then selects the detection slice position of suitable liquid container; In each section, first the attribute of analyzing container, carry out the reconstruction of Double-visual angle container section based on container attribute again, the various features such as material, density of liquid is tried to achieve in the cross section that utilizes reconstruction to obtain, and judges that by for example support vector machine of suitable sorter whether liquid is dangerous; Described liquid article safety detection method can be applicable to the various channel type X-ray safety inspection equipments that shine visual angle that are no less than 2 visual angles and have side.
2. the method for a kind of liquid article safety inspection according to claim 1, is characterized in that, only utilizes Double-visual angle radioscopic image to carry out container section reconstruction, and first analyzing container attribute comprises the information such as the fundamental type of container material quality, container section; Again nonlinear least square method is combined with method of conjugate gradient, realize the reconstruction of the container section shape under Double-visual angle condition; The method that described container section is rebuild can be applicable to the various channel type X-ray safety inspection equipments that shine visual angle that are no less than 2 visual angles and have side.
3. the method for a kind of liquid article safety inspection according to claim 1, it is characterized in that, distinguishing according to radioscopic image content is parcel or pallet, and the method for described difference parcel and pallet can be applicable to comprise single-view, various visual angles equipment at interior various conventional channels formula X-ray safety inspection equipment.
4. the method for a kind of liquid article safety inspection according to claim 1, it is characterized in that, be liquid article or on-liquid article according to each article in the shape in article region in pallet, gray scale, material information difference pallet, the method for described difference liquid article and on-liquid article can be applicable to comprise single-view, various visual angles equipment at interior various conventional channels formula X-ray safety inspection equipment.
5. the method for a kind of liquid article safety inspection according to claim 1, it is characterized in that, according to the shape in liquid article region, gray scale, material information, multiple liquid articles from beginning to end are separated, finally on each liquid container, survey slice position and select to use for subsequent probe; The system of selection of the dividing method of described liquid article from beginning to end and detection slice position can be applicable to comprise single-view, various visual angles equipment at interior various conventional channels formula X-ray safety inspection equipment.
6. the method for a kind of liquid article safety inspection according to claim 1, is characterized in that, after obtaining described Double-visual angle liquid article image, liquid is surveyed and specifically comprised the following steps:
1) Images Classification: the safety check image collecting is divided into two classes, one class is parcel image, one class is pallet image, wherein in pallet, may put the article that the various needs such as liquid article, waistband, wallet, mobile phone, coat, Ms's handbag carry out independent safety inspection;
2) image segmentation: allow that on-liquid article and liquid are mixed to be put in same pallet and to carry out safety inspection, but require the imaging that do not overlap of liquid article and other article in any visual angle, the image at 2 visual angles of 1 pallet is divided into multistage, every section of image is inner or be liquid article entirely, be called container image sections, or be non-liquid article entirely, be called on-liquid image sections;
3) detecting location is selected: the multiple liquid articles in pallet in same container image sections are likely from beginning to end, according to features such as changes in material properties, volume, local shapes, to may separate by multiple containers from beginning to end, then in each container, find the slice position of several local material stability of characteristics, for subsequent probe;
4) container attributes estimation: whether the container that first judges liquid is high density container, if high density container, also will judge that the basic configuration of container section is sorted out, if low-density container, also want analyzing container to be projected in local morphological feature, obtain in a word the information of container of various necessity;
5) low-density container section is rebuild: for low-density containers such as plastic bottles, by container section modelling, propose the algorithm of a kind of nonlinear least square method in conjunction with method of conjugate gradient, in iterative estimation cross section, whether each grid belongs to container area, finally obtains cross sectional shape;
6) high density container section is rebuild: for high density containers such as vials, by container section modelling, the algorithm of a kind of nonlinear least square method in conjunction with method of conjugate gradient proposed, in iterative estimation cross section, whether each grid belongs to container, liquid or air, finally obtain cross sectional shape, in this process, utilized step 4) the container attribute information that obtains;
7) decision-making: based on step 5) or 6) the container section shape that obtains, calculate the various features such as material, density of liquid, use support vector machine as sorter, provide the whether dangerous judgement of liquid;
Finally, as long as 1 container has 1 section by step 7) being judged as danger, system is just reported to the police to this danger container.
7. a liquid article safety inspection device, it is characterized in that, adopt brand-new Double-visual angle topological design, two described visual angles are made up of 2 groups of x-ray sources and detector, every group of x-ray source and detector are called a probe unit, wherein, described probe unit be at the bottom of the middle part being installed in transfer passage according to unit of view V1, side according to unit of view V2; At the bottom of described middle part, comprise that according to unit of view V1 the end, middle part is according to x-ray source and the first detector; Described side comprises that according to unit of view V2 side is according to x-ray source and the second detector.
8. a kind of liquid article safety inspection device according to claim 7, is characterized in that, two described visual angles are for really orthogonal.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410406756.6A CN104165896B (en) | 2014-08-18 | 2014-08-18 | Liquid goods safety inspection method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410406756.6A CN104165896B (en) | 2014-08-18 | 2014-08-18 | Liquid goods safety inspection method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104165896A true CN104165896A (en) | 2014-11-26 |
| CN104165896B CN104165896B (en) | 2017-03-22 |
Family
ID=51909807
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410406756.6A Active CN104165896B (en) | 2014-08-18 | 2014-08-18 | Liquid goods safety inspection method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104165896B (en) |
Cited By (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104459812A (en) * | 2014-12-17 | 2015-03-25 | 同方威视技术股份有限公司 | Pull type multi-view-angle article inspection system and application method thereof |
| CN105631482A (en) * | 2016-03-03 | 2016-06-01 | 中国民航大学 | Convolutional neural network model-based dangerous object image classification method |
| CN105807329A (en) * | 2016-05-30 | 2016-07-27 | 公安部第研究所 | X-ray detection device and method for identifying hazardous liquid in parcels |
| CN106353342A (en) * | 2016-08-12 | 2017-01-25 | 安徽中杰信息科技有限公司 | Method for identifying liquid dangerous goods by X-ray inspection system |
| CN106841247A (en) * | 2017-03-28 | 2017-06-13 | 合肥科的星测控科技有限公司 | A kind of hand-held detection instrument system of contactless liquid dangerous material |
| CN107615336A (en) * | 2015-02-18 | 2018-01-19 | 西门子医疗保健诊断公司 | The location-based detection of tray slot type and test tube type in vision system |
| CN108169254A (en) * | 2016-12-07 | 2018-06-15 | 清华大学 | Check equipment and inspection method |
| WO2018121504A1 (en) * | 2016-12-29 | 2018-07-05 | 同方威视技术股份有限公司 | Security scanning method, apparatus, and system |
| CN109374658A (en) * | 2018-12-24 | 2019-02-22 | 电子科技大学 | The neural network reconstruction tomograph imaging method of double-visual angle X-ray safe examination system |
| WO2019052561A1 (en) * | 2017-09-18 | 2019-03-21 | 同方威视技术股份有限公司 | Check method and check device, and computer-readable medium |
| CN111052129A (en) * | 2017-07-28 | 2020-04-21 | 美国西门子医学诊断股份有限公司 | Deep learning volumetric quantification method and apparatus |
| WO2020082171A1 (en) * | 2018-10-22 | 2020-04-30 | Voti Inc. | Tray insert for screening tray |
| CN111521728A (en) * | 2020-05-15 | 2020-08-11 | 福州大学 | Gas blasting pipeline experimental device and method with multi-dimensional concentration gradient |
| CN111735735A (en) * | 2020-06-11 | 2020-10-02 | 山东滨州烟草有限公司 | Intelligent cigarette package nondestructive identification method and identifier |
| CN113960687A (en) * | 2020-07-03 | 2022-01-21 | 株式会社日立解决方案 | X-ray image processing device and X-ray image processing method |
| CN114313781A (en) * | 2022-01-04 | 2022-04-12 | 北京航星机器制造有限公司 | Luggage tray for safety inspection and image processing method thereof |
| US11885752B2 (en) | 2021-06-30 | 2024-01-30 | Rapiscan Holdings, Inc. | Calibration method and device therefor |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101358936A (en) * | 2007-08-02 | 2009-02-04 | 同方威视技术股份有限公司 | Method and system for discriminating material by double-perspective multi energy transmission image |
| WO2009114928A1 (en) * | 2008-03-17 | 2009-09-24 | Optosecurity, Inc. | Method and apparatus for assessing characteristics of liquids |
| CN101644686A (en) * | 2008-08-08 | 2010-02-10 | 李海洋 | United on-line detector for explosive and poison and application thereof |
| CN102928448A (en) * | 2012-10-30 | 2013-02-13 | 公安部第一研究所 | Channel type four-perspective X-ray liquid goods safety inspection method and inspection device |
| EP2677304A1 (en) * | 2012-06-21 | 2013-12-25 | Entech Scientific B.V. | Method and device for identifying unknown substances in an object |
-
2014
- 2014-08-18 CN CN201410406756.6A patent/CN104165896B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101358936A (en) * | 2007-08-02 | 2009-02-04 | 同方威视技术股份有限公司 | Method and system for discriminating material by double-perspective multi energy transmission image |
| WO2009114928A1 (en) * | 2008-03-17 | 2009-09-24 | Optosecurity, Inc. | Method and apparatus for assessing characteristics of liquids |
| CN101644686A (en) * | 2008-08-08 | 2010-02-10 | 李海洋 | United on-line detector for explosive and poison and application thereof |
| EP2677304A1 (en) * | 2012-06-21 | 2013-12-25 | Entech Scientific B.V. | Method and device for identifying unknown substances in an object |
| CN102928448A (en) * | 2012-10-30 | 2013-02-13 | 公安部第一研究所 | Channel type four-perspective X-ray liquid goods safety inspection method and inspection device |
Non-Patent Citations (1)
| Title |
|---|
| 王宇石 等: "用多视角X射线安检技术探测***物", 《全国危险物质与安全应急技术研讨会议论文集》 * |
Cited By (27)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104459812A (en) * | 2014-12-17 | 2015-03-25 | 同方威视技术股份有限公司 | Pull type multi-view-angle article inspection system and application method thereof |
| US10648929B2 (en) | 2014-12-17 | 2020-05-12 | Nuctech Company Limited | Trailed multiple-viewing-angle item inspection system and method of using the same |
| CN107615336B (en) * | 2015-02-18 | 2021-10-15 | 西门子医疗保健诊断公司 | Location-based detection of tray slot type and cuvette type in a vision system |
| CN107615336A (en) * | 2015-02-18 | 2018-01-19 | 西门子医疗保健诊断公司 | The location-based detection of tray slot type and test tube type in vision system |
| CN105631482A (en) * | 2016-03-03 | 2016-06-01 | 中国民航大学 | Convolutional neural network model-based dangerous object image classification method |
| CN105807329A (en) * | 2016-05-30 | 2016-07-27 | 公安部第研究所 | X-ray detection device and method for identifying hazardous liquid in parcels |
| CN105807329B (en) * | 2016-05-30 | 2019-05-17 | 公安部第一研究所 | It is a kind of for identification package in dangerous liquid x-ray detection device and method |
| CN106353342A (en) * | 2016-08-12 | 2017-01-25 | 安徽中杰信息科技有限公司 | Method for identifying liquid dangerous goods by X-ray inspection system |
| CN108169254A (en) * | 2016-12-07 | 2018-06-15 | 清华大学 | Check equipment and inspection method |
| CN108267463A (en) * | 2016-12-29 | 2018-07-10 | 同方威视技术股份有限公司 | Safety check scan method, device and equipment |
| WO2018121504A1 (en) * | 2016-12-29 | 2018-07-05 | 同方威视技术股份有限公司 | Security scanning method, apparatus, and system |
| CN106841247A (en) * | 2017-03-28 | 2017-06-13 | 合肥科的星测控科技有限公司 | A kind of hand-held detection instrument system of contactless liquid dangerous material |
| CN111052129B (en) * | 2017-07-28 | 2024-03-08 | 美国西门子医学诊断股份有限公司 | Deep learning volume quantification method and apparatus |
| CN111052129A (en) * | 2017-07-28 | 2020-04-21 | 美国西门子医学诊断股份有限公司 | Deep learning volumetric quantification method and apparatus |
| KR20200003011A (en) * | 2017-09-18 | 2020-01-08 | 눅테크 컴퍼니 리미티드 | Inspection methods and inspection equipment and computer readable media |
| KR102240058B1 (en) | 2017-09-18 | 2021-04-15 | 눅테크 컴퍼니 리미티드 | Test methods and test equipment and computer-readable media |
| WO2019052561A1 (en) * | 2017-09-18 | 2019-03-21 | 同方威视技术股份有限公司 | Check method and check device, and computer-readable medium |
| WO2020082171A1 (en) * | 2018-10-22 | 2020-04-30 | Voti Inc. | Tray insert for screening tray |
| US11977037B2 (en) | 2018-10-22 | 2024-05-07 | Rapiscan Holdings, Inc. | Insert for screening tray |
| CN109374658B (en) * | 2018-12-24 | 2022-05-03 | 电子科技大学 | Neural network reconstruction tomography method of double-view X-ray security inspection system |
| CN109374658A (en) * | 2018-12-24 | 2019-02-22 | 电子科技大学 | The neural network reconstruction tomograph imaging method of double-visual angle X-ray safe examination system |
| CN111521728A (en) * | 2020-05-15 | 2020-08-11 | 福州大学 | Gas blasting pipeline experimental device and method with multi-dimensional concentration gradient |
| CN111735735A (en) * | 2020-06-11 | 2020-10-02 | 山东滨州烟草有限公司 | Intelligent cigarette package nondestructive identification method and identifier |
| CN113960687A (en) * | 2020-07-03 | 2022-01-21 | 株式会社日立解决方案 | X-ray image processing device and X-ray image processing method |
| US11885752B2 (en) | 2021-06-30 | 2024-01-30 | Rapiscan Holdings, Inc. | Calibration method and device therefor |
| CN114313781A (en) * | 2022-01-04 | 2022-04-12 | 北京航星机器制造有限公司 | Luggage tray for safety inspection and image processing method thereof |
| CN114313781B (en) * | 2022-01-04 | 2023-11-07 | 北京航星机器制造有限公司 | Luggage tray during safety inspection and image processing method thereof |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104165896B (en) | 2017-03-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104165896A (en) | Liquid goods safety inspection method and device | |
| Zhang et al. | Deep learning–based fully automated pavement crack detection on 3D asphalt surfaces with an improved CrackNet | |
| CN102928448B (en) | Channel-type four visual angle X ray liquid article safety detection method and testing fixture | |
| US20180195977A1 (en) | Inspection devices and methods for detecting a firearm in a luggage | |
| CN103900503B (en) | Extract the method for shape facility, safety detection method and equipment | |
| CN101592579B (en) | Method and device for automatically detecting explosive substances in luggage by using multi-view X rays | |
| US10509979B2 (en) | Inspection methods and systems | |
| EP3123208B1 (en) | Detection of items in a container | |
| EP3699579B1 (en) | Inspection method and inspection device and computer-readable medium | |
| CN102460067A (en) | Systems and methods for automated, rapid detection of high-atomic-number materials | |
| US10186050B2 (en) | Method and system for detection of contraband narcotics in human digestive tract | |
| CN105844602A (en) | Airborne LIDAR point cloud 3D filtering method based on volume elements | |
| DE202007019256U1 (en) | Device for assessing the properties of liquids | |
| CN101387610B (en) | Safety inspection method utilizing double source double energy straight-line type safety inspection CT apparatus | |
| CN102708373A (en) | Method and device for classifying remote images by integrating space information and spectral information | |
| CN105005764A (en) | Multi-direction text detection method of natural scene | |
| CN102901740B (en) | Channel type four visual angles X ray liquid state article safety inspection system | |
| CN103679205A (en) | Preceding car detection method based on shadow hypothesis and layered HOG (histogram of oriented gradient) symmetric characteristic verification | |
| Zhang et al. | On Using XMC R‐CNN Model for Contraband Detection within X‐Ray Baggage Security Images | |
| CN106651841A (en) | Analysis method for security inspection of image complexity | |
| CN111539251B (en) | Security check article identification method and system based on deep learning | |
| CN201514387U (en) | System for utilizing multi-view X ray to automatically detect explosive in luggage | |
| CN101878435A (en) | Systems and methods for reducing false alarms in detection systems | |
| CN202903699U (en) | Channel type four-visual angle X-ray liquid article safety check system | |
| CN106885813A (en) | Check the mthods, systems and devices of goods |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant |