CN117544730A - Portable photographing type scanning equipment and method for distinguishing authenticity of certificate document - Google Patents
Portable photographing type scanning equipment and method for distinguishing authenticity of certificate document Download PDFInfo
- Publication number
- CN117544730A CN117544730A CN202311514826.5A CN202311514826A CN117544730A CN 117544730 A CN117544730 A CN 117544730A CN 202311514826 A CN202311514826 A CN 202311514826A CN 117544730 A CN117544730 A CN 117544730A
- Authority
- CN
- China
- Prior art keywords
- image
- optical signal
- transmitting device
- signal transmitting
- value
- 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.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 49
- 230000003287 optical effect Effects 0.000 claims abstract description 165
- 238000001514 detection method Methods 0.000 claims abstract description 49
- 238000010586 diagram Methods 0.000 claims description 19
- 238000012549 training Methods 0.000 claims description 17
- 230000008569 process Effects 0.000 claims description 11
- 238000005520 cutting process Methods 0.000 claims description 8
- 238000007781 pre-processing Methods 0.000 claims description 7
- 230000009466 transformation Effects 0.000 claims description 5
- PXFBZOLANLWPMH-UHFFFAOYSA-N 16-Epiaffinine Natural products C1C(C2=CC=CC=C2N2)=C2C(=O)CC2C(=CC)CN(C)C1C2CO PXFBZOLANLWPMH-UHFFFAOYSA-N 0.000 claims description 4
- 238000005096 rolling process Methods 0.000 claims description 4
- 238000013135 deep learning Methods 0.000 claims description 3
- 238000010200 validation analysis Methods 0.000 claims description 3
- 238000004364 calculation method Methods 0.000 description 9
- 230000006870 function Effects 0.000 description 7
- 230000005540 biological transmission Effects 0.000 description 5
- 238000013528 artificial neural network Methods 0.000 description 4
- 230000008859 change Effects 0.000 description 4
- 238000005286 illumination Methods 0.000 description 3
- 238000007689 inspection Methods 0.000 description 3
- 238000004519 manufacturing process Methods 0.000 description 3
- 238000001228 spectrum Methods 0.000 description 3
- 230000001133 acceleration Effects 0.000 description 2
- 230000008901 benefit Effects 0.000 description 2
- 238000013527 convolutional neural network Methods 0.000 description 2
- 238000012937 correction Methods 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000002372 labelling Methods 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 230000011218 segmentation Effects 0.000 description 2
- 238000003860 storage Methods 0.000 description 2
- 208000037170 Delayed Emergence from Anesthesia Diseases 0.000 description 1
- RWSOTUBLDIXVET-UHFFFAOYSA-N Dihydrogen sulfide Chemical compound S RWSOTUBLDIXVET-UHFFFAOYSA-N 0.000 description 1
- 241000519995 Stachys sylvatica Species 0.000 description 1
- 239000003086 colorant Substances 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000008878 coupling Effects 0.000 description 1
- 238000010168 coupling process Methods 0.000 description 1
- 238000005859 coupling reaction Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 239000000835 fiber Substances 0.000 description 1
- 238000003702 image correction Methods 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000005549 size reduction Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N1/00—Scanning, transmission or reproduction of documents or the like, e.g. facsimile transmission; Details thereof
- H04N1/04—Scanning arrangements, i.e. arrangements for the displacement of active reading or reproducing elements relative to the original or reproducing medium, or vice versa
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N1/00—Scanning, transmission or reproduction of documents or the like, e.g. facsimile transmission; Details thereof
- H04N1/44—Secrecy systems
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Signal Processing (AREA)
- Inspection Of Paper Currency And Valuable Securities (AREA)
Abstract
The invention provides a portable photographing scanning device and a portable photographing scanning method for distinguishing authenticity of a certificate document, which can collect high-quality digital images and can quickly distinguish authenticity of a collected object. The first optical signal transmitting device and the second optical signal transmitting device are arranged to respectively emit white light and purple light, the irradiation angle of the optical signal transmitting device is controlled, the current is adjusted in the range of the calibrated voltage of the optical signal transmitting device, the brightness of the acquired image is adjusted, the influence of ambient light on the optically variable security feature of the document file is reduced, and the acquired image data is ensured not to influence the normal reading of information on the document file due to the problems of noise points or optically variable security feature and the like; meanwhile, the technical scheme of the application uses the image data of the certificate document and the anti-counterfeiting image data as the input data of the true and false detection model at the same time, judges the true and false of the document, and improves the accuracy of the judgment result.
Description
Technical Field
The invention relates to the technical field of image acquisition and document identification, in particular to portable photographing scanning equipment and a portable photographing scanning method for identifying authenticity of a certificate document.
Background
A photographing scanner, also called a high-speed scanner, is an office equipment which captures an image of a file to be acquired through a CCD/CMOS photosensitive element, converts the image into a digital instrument device which can be displayed, edited and stored by a computer, is generally connected with a computer terminal through a USB/RJ45 interface to complete the transmission of the image, and converts the image into various file formats through matched tool software of the computer terminal.
The high-speed shooting instrument has the advantages of small volume, quick image acquisition and the like, and can digitally acquire documents such as certificates, paperwork, archives and the like. Thus, it is one of the common identity and document information collection office equipment in government authorities, banks, and the like. However, in the long-term document file electronization process, if someone uses forged and altered certificates and document files to transact business in actual work, the traditional high-speed camera does not have the function of distinguishing the authenticity of the certificates and document files, and can only distinguish the authenticity of the certificates and document files by virtue of the experience of window personnel, thereby the lawbreaker can gain illegal benefits. In addition, the working environment of the high-speed image shooting instrument is an open type and open type environment, and the defects of large noise, poor quality and the like of the acquired image can be caused due to large personnel mobility and large change of ambient light, so that the difficulty in distinguishing the authenticity of the subsequent digital document is further improved, and the development of informatization and digital work of the document file is not facilitated. Meanwhile, for some document files added with optically variable security features, the optically variable security features are influenced by multi-factor coupling such as an optical effect visual angle, ambient light, equipment light and the like, and when a high-speed image is acquired, the optically variable security features influence normal reading of information on the document files and are hidden hazards for identification and embedding of subsequent digital documents, and as shown in fig. 11, the information in a white circular frame is shielded by the optically variable security features.
In the prior art, for the problem that the authenticity of certificates, documents and files is difficult to distinguish, the common solution is that a multispectral acquisition instrument, a special document inspection instrument, a license reader and other document inspection equipment are adopted to acquire the anti-counterfeiting characteristic images of the certificates, the documents and the files under different spectrums, and then a professional identifies the authenticity. As a patent with application number of cn20180050037. X, a verification method is provided for easily identifying authenticity by using a printed fine identification mark, and a method is proposed for obtaining an authenticity result by using a fine identification mark printed on a medium as an identification mark, collecting a feature image amplified by a microscope and a manufacturing number, storing the feature image and the manufacturing number in an authenticity data center server, and comparing the feature image and the manufacturing number by a data center in the later stage by using a detail feature image comparison technology. However, the method has the defects of large storage capacity of the server, large investment of personnel and equipment and the like because of more acquired characteristic images, is not suitable for rapidly distinguishing certificates, documents and files in large batches, and does not meet the real-time requirement for identifying true and false during the work of a common window.
Disclosure of Invention
In order to solve the problem that identification of authenticity is difficult in the digital acquisition process of files such as certificate documents, the invention provides portable photographing scanning equipment and a portable photographing scanning method for identifying authenticity of the certificate documents, which can acquire high-quality digital images and can rapidly identify authenticity of an acquired object.
The technical scheme of the invention is as follows: a portable photographic scanning device for discriminating authenticity of a document, comprising: the device comprises a control module, an optical signal transmitting device, a scanning table and a signal receiving device; the optical signal transmitting device and the signal receiving device are electrically connected with the control module;
the signal receiving device is used for receiving the transmitting signal of the optical signal transmitting device to obtain a signal value and transmitting the signal value to the control module;
the method is characterized in that:
the optical signal transmitting apparatus includes: a first optical signal transmitting device and a second optical signal transmitting device;
the first optical signal transmitting device is arranged right above the scanning table;
when the position of the second optical signal transmitting device is below the scanning plane of the scanning table, the signal transmitting angle of the second optical signal transmitting device and the included angle range alpha of the surface of the object to be acquired 2 Between 30 and 90 degrees; otherwise, the signal emission angle of the second optical signal emission device forms an included angle alpha with the surface of the object to be collected 2 Between 15 and 60 degrees;
the first light signal emitting device emits white light;
the second optical signal emitting device emits ultraviolet light;
the control module controls the optical signal transmitting device and the signal receiving device to acquire data of an object to be acquired on the scanning table, and sends the data to a preset pre-acquisition image judging model, and the pre-acquisition image judging model calibrates and adjusts the output value of the optical signal transmitting device according to the signal value acquired by the signal receiving device so that the acquired image signal meets the quality requirement of true and false judgment.
It is further characterized by:
the first optical signal transmitting device is arranged on a normal line passing through the geometric center of the scanning plane of the scanning table, and the included angle between the side edge of a triangular cone formed by the optical signals of the first optical signal transmitting device and the scanning plane of the scanning table is between 60 and 90 degrees;
it also includes: an optical signal reflection housing, the structure of the optical signal reflection housing comprising: a full-wrap reflector and a top reflector;
in the structure of the full-package type reflecting cover, other parts of the scanning equipment are fully covered in the inner cavity of the optical signal reflecting cover;
in the structure of the top reflecting cover, the optical signal transmitting device arranged right above the scanning table is covered in the inner cavity of the optical signal reflecting cover;
when the optical signal reflecting cover is a top reflecting cover, the structure is a conical structure with a narrow upper part and a wide lower part, the bottom surface of the conical structure is parallel to the scanning plane of the scanning platform, and the included angle between the side edge and the bottom surface of the conical structure is between 30 and 75 degrees;
the color temperature of the optical signal of the first optical signal transmitting device is 3000-7000K, and the vertical illuminance from the acquisition surface is 10-200 Lux; the second optical signal emitting device emits ultraviolet light with the wave band of 250 nm-400 nm, the half wave width of 10-30 nm and the intensity of 100 mu W/cm 2 ～2000μW/cm 2 ；
It also includes: and the authenticity detection model is used for carrying out authenticity judgment on the acquired file based on the sent image signal and outputting a judgment result.
A method for discriminating authenticity of a document, comprising the steps of:
s1: the object to be collected is horizontally placed on a scanning plane of a scanning table;
s2: turning on each of the signal generating devices in the optical signal emitting device, respectively, according to the following steps:
a1: acquiring a pre-acquisition image;
a2: detecting an image data brightness value based on the pre-acquired image as a pre-acquired image brightness value;
a3: calculating the signal value calibration of the optical signal transmitting device based on a preset brightness standard value and a pre-acquired image brightness value;
the calculating method for signal value calibration comprises the following steps:
b1: calculating a brightness difference value according to the brightness standard value and the pre-acquired image brightness value;
b2: calculating a current adjustment value I according to the brightness difference value Adjustment of ；
I Adjustment of ＝I+ε×ΔL/L Label (C) ×I
Wherein I is a current value corresponding to the pre-acquired image, L Label (C) As a brightness standard value, delta L is a brightness difference value, and epsilon is a constant parameter;
a4: based on the signal value calibration of the optical signal transmitting device, the image acquisition is carried out on the object to be acquired, and the image acquisition is recorded as: comparing the images;
S3: and respectively marking the contrast images corresponding to the first optical signal transmitting device and the second optical signal transmitting device as: white light contrast image and purple light contrast image;
s4: respectively carrying out image preprocessing on the white light contrast image and the purple light contrast image to divide an image area to be identified, which comprises key content;
s5: and sending all the image areas to be identified into a preset true and false detection model to obtain a true and false judgment result corresponding to the acquired object.
It is further characterized by:
the second optical signal transmitting device is used for transmitting the pre-acquired image corresponding to the second optical signal transmitting device, and detecting the brightness value of the blue channel of the image to serve as the brightness value of the pre-acquired image corresponding to the brightness value;
in step S4, the image preprocessing process includes the following steps:
c1: constructing a certificate image key point detection model based on a coding and decoding network model Unet;
the certificate image key point detection model comprises the following steps: an encoding network and a decoding network;
the coding network comprises a convolution layer and four lightweight residual modules which are sequentially connected, the output of the four lightweight residual modules is 1/4, 1/8, 1/16 and 1/16 of an input image respectively from the input direction to the output direction, the coding network downsamples the characteristics of the input image layer by layer, and the final output characteristic size is changed into 1/16 of the original input image;
The lightweight residual error module comprises: a depth separable rolling layer and an attention module are sequentially arranged, and a 1X 1 convolution layer is respectively arranged in front of the depth separable rolling layer and the attention module; the characteristic value input to the lightweight residual error module is sequentially processed by the depth separable convolution and the attention module, and the obtained attention module output characteristic and the input characteristic value are spliced and then output;
the decoding network includes: the two transposed convolutions DeconvS2 are arranged in sequence, the output of the two transposed convolutions is 1/8 and 1/4 of the input image respectively, one convolution layer is arranged between the two transposed convolutions DeconvS2, and two continuous convolutions layers are arranged behind the latter transposed convolutions; the decoding network uses the deconvolution to up-sample the output characteristics of the encoding network, and finally outputs a prediction thermodynamic diagram after recovering to 1/4 of the original input size;
c2: collecting historical data of a certificate image, marking four edge corner points of the certificate image by using a Lableme tool to obtain a data set, dividing the data set into two groups, and respectively marking the two groups as: a training data set and a validation data set;
c3: training the certificate image key point detection model based on the training data set to obtain a trained certificate image key point detection model;
c4: inputting the image to be identified into the trained certificate image key point detection model, wherein the four key points output correspond to the four corner points of the certificate area of the image to be identified;
c5: carrying out affine transformation on the image to be identified according to the detected key points of the image to be identified, correcting the image to be identified to a standard posture and a standard size, and then cutting to obtain an image of a certificate area;
in step S5, the authenticity detection model is constructed based on a deep learning network model, and is input as an image area to be identified, and output as an authenticity judgment result and a confidence level.
The portable photographing scanning equipment and the portable photographing scanning method for distinguishing the authenticity of the document are characterized in that the first optical signal transmitting device and the second optical signal transmitting device are arranged to respectively emit white light and purple light, the irradiation angle of the optical signal transmitting device is controlled, the current size is adjusted within the calibrated voltage range of the optical signal transmitting device, the brightness of an acquired image is adjusted, the influence of ambient light on the optically variable security feature on a document file is reduced, the normal reading of information on the document file is not influenced by the problems of noise points or optically variable anti-counterfeiting feature and the like on the acquired image data, and the accuracy of the subsequent authenticity distinguishing is further ensured; meanwhile, the technical scheme of the application uses the image data of the certificate document and the anti-counterfeiting image data as the input data of the true and false detection model at the same time, judges the true and false of the document, and improves the accuracy of the judgment result; the acquired image is processed by the image self-adaptive clipping method based on the key point detection, a certificate image key point detection model is constructed by utilizing a lightweight neural network structure, the acquired image is subjected to image correction, and then the image of a certificate area comprising key point information is accurately and rapidly clipped, so that the speed and accuracy of judging the authenticity of the file are improved; the method has low requirements on hardware deployment environment, can rapidly complete the detection of the key points of the certificate image without depending on a graph acceleration chip such as GPU, NPU, TPU and the like, completes the identification of the authenticity of the certificate document, is suitable for the embedded equipment deployment environment with low power consumption and low calculation power, is convenient for developing a portable photographing scanner, solves the problem of insufficient generalization capability of an algorithm, provides convenience for a service window, improves the service handling efficiency, and has important value and significance in practical application.
Drawings
FIG. 1 is a schematic diagram of a portable photographing scanning apparatus for discriminating authenticity of a document in the present application, which is a structural example 1;
FIG. 2 is a schematic diagram of an optical path of an optical signal emitting device;
FIG. 3 is a schematic flow chart of a method for distinguishing authenticity of a document of a certificate in the present application;
FIG. 4 is a schematic diagram of an image preprocessing process;
FIG. 5 is a configuration example 2 of a portable photographing scanning device for discriminating authenticity of a document;
FIG. 6 is a schematic diagram of a portable photographic scanning apparatus for discriminating authenticity of a document, according to example 3
FIG. 7 is a configuration example 4 of a portable photographing scanning device for discriminating authenticity of a document;
FIG. 8 is a schematic diagram of a network structure of an image keypoint detection model;
FIG. 9 is a schematic diagram of a lightweight residual block;
FIG. 10 is an example of an image region to be identified
FIG. 11 is an illustration of an optically variable security feature occlusion;
fig. 12 is a schematic view of a position setting of an optical signal emitting device on a top plate in the example of fig. 1.
Detailed Description
The present invention includes a portable photographic scanning device for discriminating authenticity of a document, comprising: a control module 24, an optical signal transmitting device, a scanning table 13 and a signal receiving device 25; the optical signal transmitting means and the signal receiving means 25 are electrically connected to the control module 24. The signal receiving device 25 is configured to receive the transmission signal of the optical signal transmitting device, obtain a signal value, and transmit the signal value to the control module 24.
In order to the outward appearance and the form of the high appearance of change tradition that reduce, reduce the influence of environment light to the security feature that becomes of light on the certificate to according to gathering the image, reach the function of distinguishing certificate archives true and false fast, the optical signal transmitting device in this application includes: a first optical signal emitting device 21 and a second optical signal emitting device 22. In order to ensure that the optical signal reflected by the scanning table 13 can be received, the signal receiving device 25 is arranged above the scanning table 13 in the present application, and in fact, the signal collected by the signal receiving device 25 is the reflection signal of the optical signal transmitting device at the scanning table 13 and the reflection signal of the reflection cover. If there is no reflective cover or a non-totally enclosed reflective cover, the signal collected by the signal receiving device 25 also includes reflection of natural light at the scanning stage 13.
The first light signal emitting device 21 emits white light; the first optical signal transmitting device 21 is used for collecting information such as pictures and characters of the collected object.
The second optical signal emitting device 22 emits ultraviolet light; the second optical signal transmitting device 22 is mainly used for collecting the anti-counterfeiting information on the collected object. The existing anti-counterfeiting images on some certificates are usually made of ultraviolet-sensitive materials such as fluorescent fibers, fluorescent seals and the like, and can be detected only by ultraviolet irradiation.
Some certificates also comprise optical holographic anti-counterfeiting images, the optical holographic anti-counterfeiting is provided with a stripe grating structure, different patterns or marks can be displayed under the reflection of light rays with different angles, and in order to avoid the influence of the anti-counterfeiting images on the information identification of the certificate images, when the optical holographic anti-counterfeiting features are manufactured, the anti-counterfeiting images cannot appear at the angle perpendicular to the certificate under white light, so that the first optical signal transmitting device 21 is arranged right above the scanning table 13 in the application, and the collected image data under the white light cannot have the anti-counterfeiting images; meanwhile, according to different characteristics of optical holographic anti-counterfeiting on the file to be acquired, the second optical signal transmitting device 22 for transmitting ultraviolet light is arranged at other angular positions, so that the influence of optical holographic anti-counterfeiting can be avoided under ultraviolet light, and an ultraviolet light anti-counterfeiting image can be accurately acquired.
In the technical scheme, the image data and the anti-counterfeiting image data of the certificate file are clearly read and are sent into the true and false detection model to be detected as input data, so that the accuracy of the true and false detection result is improved.
The first optical signal emitter 21 is arranged on the normal line of the geometric center of the scanning plane of the overscan table 13, and the side of the triangular cone formed by the optical signals of the first optical signal emitter 21 forms an included angle alpha with the scanning plane of the overscan table 13 1 Is in the range of 60 DEG to 90 deg.
The color temperature of the optical signal of the first optical signal transmitting device 21 is 3000-7000K, and the vertical illumination from the acquisition surface is 10-200 Lux; the second optical signal emitting device 22 emits ultraviolet light with a wavelength of 250nm to 400nm, a half-wave width of 10 to 30nm and an intensity of 100 mu W/cm 2 ～2000μW/cm 2 。
The second optical signal emitting device 22 is arranged at different positions according to the requirements of the actual detected certificate and file.
As shown in FIG. 2, when the position of the second optical signal emitting device 22 is below the scanning plane of the scanning table 13, the signal emitting angle thereof is within the range alpha of the surface angle of the object to be collected 2 Between 30 and 90 degrees. When the second optical signal emitting device 22 is positioned above the scanning stage 13 or the second optical signal emitting device 22 is positioned above the scanning stage 13, the signal emitting angle of the second optical signal emitting device 22 forms an angle alpha with the surface of the object to be collected 2 Between 15 and 60 degrees.
The method also comprises a pre-acquisition image judging model and an authenticity detecting model, wherein the control module 24 controls the optical signal transmitting device and the signal receiving device 25 to acquire data of an object to be acquired on the scanning table 13, and sends the data into the pre-acquisition image judging model, and in the pre-acquisition image judging model, the output value of the optical signal transmitting device is calibrated and adjusted according to the signal value acquired by the signal receiving device 25, so that the acquired image signal meets the quality requirement of authenticity judgment. The authenticity detection model performs authenticity judgment on the acquired file based on the sent image signal, and outputs a judgment result.
When the method is concretely implemented, the pre-acquisition image judgment model and the true and false detection model can be arranged in a remote data service center, the scanning equipment is connected with the data service center through internal network encryption, acquired data are sent to the data center in real time, the judgment result fed back by the data center is received, and when the scanning equipment is modified based on the existing portable photographing equipment, the method can be used. The trained model can be directly carried into the control module, off-grid work is completely realized, and the working efficiency is further improved.
In order to improve the acquisition quality of the image on the scanning table, the optical signal reflecting cover 23 is further provided in the application, and the structure of the optical signal reflecting cover 23 comprises: full-wrap reflectors and top reflectors.
In the structure of the full-package type reflection cover, other parts of the scanning device are arranged to be covered in the inner cavity of the optical signal reflection cover 23.
In the structure of the top reflection cover, an optical signal emitting device disposed directly above the scanning table 13 is covered in the inner cavity of the optical signal reflection cover 23.
When the optical signal reflecting cover 23 is a top reflecting cover, the structure is a conical structure with narrow top and wide bottom, the bottom surface of the conical structure is parallel to the scanning plane of the scanning platform, and the included angle alpha between the side edge and the bottom surface of the conical structure 3 The range is between 30 DEG and 75 deg.
In the photographing scanner of the embodiment shown in fig. 1, a top plate 12 is installed above a scanning table 13 based on a support rod 11, and a first optical signal transmitting device 21 and a second optical signal transmitting device 22 are installed at a lower end surface of the top plate 12. The signal receiving means 25 is placed on top.
The first optical signal emitting device 21 emits white lightThe light color temperature is 5000K, and the vertical illuminance of the light source from the bottom plate 13 is 100Lux; α1 is 50++5°. The ultraviolet light wave band of the second optical signal transmitting device 22 is 365nm, the half-wave width is 30nm, and the ultraviolet light intensity from the bottom plate 13 is 800 mu W/cm 2 The method comprises the steps of carrying out a first treatment on the surface of the α2 is 45°±5°; a signal reflecting cover 23 is provided on top of the scanning table 13. Inclination angle α of optical signal reflection cover 23 3 35 deg. + -5 deg..
In particular, the signal receiving device 25 is realized based on an existing camera. The first optical signal emitting device 21 and the second optical signal emitting device 22 are based on the light emitting module of white light and ultraviolet light in the prior art, such as: the first optical signal emitting means 21 and the second optical signal emitting means 22 are implemented based on a dot matrix LED, a plate-like LED or a tube-like LED as in the embodiment shown in fig. 12. The camera of the optical signal reflecting cover 23 is placed right above the scanning table 13, the lattice LEDs of the first optical signal emitting device 21 are uniformly distributed on the outer periphery of the optical signal reflecting cover 23, and the second optical signal emitting device 22 is uniformly distributed on the outer periphery of the first optical signal emitting device 21. Ensures that the upper light can uniformly irradiate on the scanning table 13 at the bottom, thereby improving the scanning effect.
In order to meet the requirement of acquiring better images in complex environments, a fully-closed device can be adopted. In the embodiment shown in fig. 5, the first optical signal emitting device 21 is placed on top of the scanning table 13; the position of the second optical signal emitting means 22 is below the scanning plane of the scanning stage 13; the signal receiving means 25 is placed on top. The optical signal reflecting cover 23 is of a full-closed type, a sliding door structure is arranged at the same time, a door shaft is arranged at the bottom, a door is opened to place an object to be collected on the scanning table 13, the door is closed during scanning, and the door is opened to take out the object to be collected after scanning; the scanning process is ensured to be completely free from the influence of external light.
Another embodiment of the totally enclosed device referring to fig. 7, a transparent slot is formed at the bottom of the totally enclosed optical signal reflecting cover 23, and after the object to be collected is placed on the scanning table 13, the scanning table 13 and the object to be collected are pushed into the inner cavity of the totally enclosed optical signal reflecting cover 23 together based on the bottom transparent slot, so as to complete the scanning operation. After scanning, the scanning table 13 and the object to be collected are pulled out together from the transparent groove at the bottom of the totally-enclosed optical signal reflecting cover 23. When the scanning table 13 is pushed into the inner cavity of the totally-enclosed optical signal reflecting cover 23, the bottom transparent groove is completely blocked, the arrangement ensures that the scanning process is completely not influenced by external light, a door spindle structure is not required to be arranged independently, and the complexity of the whole structure is reduced.
In the embodiment shown in fig. 5, the first light signal emitting device emits white light with a color temperature of 3000K, and the vertical illuminance of the light source is 80Lux; the included angle alpha 1 between the signal emission angle and the surface of the object to be collected is 30 degrees plus or minus 5 degrees;
a second signal device with ultraviolet light wave band of 280nm, half wave width of 30nm and ultraviolet light intensity of 500 mu W/cm 2 The method comprises the steps of carrying out a first treatment on the surface of the The included angle alpha 2 between the signal emission angle and the surface of the object to be collected is 35 degrees plus or minus 5 degrees.
The tilt angle α3 of the totally enclosed optical signal reflecting cover 23 is 90 ° ± 5 °.
In order to reduce the application cost, the prior high-speed shooting instrument is not greatly changed, and a reflecting cover can be omitted by applying the method. In the embodiment shown in fig. 6, the first optical signal emitting device 21, the second optical signal emitting device 22 and the signal receiving device 25 are all disposed above the scanning table 13 through the top support rod 12, and the control module 24 is integrated on the original control chip below the scanning table 13.
The embodiment shown in fig. 7 is still based on the prior art high-speed imaging apparatus, in which the signal receiving device 25 is disposed above the scanning table 13, the second optical signal transmitting device 22 is disposed above the scanning table 13, and is disposed on the inner side wall of the enclosed optical signal reflection housing 23.
In the embodiment shown in fig. 7, the first optical signal transmitting means 21 is not provided, but only the second optical signal transmitting means 22 is provided. In some cases, in order to improve the detection speed, the anti-counterfeiting image under the purple light can be collected only at the window, and the image under the white light can be used for using historical data, or a conventional file image under the white light is not needed, and the authenticity of the file can be judged only through the anti-counterfeiting image under the purple light. The specific setting is adaptively selected according to actual conditions.
When the portable photographing scanning device is specifically applied, the portable photographing scanning device for identifying authenticity of the certificate document further comprises a data transmission module, is realized based on a limited or wireless network module or a USB transmission module and a serial port module, and the control module 24 sends all received scanned image data and authenticity judgment results to the server for archiving based on the data transmission module.
The method for distinguishing the authenticity of the certificate document based on the scanning equipment during photographing comprises the following steps.
S1: the object to be acquired is laid flat on the scanning plane of the scanning table 13. In practical use, the scanning table 13 will be horizontally arranged, and the collected objects including various certificates, documents and files are horizontally placed on the scanning plane at the top end of the scanning table 13.
S2: turning on each of the signal generating devices in the optical signal emitting device is performed according to the following steps, respectively.
a1: transmitting an optical signal according to preset voltage and current values, and acquiring a pre-acquired image;
a2: based on the pre-acquired image, an image data luminance value is detected as a pre-acquired image luminance value.
The specific image brightness value calculation method is realized based on the calculation method in the prior art.
a3: and calculating the signal value calibration of the optical signal transmitting device based on the preset brightness standard value and the pre-acquired image brightness value.
The calculating method for signal value calibration comprises the following steps:
b1: calculating a brightness difference value according to the brightness standard value and the pre-acquired image brightness value;
b2: calculating a current adjustment value I according to the brightness difference value Adjustment of ；
I Adjustment of ＝I+ε×ΔL/L Label (C) ×I
Wherein I is a current value corresponding to the pre-acquired image, L Label (C) As a standard luminance value, Δl is a luminance difference value, and epsilon is a constant parameter. The second optical signal emitting device 22 detects the brightness value of the blue channel of the image as the brightness value of the corresponding pre-acquired image.
a4: based on the signal value calibration of the optical signal transmitting device, the image acquisition is carried out on the object to be acquired, and the image acquisition is recorded as: and (5) comparing the images.
The constant parameter epsilon is set according to the characteristics of different document files under different types of light rays, and the values of the constant parameter epsilon are different because the base colors and the materials of different files are different and the reverse setting light ray intensities under different types of light rays are also different. When the method is specifically applied, the constant parameter epsilon corresponding to different types of files can be obtained by calculation according to the historical data. In this embodiment, epsilon under white light is set for the driver's license 1 Epsilon at 365nm UV light at 0.08 2 0.05.
Regarding the luminance standard value L Label (C) Is set according to the type of light rays and whether the optical signal reflecting cover 23 is closed, and the brightness standard value L corresponding to each device can be obtained through experiments in specific application Label (C) . For the embodiment shown in fig. 1, a luminance standard value L of white light is set Label W ＝20(cd/m 2 ) Brightness value L of blue channel of standard ultraviolet light image Label UV ＝16(cd/m 2 )。
Image acquisition is performed as based on the embodiment shown in fig. 1. The first light signal emitting device emits white light with a color temperature of 5000K, and the vertical illuminance of the light source from the bottom plate 13 is 100Lux. The ultraviolet light wave band of the second signal device is 365nm, the half wave width is 30nm, and the ultraviolet light intensity from the bottom plate 13 is 800 mu W/cm 2 The method comprises the steps of carrying out a first treatment on the surface of the A first signal transmitting device, wherein the signal transmitting angle forms an included angle alpha with the surface of the object to be acquired 1 At 50 ° ± 5 °; a second signal transmitting device, wherein the signal transmitting angle forms an included angle alpha with the surface of the object to be acquired 2 At 45 ° ± 5 °; inclination angle alpha of optical signal reflecting cover 23 3 35 ° ± 5 °; the signal receiving apparatus 25 is a CMOS image sensor used, and has a resolution of 1280×960 pixels (120 ten thousand pixels) for capturing image information. The aperture value F was 2.8.
The initial signal values of the first optical signal transmitting means 21 and the second optical signal transmitting means 22 are preset in the control module 24. In the present embodiment, the two rated voltages are 3v, and the initial current value I of the first optical signal transmitting device 21 1 =350 mA, second optical communicationInitial current value I of number transmitting device 22 2 ＝300mA。
The specific implementation process is as shown in fig. 3, a document file to be acquired is put in, a pre-acquired image under the white light condition is acquired, and a document image 1 to be identified is obtained;
according to the image brightness value L of the document image 1 to be identified 1 ＝15(cd/m 2 ) And the brightness value L of the standard white light image calibrated before Label W ＝20(cd/m 2 ) Calculating DeltaL 1 ＝L Label (C) -L 1 ＝5(cd/m 2 )。
The current adjustment value is I 1 tone ＝I 1 +ε 1 ×ΔL 1 /L Label W ×I 1 ＝350+0.08×5/20×350＝357mA。
Thus, the current value I of the first optical signal device is changed by the control device 1 tone =357 mA and voltage value V 1 The output of =3v to the first optical signal emitting device, a contrast image under white light condition was obtained, denoted as: the document image 2 to be identified.
The control assembly turns off the first optical signal device and turns on the second optical signal emitting device 22; according to preset parameters, the second optical signal transmitting device is turned on to provide a voltage value V for the second optical signal transmitting device 2 =3v and current value I 2 =300 mA. Acquiring a pre-acquired image under 365nm light: a document image 3 to be identified.
Brightness value L of blue channel of image according to document image 3 to be identified 3 12, luminance value L of blue channel of standard uv image calibrated before Label UV =16, calculate Δl 2 ＝L Label UV -L 3 ＝4。
The current adjustment value is I 2 tone ＝I 2 +ε 2 ×ΔL 2 /L Label UV ×I 2 ＝300+0.05×4/16×300＝303.75mA。
Changing the current value I of the second optical signal transmitting means 22 2 tone =303.75ma and voltage value V 2 The output of =3v to the second optical signal emitting device 22, and a contrast image under the condition of purple light bar is obtained: a document image 4 to be identified.
S3: the contrast images corresponding to the first optical signal transmitting device 21 and the second optical signal transmitting device 22 are respectively denoted as: white contrast image and violet contrast image.
In this embodiment, the document image 2 to be identified is a white light contrast image, and the document image 4 to be identified is a purple light contrast image.
S4: and respectively carrying out image preprocessing on the white light contrast image and the purple light contrast image, and dividing an image area to be identified, which comprises key contents.
The image preprocessing process comprises the following steps:
c1: and constructing an image key point detection model based on the coding and decoding network model Unet.
As shown in fig. 8, the document image keypoint detection model includes: an encoding network and a decoding network;
the coding network comprises a convolution layer and four lightweight residual modules which are sequentially connected, the output of the four lightweight residual modules is 1/4, 1/8, 1/16 and 1/16 of an input image respectively from the input direction to the output direction, the coding network downsamples the characteristics of the input image layer by layer, and the final output characteristic size is changed into 1/16 of the original input image.
The coding network in the application is constructed by using lightweight Bottleneck with an attention mechanism, the feature input size is uniformly scaled to 256 (width) times 192 (height) times 3 (channel), and the coding network finally outputs 16 times 12 times 512 features; the decoding network uses transposed convolution to achieve feature size reduction and finally outputs 64 x 48 x 1 features.
As shown in fig. 9, the lightweight residual module includes: a depth separable convolution and an attention module which are arranged in sequence, wherein a convolution layer of 1 multiplied by 1 is respectively arranged in front of the depth separable convolution and the attention module; the characteristic value input into the lightweight residual error module is sequentially processed by the depth separable convolution and the attention module, and the obtained attention module output characteristic and the input characteristic value are spliced and then output;
the decoding network includes: the two transposed convolutions DeconvS2 are arranged in sequence, the output of the two transposed convolutions is 1/8 and 1/4 of the input image respectively, one convolution layer is arranged between the two transposed convolutions DeconvS2, and two continuous convolutions layers are arranged behind the latter transposed convolutions; the decoding network uses the deconvolution to up-sample the output characteristics of the encoding network, and finally outputs a prediction thermodynamic diagram after recovering to 1/4 of the original input size; the size of the output image is set to be 1/4 of the size of the initial input image, the output image with the size is also enough to support identification of the characteristic points of the certificate area, the output image is not required to be amplified to be the same size as the input image, the calculation speed is improved, the structural complexity of a decoding network is reduced, the identification precision is ensured, and meanwhile, the model training difficulty is also greatly reduced.
The certificate image key point detection model is a lightweight coding and decoding network designed for the certificate size, and the characteristic semantic information under different scales is effectively cascaded through structural design in the coding network and the decoding network, so that the key point detection of the high-precision certificate image is realized, the recognition of certificate areas under different angles and different light rays can be adapted, and further, the accurate cutting task is realized;
the coding and decoding network uses depth separable convolution to replace traditional convolution, the coding network downsamples the original features layer by layer, and the final feature size is changed into 1/16 of the original input; the decoding network uses the deconvolution to upsample the output characteristics of the encoding network, ultimately returning to 1/4 of the original input size. Through the structural design of the depth coding and decoding network, the characteristic semantic information of noise robustness under different scales can be extracted, so that the final key point detection result has stronger anti-interference performance on illumination, angle change and the like.
c2: collecting historical data of a certificate image, marking four edge corner points of the certificate image by using a Lableme tool to obtain a data set, dividing the data set into two groups, and respectively marking the two groups as: a training data set and a validation data set; the labeling sequence is carried out according to the upper left corner, the upper right corner, the lower right corner and the lower left corner of the certificate, and finally the COCO JSON format labeling file is exported.
c3: and training the certificate image key point detection model based on the training data set to obtain a trained certificate image key point detection model.
After the certificate image key point detection model outputs the prediction thermodynamic diagram, marking each key point of the input image as a characteristic point, and respectively constructing a corresponding true value thermodynamic diagram to form an input-output matching diagram similar to a semantic segmentation task;
the truth thermodynamic diagram H (x, y) is constructed in the following manner:
wherein the subscript k denotes the kth feature point, x k And y k And (5) representing coordinates corresponding to the kth feature point, wherein sigma is the standard deviation of the Gaussian kernel.
In the training process, an average square error loss function L is used for measuring the difference between the output predicted thermodynamic diagram and the true thermodynamic diagram;
the average square error loss function L has the following calculation formula:
wherein H is k (x, y) represents a predicted thermodynamic diagram of the model predicted,representing a true value thermodynamic diagram actually corresponding to the training data;
the average square error loss function L is convenient to derive during calculation, the calculation efficiency is improved, the back propagation speed is high, and the training efficiency is greatly improved.
c4: and inputting the image to be identified into a trained certificate image key point detection model, wherein the four outputted key points correspond to four corner points of a certificate area of the image to be identified.
As shown in fig. 4, after the input image is processed by the certificate image key point detection model, the truth heat maps corresponding to the four corner points output by the network are combined, and a graph with the four corner points is obtained. Four white spots on the black matrix, namely four corner points, can obtain the certificate area.
c5: and carrying out affine transformation on the image to be identified according to the detected key points of the image to be identified, correcting the image to be identified to the standard posture and standard size, and then cutting to obtain the image of the certificate area.
In this embodiment, the standard gesture corresponds to a document long edge at a horizontal angle, and the standard dimension corresponds to a width 1920 (width) x1280 (height) pixels, at which dimension clear identification of document image features can be ensured. Based on the existing image affine transformation technology, angle adjustment is carried out on an 'input image' based on four corner points according to a standard gesture, the 'input image' is subjected to size adjustment according to a ratio by comparing the sizes among the four corner points with a standard size, an input image corrected to the standard gesture and the standard size is obtained, then the regions outside the corner points are cut according to the positions of the four corner points, and rectangular regions among the obtained corner points are the certificate image obtained after the final perspective transformation correction cutting in fig. 4. Then, according to different types of objects to be detected, cutting the image area to be identified according to a preset cutting template. As in the example of the card-type document in fig. 10, the area of the clipping image in the corresponding preset template is the area marked in the white box, including: head portrait area, personal information area, issuing department stamping area, anti-fake image area, etc., these areas are the image area to be identified.
According to the self-adaptive clipping method for the certificate image based on the key point detection, the key point detection model of the certificate image is constructed based on the lightweight neural network, the characteristic semantic information under different scales is effectively cascaded, the key point detection of the high-precision certificate image is realized, the positioning of the corner points of the certificate area is completed, and the whole positioning process does not need manual parameter adjustment; meanwhile, in the coding network, by setting 1/4, 1/8, 1/16 and 1/16 lightweight residual modules for downsampling and 1/8 and 1/4 transposed convolution in the decoding network for upsampling, the characteristic semantic information of noise robustness under different scales can be extracted, so that a final key point detection result has stronger anti-interference performance on illumination, angle change and the like, four corner points of a certificate image are ensured to be accurately extracted, and further the edge of a cut certificate region is ensured to be more accurate; in the method, the document image key point detection model constructs an input-output matching diagram based on a semantic segmentation task principle, each output characteristic point has accurate coordinate information and classification labels, and four key points corresponding to an image to be identified can be obtained very efficiently based on thermodynamic diagrams of each key point to serve as corner points of the document; the method is based on a lightweight neural network structure to construct a certificate image key point detection model, the traditional convolution is replaced by the depth separable convolution in the coding network in the model, the model size is obviously reduced, the reasoning speed is accelerated on the premise of ensuring the accuracy, the rapid reasoning under a CPU is supported, the requirement on the hardware deployment environment is low, and the certificate image key point detection can be rapidly completed without depending on a GPU, NPU, TPU and other graphic acceleration chips.
S5: and sending all the image areas to be identified into a preset true and false detection model to obtain a true and false judgment result corresponding to the collected object.
The true and false detection model in the application is realized based on a deep learning network model in the prior art, such as a convolutional neural network CNN; the input of the model is the image area to be identified of the certificate file under the white light and the image area to be identified of the certificate file anti-counterfeiting pattern under the purple light, and the output is the true and false judgment result and the confidence level. The specific construction method and the training method are realized based on the prior art. The training data of the true and false detection model is constructed based on historical data, and training is carried out based on a training mode with labels.
The method is based on true and false judgment of certificates, files and the like, and when the input data is the image area to be identified of the certificate file under white light and the image area to be identified of the certificate file anti-counterfeiting pattern under purple light, the feedback false identification confidence can reach 95.3%. On the premise of not collecting white light, the document false-proof pattern is judged by collecting the document false-proof pattern under the purple light, so that the recognition efficiency can be accelerated. However, since the number of judgments is reduced, which results in a reduced accuracy of the judgment result, and once a person with a heart certificate counterfeiter is induced, the anti-counterfeit feature of ultraviolet light is simulated, thereby breaking the anti-counterfeit function of the device, it is suggested that the use of two or more spectra in combination will increase the ability to distinguish the authenticity of the certificate.
After the technical scheme of the invention is used, the control assembly consisting of the first optical signal transmitting device, the second optical signal transmitting device, the optical signal reflecting cover, the signal receiving device and the control module is used for presenting different physical feedback phenomena under different spectrums based on the document file, respectively collecting optical signals sent by the two signal transmitting devices, adjusting the current correction acquisition signal intensity under rated voltage, acquiring the image of the document to be identified, processing the acquired image based on the image self-adaptive cutting method of key point detection, constructing an image key point detection model by utilizing a lightweight neural network, and carrying out logic judgment, thereby achieving the function of rapidly distinguishing the authenticity of the document file. According to the technical scheme, personnel and equipment cost brought by using the special text inspection instrument and the distinguishing instrument can be effectively reduced, the number of image acquisition is reduced, the acquisition quality requirement of the images is reduced, particularly, accurate object color does not need to be acquired, the calculation power requirement and the storage requirement of a server are further reduced, various problems caused by insufficient generalization capability of an optimization algorithm are further reduced, convenience is brought to a service window, the business handling efficiency is improved, and the method has important value and significance in practical application.
Claims (10)
1. A portable photographic scanning device for discriminating authenticity of a document, comprising: the device comprises a control module, an optical signal transmitting device, a scanning table and a signal receiving device; the optical signal transmitting device and the signal receiving device are electrically connected with the control module;
the signal receiving device is used for receiving the transmitting signal of the optical signal transmitting device to obtain a signal value and transmitting the signal value to the control module;
the method is characterized in that:
the optical signal transmitting apparatus includes: a first optical signal transmitting device and a second optical signal transmitting device;
the first optical signal transmitting device is arranged right above the scanning table;
when the position of the second optical signal transmitting device is below the scanning plane of the scanning table, the signal transmitting angle of the second optical signal transmitting device and the included angle range alpha of the surface of the object to be acquired 2 Between 30 and 90 degrees; otherwise, the signal emission angle of the second optical signal emission device forms an included angle alpha with the surface of the object to be collected 2 Between 15 and 60 degrees;
the first light signal emitting device emits white light;
the second optical signal emitting device emits ultraviolet light;
the control module controls the optical signal transmitting device and the signal receiving device to acquire data of an object to be acquired on the scanning table, and sends the data to a preset pre-acquisition image judging model, and the pre-acquisition image judging model calibrates and adjusts the output value of the optical signal transmitting device according to the signal value acquired by the signal receiving device so that the acquired image signal meets the quality requirement of true and false judgment.
2. The portable photographic scanning apparatus for discriminating authenticity of a document according to claim 1, wherein: the first optical signal transmitting device is arranged on a normal line passing through the geometric center of the scanning plane of the scanning table, and the included angle between the side edge of a triangular cone formed by the optical signals of the first optical signal transmitting device and the scanning plane of the scanning table is 60-90 degrees.
3. The portable photographic scanning apparatus for discriminating authenticity of a document according to claim 1, wherein: it also includes: an optical signal reflection housing, the structure of the optical signal reflection housing comprising: a full-wrap reflector and a top reflector;
in the structure of the full-package type reflecting cover, other parts of the scanning equipment are fully covered in the inner cavity of the optical signal reflecting cover;
in the structure of the top reflecting cover, the optical signal transmitting device arranged right above the scanning table is covered in the inner cavity of the optical signal reflecting cover.
4. The portable photographic scanning apparatus for discriminating authenticity of a document according to claim 1, wherein: when the optical signal reflecting cover is a top reflecting cover, the structure is a conical structure with a narrow upper part and a wide lower part, the bottom surface of the conical structure is parallel to the scanning plane of the scanning platform, and the included angle between the side edge of the conical structure and the bottom surface is 30-75 degrees.
5. The portable photographic scanning apparatus for discriminating authenticity of a document according to claim 1, wherein: the color temperature of the optical signal of the first optical signal transmitting device is 3000-7000K, and the vertical illuminance from the acquisition surface is 10-200 Lux; the second optical signal emitting device emits ultraviolet light with wave bands of 250-400 nm, half wave widths of 10-30 nm and intensity of 100 mu W/cm 2 ~2000µW/cm 2 。
6. The portable photographic scanning apparatus for discriminating authenticity of a document according to claim 1, wherein: it also includes: and the authenticity detection model is used for carrying out authenticity judgment on the acquired file based on the sent image signal and outputting a judgment result.
7. A method for discriminating authenticity of a document, comprising the steps of:
s1: the object to be collected is horizontally placed on a scanning plane of a scanning table;
s2: turning on each of the signal generating devices in the optical signal emitting device, respectively, according to the following steps:
a1: acquiring a pre-acquisition image;
a2: detecting an image data brightness value based on the pre-acquired image as a pre-acquired image brightness value;
a3: calculating the signal value calibration of the optical signal transmitting device based on a preset brightness standard value and a pre-acquired image brightness value;
The calculating method for signal value calibration comprises the following steps:
b1: calculating a brightness difference value according to the brightness standard value and the pre-acquired image brightness value;
b2: calculating a current adjustment value I according to the brightness difference value Adjustment of ；
I Adjustment of =I+Ɛ×ΔL/L Label (C) ×I
Wherein I is a current value corresponding to the pre-acquired image, L Label (C) As a brightness standard value, Δl is a brightness difference value, Ɛ is a constant parameter;
a4: based on the signal value calibration of the optical signal transmitting device, the image acquisition is carried out on the object to be acquired, and the image acquisition is recorded as: comparing the images;
s3: and respectively marking the contrast images corresponding to the first optical signal transmitting device and the second optical signal transmitting device as: white light contrast image and purple light contrast image;
s4: respectively carrying out image preprocessing on the white light contrast image and the purple light contrast image to divide an image area to be identified, which comprises key content;
s5: and sending all the image areas to be identified into a preset true and false detection model to obtain a true and false judgment result corresponding to the acquired object.
8. The method for distinguishing between true and false documents according to claim 7, wherein: and the second optical signal transmitting device is used for detecting the brightness value of the blue channel of the image as the brightness value of the corresponding pre-acquired image corresponding to the brightness value of the blue channel of the image.
9. The method for distinguishing between true and false documents according to claim 7, wherein: in step S4, the image preprocessing process includes the following steps:
c1: constructing a certificate image key point detection model based on a coding and decoding network model Unet;
the certificate image key point detection model comprises the following steps: an encoding network and a decoding network;
the coding network comprises a convolution layer and four lightweight residual modules which are sequentially connected, the output of the four lightweight residual modules is 1/4, 1/8, 1/16 and 1/16 of an input image respectively from the input direction to the output direction, the coding network downsamples the characteristics of the input image layer by layer, and the final output characteristic size is changed into 1/16 of the original input image;
the lightweight residual error module comprises: a depth separable rolling layer and an attention module are sequentially arranged, and a 1X 1 convolution layer is respectively arranged in front of the depth separable rolling layer and the attention module; the characteristic value input to the lightweight residual error module is sequentially processed by the depth separable convolution and the attention module, and the obtained attention module output characteristic and the input characteristic value are spliced and then output;
The decoding network includes: the two transposed convolutions DeconvS2 are arranged in sequence, the output of the two transposed convolutions is 1/8 and 1/4 of the input image respectively, one convolution layer is arranged between the two transposed convolutions DeconvS2, and two continuous convolutions layers are arranged behind the latter transposed convolutions; the decoding network uses the deconvolution to up-sample the output characteristics of the encoding network, and finally outputs a prediction thermodynamic diagram after recovering to 1/4 of the original input size;
c2: collecting historical data of a certificate image, marking four edge corner points of the certificate image by using a Lableme tool to obtain a data set, dividing the data set into two groups, and respectively marking the two groups as: a training data set and a validation data set;
c3: training the certificate image key point detection model based on the training data set to obtain a trained certificate image key point detection model;
c4: inputting the image to be identified into the trained certificate image key point detection model, wherein the four key points output correspond to the four corner points of the certificate area of the image to be identified;
c5: and carrying out affine transformation on the image to be identified according to the detected key points of the image to be identified, correcting the image to be identified to the standard posture and standard size, and then cutting to obtain the image of the certificate area.
10. The method for distinguishing between true and false documents according to claim 7, wherein: in step S5, the authenticity detection model is constructed based on a deep learning network model, and is input as an image area to be identified, and output as an authenticity judgment result and a confidence level.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311514826.5A CN117544730A (en) | 2023-11-15 | 2023-11-15 | Portable photographing type scanning equipment and method for distinguishing authenticity of certificate document |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311514826.5A CN117544730A (en) | 2023-11-15 | 2023-11-15 | Portable photographing type scanning equipment and method for distinguishing authenticity of certificate document |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117544730A true CN117544730A (en) | 2024-02-09 |
Family
ID=89787586
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311514826.5A Pending CN117544730A (en) | 2023-11-15 | 2023-11-15 | Portable photographing type scanning equipment and method for distinguishing authenticity of certificate document |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117544730A (en) |
-
2023
- 2023-11-15 CN CN202311514826.5A patent/CN117544730A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| KR102549881B1 (en) | Validation of security documents | |
| US9016577B2 (en) | Machine-readable symbols | |
| AU2008335637B2 (en) | Document verification using dynamic document identification framework | |
| US9483677B2 (en) | Machine-readable symbols | |
| US8194933B2 (en) | Identification and verification of an unknown document according to an eigen image process | |
| EP2102785B1 (en) | Apparatus and method for secure detection of an item and a method of securing access to information associated with the item | |
| US20060017959A1 (en) | Document classification and authentication | |
| CN103477369A (en) | Systems and methods for detecting an optically variable material | |
| US20230267602A1 (en) | Method and system for optical product authentication | |
| MX2008005018A (en) | Drive socket design for hold and drive fastener. | |
| US20070269109A1 (en) | Method and apparatus for processing selected images on image reproduction machines | |
| CN117544730A (en) | Portable photographing type scanning equipment and method for distinguishing authenticity of certificate document | |
| JP7467795B2 (en) | How to authenticate magnetic induction marks with a portable device | |
| JP4649658B2 (en) | Hidden image or hidden information authentication method and authentication system | |
| WO2020251380A1 (en) | Method for validation of authenticity of an image present in an object, object with increased security level and method for preparation thereof, computer equipment, computer program and appropriate reading means | |
| US20230300270A1 (en) | Document reader with improved illumination | |
| JP2023511569A (en) | Methods for authenticating security documents | |
| AU2005201622B2 (en) | Secure recorded documents |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |