Network freight certificate automatic auditing method and system

Abstract

The invention discloses an automatic auditing method and system for a network freight certificate, and relates to the technical field of network freight, and the method comprises the steps: obtaining a certificate image, and carrying out the preprocessing of denoising and contrast enhancement of the certificate image; carrying out binarization processing on the obtained enhanced image, extracting a target boundary, and carrying out perspective correction processing on the obtained edge image to obtain a standard image; extracting anti-counterfeiting features according to the edge image and the standard image, wherein the anti-counterfeiting features comprise texture features, color features, shape features and frequency domain features; and determining an anti-counterfeiting detection score according to the anti-counterfeiting feature, and completing anti-counterfeiting evaluation of the certificate by combining the scores of the image quality and the feature vector quality. The full-process automation of certificate auditing is realized, the forgery recognition rate is improved, the omission ratio of expired certificates is reduced, the online average auditing time is shortened, and the manual intervention rate is reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of network freight, in particular to a network freight certificate automatic auditing method and system. BACKGROUND

[0002] The statements in this section merely provide background information related to the present application and do not necessarily constitute the prior art.

[0003] In recent years, with the deep integration of digital technology and road freight field, the emerging industry of network freight has shown a vigorous development trend, and the network freight platform has become the key support for the majority of truck drivers to undertake transportation business. However, the current network freight platform still has obvious shortcomings in certificate auditing technology, which is embodied in multiple dimensions: in terms of auditing efficiency, manual auditing of single certificate takes more than 2 hours on average, and serious auditing backlog problem occurs during peak period, which greatly affects the business promotion speed; in terms of auditing accuracy, the success rate of manual identification of counterfeit certificates is less than 65%, and the missed detection rate of expired certificates is more than 18%, which brings safety hazards to the platform operation; in terms of data connection, the platform cannot realize real-time connection with the transportation and management system, and can only rely on offline information verification, forming a significant data island; in terms of risk control and dispute handling, the platform lacks effective counterfeit certificate early warning mechanism and electronic evidence for dispute tracing, making it difficult to deal with potential risks; in terms of user experience, the feedback cycle of certificate auditing often exceeds 24 hours, and the auditing process is not transparent, resulting in poor user experience of the driver group.

[0004] In order to improve this situation, some solutions have appeared in the industry, but the limitations are also prominent: among them, the OCR+rule engine solution can only complete the extraction of certificate information, and cannot effectively verify the authenticity of the information; the single data source verification solution is difficult to fully meet the auditing needs due to limited coverage; and the traditional image comparison solution has generally low recognition rate when facing PS processing or high imitation certificates, and cannot effectively solve the problem of counterfeit certificate identification. SUMMARY

[0005] In order to solve the above problems, the present application provides a network freight certificate automatic auditing method and system, which realizes the full-process automation of certificate auditing, improves the identification rate of counterfeit certificates, reduces the missed detection rate of expired certificates, reduces the average online auditing time, and reduces the manual intervention rate.

[0006] In order to achieve the above purpose, the present application adopts the following technical solutions: In a first aspect, the present application provides a network freight certificate automatic auditing method, comprising: obtaining a certificate image and pre-processing it by denoising and contrast enhancement; The obtained enhanced image is binarized, and a target boundary is extracted. The obtained edge image is subjected to perspective correction to obtain a standard image. Anti-fake features including texture features, color features, shape features and frequency domain features are extracted according to the edge image and the standard image. An anti-fake detection score is determined according to the anti-fake features, and an anti-fake evaluation of the certificate is completed in combination with the scores of the image quality and the feature vector quality.

[0007] As an optional implementation, the preprocessing process includes: performing a denoising operation on the brightness channel of the certificate image, performing convolution smoothing by using a two-dimensional Gaussian kernel to obtain a denoised image, and performing contrast enhancement on the denoised image by histogram equalization.

[0008] As an optional implementation, edges are detected on the binary image, an expansion operation is performed on the edge image to repair breaks, a contour is extracted and a candidate quadrilateral region is obtained by polygon approximation, and corner point positions and geometric proportions are verified. The edge image is mapped to a standard size, and if corner point detection fails, a Hough straight line is used to intersect points or a minimum circumscribed rectangle is used as a degenerate scheme.

[0009] As an optional implementation, texture consistency scores, color distribution anomaly scores and shape consistency scores are determined according to the anti-fake features, and the anti-fake detection score is obtained by weighting.

[0010] As an optional implementation, the calculation process of the image quality score is that the image quality score is obtained by weighted calculation of a Laplacian variance, a brightness uniformity, a histogram dynamic range, an HSV distribution skewness and a smooth signal-to-noise ratio of the standard image.

[0011] As an optional implementation, the calculation process of the feature extraction score is that the feature extraction score is obtained by weighted calculation of a LBP variance calculated according to the texture feature vector, a color anomaly score calculated according to the color feature, an edge continuity ratio calculated according to the shape feature and the frequency domain feature.

[0012] In a second aspect, the present application provides an automatic network freight certificate auditing system, comprising: A preprocessing module configured to acquire a certificate image and perform denoising and contrast enhancement preprocessing on the certificate image; A standard processing module configured to binarize the obtained enhanced image, extract a target boundary, and obtain a standard image by performing perspective correction on the obtained edge image; A feature extraction module configured to extract anti-fake features including texture features, color features, shape features and frequency domain features according to the edge image and the standard image; The anti-counterfeiting detection module is configured to determine an anti-counterfeiting detection score according to the anti-counterfeiting features, and complete the anti-counterfeiting evaluation of the certificate in combination with the scores of the image quality and the feature vector quality.

[0013] In a third aspect, the present application provides an electronic device comprising a memory and a processor, and computer instructions stored in the memory and running on the processor, when the computer instructions are run by the processor, the method of the first aspect is completed.

[0014] In a fourth aspect, the present application provides a computer readable storage medium for storing computer instructions, when the computer instructions are executed by the processor, the method of the first aspect is completed.

[0015] In a fifth aspect, the present application provides a computer program product comprising a computer program, when the computer program is executed by the processor, the method of the first aspect is completed.

[0016] Compared with the prior art, the present application has the following beneficial effects: The present application provides a network freight certificate automatic auditing method and system, after the certificate image is preprocessed by denoising and contrast enhancement, the enhanced image is binarized and the target boundary is extracted, the standard image is obtained after the edge image is processed by perspective correction; the anti-counterfeiting features including texture features, color features, shape features and frequency domain features are extracted according to the edge image and the standard image; and the anti-counterfeiting detection score is determined according to the anti-counterfeiting features, and the anti-counterfeiting evaluation of the certificate is completed in combination with the scores of the image quality and the feature vector quality. Through real-time comparison of online AI intelligent identification, AI anti-counterfeiting + multi-source real-time verification + risk engine + block chain storage is adopted to realize the full-process automation of certificate auditing, improve the identification rate of counterfeiting, reduce the missed detection rate of expired certificates, reduce the average online auditing time, and reduce the artificial intervention rate.

[0017] The advantages of the additional aspects of the present application will be partially given in the following description, partially will become obvious from the following description, or will be known by the practice of the present application. BRIEF DESCRIPTION OF DRAWINGS

[0018] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or the prior art description. Obviously, the drawings in the following description only belong to the embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor on the basis of the provided drawings.

[0019] Figure 1 The network freight certificate automatic auditing method flow chart provided for embodiment 1 of the present application. DETAILED DESCRIPTION

[0020] The application will be further described below with reference to the drawings and examples.

[0021] It should be noted that the following detailed description is exemplary in nature and is intended to provide further description of the application. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs.

[0022] It should be noted that the terms used herein are only intended to describe specific embodiments and are not intended to limit the exemplary embodiments according to the present application. As used herein, the singular form is intended to include the plural form unless the context clearly indicates otherwise, and it should also be understood that the terms "comprise" and "include" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device that includes a series of steps or units does not have to be limited to only those steps or units clearly listed, but can include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0023] The embodiments in the present application and the features in the embodiments can be combined with each other without conflict.

[0024] Embodiment 1 The present embodiment provides a network freight certificate automatic auditing method, which aims to realize full-process automation of certificate auditing, improve the identification rate of counterfeiting, improve the real-time nature of data verification, and reduce the rate of manual intervention; and build a four-fold security system: AI anti-counterfeiting detection, multi-source real-time verification, intelligent risk assessment, and blockchain evidence storage.

[0025] As shown in Figure 1 The method comprises the following steps: Obtain the certificate image and perform denoising and contrast enhancement preprocessing on it; Perform binarization processing on the obtained enhanced image, extract the target boundary, and obtain a standard image after perspective correction processing on the obtained edge image; Extract anti-counterfeiting features from the edge image and the standard image, including texture features, color features, shape features, and frequency domain features; Determine the anti-counterfeiting detection score according to the anti-counterfeiting features, and complete the anti-counterfeiting evaluation of the certificate in combination with the scoring of the image quality and the feature vector quality.

[0026] The method of the present embodiment will be described in detail below.

[0027] S1: The user end uploads certificate information.

[0028] The certificate types include: ID card, driver's license, driving license, road transport certificate, and qualification certificate.

[0029] Among them, the certificate information collection mode includes: mobile terminal high-definition shooting (including living body detection), scanning uploading (≤5MB, JPG / PDF format).

[0030] S2: AI anti-counterfeiting processing.

[0031] The following operations are performed on the certificate image by referencing the OpenCV library: S2-1: Denoising the certificate image by Gaussian blur, which is used to suppress high-frequency noise of the certificate image and smooth fine textures, thereby improving the stability of subsequent threshold segmentation and edge detection.

[0032] Specifically: (1) Convert the input original certificate image to YCrCb or Lab space, and only perform denoising operation on the luminance channel.

[0033] (2) Convolution smoothing is performed using a two-dimensional Gaussian kernel G(x, y; σ) to obtain a denoised image I denoise .

[0034] Among them, the kernel size is adaptively adjusted according to the image resolution, and σ usually takes a value of 0.6~1.8, and the odd kernel size ksize=2×ceil(3σ)+1.

[0035] Among them, by calculating the high-frequency energy attenuation ratio R∈[0.4, 0.8], the σ is automatically adjusted when the transition is too smooth or not enough, to realize quality control.

[0036] S2-2: Contrast enhancement is performed by histogram equalization to improve the brightness and contrast of the denoised image and reduce the impact of uneven illumination on subsequent recognition.

[0037] Specifically: (1) According to the entropy H and luminance variance σL of the denoised image I denoise , determine whether to use global equalization or adaptive CLAHE algorithm; where the CLAHE parameter tileGridSize is usually (8×8), and clipLimit takes 2.0~4.0.

[0038] (2) Perform light bilateral filtering on the result to suppress noise amplification and output an enhanced image I enh .

[0039] Among them, by detecting the contrast enhancement ratio C=σL after / σL before , C should be in the range of [1.2, 2.5], and the parameters are automatically adjusted when exceeding the limit.

[0040] S2-3: Binaryzation processing, which is used to separate the foreground from the background to provide clean input for structure extraction and contour fitting.

[0041] Specifically: The enhanced image I enh is automatically selected according to the uniformity of illumination by using an Otsu threshold method or an adaptive threshold method; morphological closing and opening operations are performed to remove noise points and fill in holes, filter connected domains with small areas, retain the main foreground region, and obtain a binary image I bin .

[0042] wherein the foreground proportion p is in the range of [0.05, 0.85], and the threshold or parameter C is automatically adjusted to maintain balance.

[0043] S2-4: Edge detection, used to extract target boundary features for perspective correction and template matching.

[0044] Specifically: (1) For the binary image I bin , the Canny algorithm is used to detect edges, and the threshold is adaptively calculated according to the median of the gradient.

[0045] (2) For the edge image I edge , an expansion operation is performed to repair breaks; the contour is extracted and a candidate quadrilateral region is obtained by polygon approximation.

[0046] wherein the edge density d is in the range of [0.02, 0.15], and the threshold range is automatically adjusted to ensure that the edge structure is clear and complete.

[0047] S2-5: Perspective correction, used to restore the target plane captured at an angle or in perspective to a standard orthographic view to standardize the size and direction.

[0048] Specifically: (1) For the edge image I edge , the largest area quadrilateral is selected from the contour as a candidate region, and the corner point position and geometric proportion are verified.

[0049] (2) The homography matrix H is calculated and perspective transformation is performed to map the edge image to a standard size WstdxHstd.

[0050] (3) If the corner point detection fails, a Hough line intersection point or a minimum circumscribed rectangle is used as a degenerate scheme.

[0051] (4) The final output is the corrected standard image I std .

[0052] wherein the corner point re-projection error ε is less than 2px, and the boundary line residual error is less than 1.5px, and if not satisfied, the calculation is automatically recalculated or degenerated to an affine transformation.

[0053] S3: Extract anti-counterfeiting features based on edge images and standard images, including texture features, color features, shape features, and frequency domain features.

[0054] Specifically: S3-1: Extract texture features; the purpose is to extract the detailed texture, fiber structure and local repeating patterns of the image surface, which can be used to describe the microstructural differences of anti-counterfeiting paper, watermark embossing and other materials.

[0055] Specifically: (1) For standard image I std Local texture encoding is extracted using the Local Binary Pattern (LBP) algorithm, and multi-scale LBP histograms are calculated.

[0056] (2) Calculate the gray co-occurrence matrix (GLCM) and extract statistics such as energy, contrast, homogeneity, and entropy.

[0057] (3) Construct a multi-scale, multi-directional Gabor filter bank, extract the response intensity distribution in each direction, and thereby construct the texture feature vector F. texture .

[0058] Among them, the parameter settings and adaptive strategy are as follows: LBP radius r=1~3, neighborhood points p=8~16; Gabor kernel scale is adjusted with resolution, covering the main frequency band 0.1~0.4π; the filtering intensity is adaptively adjusted according to image noise and variance.

[0059] Quality control and rollback mechanism: If the local variance is too low (texture smoothing), the contrast enhancement map is called and recalculated; if the noise is too strong, the smoothed version is used to re-extract.

[0060] S3-2: Extract color features; the purpose is to extract the color distribution and spectral ratio features of the image, which can be used to identify special inks, anti-counterfeiting pigments and differences in ultraviolet / infrared response.

[0061] Specifically: (1) Standard image I std Convert from BGR space to Lab and HSV spaces to obtain standard color image I std_color .

[0062] (2) For standard color image I std_color Normalized color histograms (32~64 bins) are calculated for each channel, and statistical indicators such as mean, variance, and skewness are extracted.

[0063] (3) Calculate the color ratio characteristics (R / G, R / B, G / B) and the cross-spectral ratio characteristics such as UV / VIS, and construct the color feature vector F. color .

[0064] Among them, the parameters and adaptive strategy are as follows: the CLAHE parameter is adjusted according to the image brightness variance to balance the influence of illumination; if the illumination is uneven, brightness correction is performed before extraction.

[0065] Among them, the quality control and rollback mechanism is as follows: if the brightness non-uniformity is higher than the threshold, the illumination correction is re-executed; if there is no color information, the empty vector is recorded and the module is skipped.

[0066] S3-3: Extract shape features; the purpose is to extract significant boundaries and geometric structures for detecting pattern outlines, watermark shapes, and logo geometric consistency.

[0067] Specifically: For edge image I edge The main contours are extracted using OpenCV's findContours method; a Fourier descriptor transform is then performed on the main contours to extract the low-frequency shape feature vector F. shape .

[0068] Among them, the parameters and adaptive strategy are as follows: the top N (3~5) largest contours are automatically selected for feature calculation based on the target complexity.

[0069] Among them, the quality control and rollback mechanism is as follows: if the number of edge contours is insufficient, rollback to the edge detection stage for recalculation; if the contour is severely broken, perform morphological repair and re-extraction.

[0070] S3-4: Extract frequency domain features; the purpose is to analyze the frequency energy distribution and periodic structure of the image, and identify hidden patterns, anti-counterfeiting grids or stripe features.

[0071] Specifically: For standard image I std Perform a Fast Fourier Transform (FFT) to obtain the amplitude spectrum and perform logarithmic normalization; calculate the energy ratios of low, high, and mid frequencies, as well as the dominant frequency direction and peak intensity; perform a Discrete Wavelet Transform (DWT) on the image and statistically analyze the sub-band frequency domain eigenvectors F. freq .

[0072] Among them, the parameters and adaptive strategy are as follows: the main frequency peak threshold is set to 0.3×max(FFT), the wavelet decomposition layer is 2~3 layers, and the adjustment is made adaptively according to the image size.

[0073] Among them, the quality control and backoff mechanism is as follows: if the signal-to-noise ratio of the main frequency is lower than 3dB, the frequency domain features are marked as low confidence; if the image has strong reflection, brightness normalization is performed and the features are re-extracted.

[0074] S4: Anti-counterfeiting detection score.

[0075] S4-1: Texture features based on multi-scale LBP. The smaller the LBP variance value, the more consistent the texture. The purpose is to evaluate the uniformity and repeatability of local texture in an image, and to detect the consistency of watermarks, paper textures or anti-counterfeiting backgrounds.

[0076] Specifically: (1) Based on the texture feature vector F texture The multi-scale LBP algorithm is used to calculate the binary pattern of the neighborhood of each pixel and construct the histogram distribution. (2) Calculate the LBP variance Var(LBP), which reflects the degree of dispersion of local texture distribution; (3) The variance is mapped to the consistency score using a normalization function: S texture = exp(-k×Var(LBP)), where k is the adjustment coefficient (0.3~0.6); S texture ∈[0,1], the smaller the variance, the more stable the texture and the higher the score.

[0077] Among them, the quality control and rollback mechanism is as follows: if the image noise is high or the gray-level variance is <0.01, smoothing is enabled before extracting LBP; if S texture If the value is less than 0.4 and the local energy is low, it is marked as "texture anomaly".

[0078] S4-2: Detecting color distribution anomalies based on HSV spatial distribution color features, calculating the peak value and skewness of the color distribution, and calculating anomaly scores based on statistics. The aim is to detect color distribution anomalies, ink shifts, or uneven color distribution in counterfeit areas of an image.

[0079] Specifically: (1) Based on the color feature vector F color Calculate the color histograms of the H, S, and V channels (usually 32 bins), and count the number of peaks P and the skewness Skew.

[0080] (2) Construct statistical outlier scores using color concentration and skewness: E color =w1×|Skew|+w2×(1-P / P max ); (3) The normalized outlier scores are used to obtain the final score: S color =1-normalize(E color ); (4) If the color distribution is smooth and the peaks are concentrated, S color The score is close to 1; if the color distribution is abnormal or multi-peaked, the score will decrease.

[0081] Specifically, when the average brightness is <50 (too dark) or >220 (overexposed), brightness correction is enabled and the calculation is recalculated; if the HSV channel fluctuates too much, the calculation is recalculated in Lab space.

[0082] S4-3: Shape feature edge continuity analysis based on contour analysis, using Canny edge detection. The purpose is to detect the continuity and integrity of pattern or structural edges to identify forgery or printing breaks.

[0083] Specifically: (1) Based on the shape feature vector F shape Canny edge detection is used to extract target boundaries; the edge length L of the connected region is calculated. total and the number of fracture segments N gap .

[0084] (2) Define the edge continuity ratio: C edge =1-(N gap / (L total / N gap )).

[0085] (3) Normalized C edge Obtain the shape score S shape S shape For edges ∈[0,1], the more continuous and less broken the edges, the higher the score.

[0086] S5: Determine the anti-counterfeiting detection score based on the anti-counterfeiting features, and combine the scores for image quality and feature vector quality to complete the anti-counterfeiting assessment of the document.

[0087] Specifically: S5-1: Calculation of image quality score for standard images.

[0088] S quality =0.3·S clarity +0.25·S brightness +0.2·S contrast +0.15·S color +0.1·S noise ; Among them, S clarity S represents the Laplacian variance; brightness The brightness uniformity after equalization; S contrast For the dynamic range of the histogram; S color S represents the skewness of the HSV distribution. noise To smooth the signal-to-noise ratio.

[0089] For example, ≥85: excellent image quality; 70–84: acceptable; <70: re-acquisition is required.

[0090] S5-2: Calculation of eigenvector quality score.

[0091] S feature =0.25·S LBP+0.25·S GLCM + 0.2·S color +0.15·S shape +0.15·S freq ; in, S LBP =100×exp(-α·Var(LBP)); S GLCM =100×(Energy / (1+Contrast)); S color =100×(1–w1|Skew|–w2(1–P / Pmax)); S shape =100×C edge ; S freq =100×normalize(SNR).

[0092] The grading system can be categorized as follows: ≥90: features are complete and reproducible; 70–89: features are stable; <70: features are unstable or missing.

[0093] S5-3: Calculation of anti-counterfeiting detection score.

[0094] S antifake =0.25·S texture +0.3·S color +0.25·S shape +0.2·S fusion ; Among them, S texture For texture consistency score; S color Score for color distribution anomalies; shape Score for shape consistency; S fusion For S total The weighted score.

[0095] Grading standards, for example: ≥90, anti-counterfeiting features are complete and verification is passed; 75–89, main features are consistent; 60–74, some abnormalities, manual review is required; <60, suspected of being counterfeit or distorted.

[0096] S5-4: Perform final weighting based on importance.

[0097] S total =0.4·S quality +0.35·S feature +0.25·S antifake ; The rating is defined as follows: ≥90, excellent (fully passed), high image quality and complete anti-counterfeiting features; 80–89, good (passable), slight deviation in features, which does not affect the result; 60–79, qualified (manual review), some features are missing or the signal-to-noise ratio is low; <60, unqualified (rejected), the image or anti-counterfeiting features are seriously abnormal.

[0098] As shown in Table 1-3.

[0099] Table 1. Image quality score;

[0100] Table 2. Feature vector quality score;

[0101] Table 3 Anti-counterfeiting detection scores;

[0102] In this embodiment, a rule-chain execution logic is designed: (1) Document suspected of being counterfeit: If any of the anti-counterfeiting detection score, image quality score and feature vector quality score is not met, the AI ​​verification anti-counterfeiting identification authentication will fail, or the watermark verification will fail, or the dynamic texture recognition will fail.

[0103] (2) Abnormal status of certificate: validity period T < the day; certificate number is inconsistent with the real name database, etc.

[0104] (3) High-risk users: ≥3 suspected forgery or ≥5 abnormal statuses submitted by the same device / person in the past 90 days; the similarity between the device fingerprint (the device model and terminal unique identifier collected by the APP) and the "professional certificate agency" label is ≥0.8.

[0105] This embodiment features dynamic rule hot updating: Threshold management: All thresholds can be configured in a gray-scale manner and subdivided by channel / region / certificate type; any upward / downward adjustment must be observed for ≥48 hours through offline playback and online 5% traffic gray-scale observation.

[0106] False alarm handling: Rules with a false alarm rate greater than 0.5% in a week will automatically enter "observation mode" and will be restored after third-party review.

[0107] This embodiment is designed with blockchain evidence storage, and key nodes are uploaded to the blockchain in real time.

[0108] The evidence includes: document image hash (SHA-256), verification request / response (de-identified), and decision log.

[0109] Interface security: Interaction with external systems uses HTTPS + Chinese national cryptographic SM2 / SM4, bidirectional TLS at the interface level, and all requests are uploaded to the blockchain according to the request body hash, which facilitates auditing and traceability.

[0110] This embodiment provides an automated review method for online freight documents. It utilizes online AI intelligent recognition and real-time comparison with government systems, with consistency rate with manual review, rejection / re-review rate, and average processing time as the main indicators. A four-fold security closed-loop system (AI anti-counterfeiting + multi-source real-time verification + risk engine + blockchain evidence storage) is employed to reduce the average online review time, improve the counterfeit detection rate, and reduce the rate of missed expired documents.

[0111] It should be noted that all data acquisition is conducted in accordance with laws and regulations and with user consent, and the data is used legally.

[0112] Example 2 This embodiment provides an automated verification system for online freight documents, including: The preprocessing module is configured to acquire the document image and perform noise reduction and contrast enhancement preprocessing on it; The standard processing module is configured to binarize the obtained enhanced image, extract the target boundary, and obtain a standard image after perspective correction of the obtained edge image. The feature extraction module is configured to extract anti-counterfeiting features based on edge images and standard images, including texture features, color features, shape features, and frequency domain features; The anti-counterfeiting detection module is configured to determine the anti-counterfeiting detection score based on the anti-counterfeiting features, and combine the scores of image quality and feature vector quality to complete the anti-counterfeiting assessment of the document.

[0113] It should be noted that the above modules correspond to the steps described in Embodiment 1, and the examples and application scenarios implemented by the above modules and the corresponding steps are the same, but are not limited to the content disclosed in Embodiment 1. It should also be noted that the above modules, as part of the system, can be executed in a computer system such as a set of computer-executable instructions.

[0114] In further embodiments, the following is also provided: An electronic device includes a memory and a processor, as well as computer instructions stored in the memory and running on the processor, wherein the computer instructions, when executed by the processor, perform the method described in Embodiment 1. For brevity, further details are omitted here.

[0115] It should be understood that in this embodiment, the processor can be a central processing unit (CPU), or it can be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor, etc.

[0116] Memory may include read-only memory and random access memory, and provides instructions and data to the processor. A portion of memory may also include non-volatile random access memory. For example, memory may also store information about the device type.

[0117] A computer-readable storage medium for storing computer instructions, which, when executed by a processor, perform the method described in Embodiment 1.

[0118] The method in Example 1 can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules within the processor. The software modules can reside in readily available storage media in the field, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, or registers. This storage medium is located in memory, and the processor reads information from the memory and, in conjunction with its hardware, completes the steps of the above method. To avoid repetition, a detailed description is not provided here.

[0119] A computer program product includes a computer program that, when executed by a processor, implements the method described in Embodiment 1.

[0120] The present invention also provides at least one computer program product tangibly stored on a non-transitory computer-readable storage medium. The computer program product includes computer-executable instructions, such as instructions included in program modules, which execute in a device on a target real or virtual processor to perform the processes / methods described above. Typically, program modules include routines, programs, libraries, objects, classes, components, data structures, etc., that perform specific tasks or implement specific abstract data types. In various embodiments, the functionality of program modules can be combined or divided among program modules as needed. The machine-executable instructions for the program modules can execute within a local or distributed device. In a distributed device, the program modules can reside in both local and remote storage media.

[0121] The computer program code used to implement the methods of the present invention may be written in one or more programming languages. This computer program code may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when executed by the computer or other programmable data processing device, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code may be executed entirely on a computer, partially on a computer, as a stand-alone software package, partially on a computer and partially on a remote computer, or entirely on a remote computer or server.

[0122] In the context of this invention, computer program code or related data may be carried by any suitable carrier to enable a device, apparatus, or processor to perform the various processes and operations described above. Examples of carriers include signals, computer-readable media, and the like. Examples of signals may include electrical, optical, radio, sound, or other forms of propagation signals, such as carrier waves, infrared signals, etc.

[0123] Those skilled in the art will recognize that the units and algorithm steps described in connection with the various examples of this embodiment can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this invention.

[0124] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of the present invention are still within the scope of protection of the present invention.

Claims

1. A method for automated verification of online freight documents, characterized in that, include: Acquire the document image and preprocess it by denoising and contrast enhancement; The enhanced image is binarized and the target boundary is extracted. The obtained edge image is then subjected to perspective correction to obtain the standard image. Anti-counterfeiting features are extracted from edge images and standard images, including texture features, color features, shape features and frequency domain features; The anti-counterfeiting detection score is determined based on the anti-counterfeiting features, and the anti-counterfeiting assessment of the document is completed by combining the scores of image quality and feature vector quality.

2. The automated verification method for online freight documents as described in claim 1, characterized in that, The preprocessing process includes: performing denoising on the brightness channel of the ID card image, using a two-dimensional Gaussian kernel for convolution smoothing to obtain a denoised image, and enhancing the contrast of the denoised image through histogram equalization.

3. The automated verification method for online freight documents as described in claim 1, characterized in that, Edges are detected in the binary image, and dilation is performed on the edge image to repair breaks. Contours are extracted and candidate quadrilateral regions are obtained by polygon approximation. The corner positions and geometric proportions are verified. The edge image is mapped to a standard size. If corner detection fails, Hough lines are used to find the intersection points or the minimum bounding rectangle as a degradation scheme.

4. The automated verification method for online freight documents as described in claim 1, characterized in that, Based on the anti-counterfeiting features, a score is determined for texture consistency, color distribution anomaly, and shape consistency. These scores are then weighted to obtain the anti-counterfeiting detection score.

5. The automated verification method for online freight documents as described in claim 1, characterized in that, The image quality score is calculated by weighting the Laplacian variance, brightness uniformity, histogram dynamic range, HSV distribution skewness, and smooth signal-to-noise ratio of the standard image.

6. The automated verification method for online freight documents as described in claim 1, characterized in that, The feature extraction score is calculated as follows: the LBP variance is calculated based on the texture feature vector, the color anomaly score is calculated based on the color features, the edge continuity ratio is calculated based on the shape features, and the score is obtained by weighted calculation based on the frequency domain features.

7. An automated verification system for online freight documents, characterized in that, include: The preprocessing module is configured to acquire the document image and perform noise reduction and contrast enhancement preprocessing on it; The standard processing module is configured to binarize the obtained enhanced image, extract the target boundary, and obtain a standard image after perspective correction of the obtained edge image. The feature extraction module is configured to extract anti-counterfeiting features based on edge images and standard images, including texture features, color features, shape features, and frequency domain features; The anti-counterfeiting detection module is configured to determine the anti-counterfeiting detection score based on the anti-counterfeiting features, and combine the scores of image quality and feature vector quality to complete the anti-counterfeiting assessment of the document.

8. An electronic device, characterized in that, It includes a memory and a processor, as well as computer instructions stored in the memory and running on the processor, which, when executed by the processor, perform the method according to any one of claims 1-6.

9. A computer-readable storage medium, characterized in that, Used to store computer instructions, which, when executed by a processor, perform the method described in any one of claims 1-6.

10. A computer program product, characterized in that, Includes a computer program, which, when executed by a processor, implements the method described in any one of claims 1-6.

Images