Cloud networking video AI image enhancement method, system, device, medium and terminal

Abstract

The application belongs to the technical field of image processing, and discloses a cloud networking video AI image enhancement method, system, device, medium and terminal. The cloud networking video AI image enhancement method comprises the following steps: performing image analysis on a road material database composed of roads, vehicles and geological structures, combing image attributes, and classifying and collating, cooperating with an image enhancement algorithm, and obtaining a clear picture through material replacement, picture repair, contrast enhancement and intelligent deployment. The application enhances the definition of standard definition and below images, so that it meets the requirements of high-definition video. The application changes the original unclear image into a high-definition image through the algorithm of material replacement. Since the highway video is relatively single compared with the public security monitoring image content, the picture improvement effect is relatively obvious, thereby improving the accuracy of the intelligent analysis and event detection functions in the later stage, and the application effect is obvious. The application reduces the construction cost of the monitoring device. The application has good compatibility.

Description

Technical Field

[0001] This invention belongs to the field of image processing technology, and in particular relates to a cloud-connected video AI image enhancement method, system, device, medium and terminal. Background Technology

[0002] Since the implementation of the video cloud networking project, the goal of uploading all traffic segment surveillance videos to the cloud has been achieved, allowing the Ministry of Transport, provincial governments, and application units to directly access relevant road videos. However, due to the massive number of videos and the varying construction times of the monitoring points, many video points are still using relatively outdated cameras, resulting in insufficient video image clarity. This hinders the smooth implementation of subsequent intelligent analysis and event detection of on-site videos. Furthermore, the sheer number of cameras requiring replacement necessitates not only substantial investment but also a lengthy timeframe.

[0003] From a comprehensive perspective, traffic video images primarily consist of visual subjects such as highways, bridges, tunnels, vehicles, surrounding greenery, and geology. Furthermore, over a considerable distance along a single road, the main video information does not change significantly. Therefore, establishing a road material database based on highway video information, and using intelligent AI to supplement and refine the material details of traffic videos, can greatly improve image clarity and enable functions such as intelligent analysis and event detection in high-definition images.

[0004] Based on the above analysis, the problems and shortcomings of the existing technology are as follows:

[0005] (1) The image clarity effect after existing technology is poor.

[0006] (2) The existing technology does not change the material algorithm, so the processed image cannot provide accurate information for subsequent intelligent analysis and event detection functions.

[0007] (3) Existing monitoring equipment has high construction costs and poor compatibility. Summary of the Invention

[0008] To overcome the problems existing in related technologies, the present invention discloses an embodiment of a cloud-connected video AI image enhancement method, system, device, medium, and terminal.

[0009] The technical solution is as follows: A cloud-connected video AI image enhancement system includes:

[0010] The video source data processing module is used to parse and analyze the video data uploaded from various roads through a private protocol, and output video data that meets the requirements.

[0011] The road material database is used to extract materials from road cameras, organize materials, and intelligently create materials after machine learning, forming a road material database containing road, vehicle, and geological information.

[0012] The intelligent AI enhancement algorithm module is used to perform image analysis on images of roads, vehicles, and geology in the road material database, sort out image attributes, and classify them. In conjunction with the image enhancement algorithm, it enhances road, vehicle, and geology images through material replacement, image repair, contrast enhancement, and intelligent allocation.

[0013] The video output module is used to output enhanced video images of roads, vehicles, and geological images.

[0014] Another objective of this invention is to provide a cloud-based video AI image enhancement method, comprising: utilizing an established road material database, and through AI video recognition, intelligently upgrading existing standard-definition images to high-definition images, obtaining high-definition road images without replacing existing cameras. This reduces hardware investment and improves monitoring effectiveness.

[0015] In one embodiment, the cloud-connected video AI image enhancement method further includes: performing image analysis on the acquired road material database consisting of roads, vehicles, and geology, sorting out image attributes, classifying and organizing them, and using image enhancement algorithms to obtain a clear image through material replacement, image repair, contrast enhancement, and intelligent adjustment.

[0016] The image analysis includes:

[0017] Fitting a multidimensional material model data volume involves first establishing a three-dimensional coordinate system for the material image using OpendTect software, based on the number and type of the imported material image data.

[0018] Next, import the processed material image data into the OpendTect software;

[0019] Then confirm the spacing of the survey lines and the channel spacing of the material data during material acquisition. Input the spacing-related parameters in the Manipulate module of the OpendTect software and determine the corresponding calculation function.

[0020] Finally, the material image data is interpolated using calculation commands to construct a three-dimensional spatial information map of the material.

[0021] The image attribute analysis includes the analysis of coherence attributes, instantaneous attributes, velocity attributes, and range attributes.

[0022] In one embodiment, based on the constructed 3D spatial information map, the cross-sectional view and attribute analysis type of the attribute analysis are determined, and the corresponding attribute type is selected using the Attribute module in the image analysis function. First, coherent attributes are applied to the horizontal cross-sectional view. Coherent attributes quantify the similarity of material waveforms in the axial and vertical directions to obtain the 3D spatial information of the material and obtain the preliminary suspected anomalies. Second, instantaneous attributes are applied to the longitudinal cross-sectional view. Instantaneous attributes highlight the slight changes in horizontal continuity. Then, velocity attributes are applied to the transverse cross-sectional view. Velocity attributes obtain a spectrum containing rich information.

[0023] In one embodiment, the classification and organization includes:

[0024] (1) Project STECs from different monitoring line-of-sight directions into VTECs; assume that the full-time images monitored by the local image processing unit of the high-definition camera are represented as STECs. GNSS The still image of the local image processing unit of the high-definition camera is represented as STEC. LEO-BTM The dynamic image is represented as STEC. LEO-UP Then, when projected onto the direction of the traffic segment, they are represented as VTEC. GNSS VTEC LEO-BTM and VTEC LEO-UP To avoid large projection errors, select the elevation angle intersecting the road surface as much as possible.

[0025]

[0026] In the formula, α and β represent STEC, respectively. GNSS and STEC LEO-BTM The distance of the traffic segment corresponding to the puncture point, γ represents The distance between high-definition cameras and traffic sections; among them, R is the average radius of the projection; H is the height of the projection; z1 and z2 represent STEC respectively. GNSS STEC LEO-BTM The distance of the traffic section at the corresponding high-definition camera location;

[0027] (2) The TEC values ​​of the dynamic and static images of the traffic segment were calculated by introducing the IRI model, and the calculated values ​​were VTEC and VTEC, respectively. IRI-BTM and VTEC IRI-UP Based on the material TEC upper and lower part ratio relationship represented by the IRI model, the preliminary classification of all-time image monitoring values ​​is calculated, as shown in the following formula:

[0028]

[0029] In the formula, VTEC LEO-BTM-ALLFor the total quantity of materials, VTEC LEO-UP-ALL For material properties, ξ LEO-BTM ξ represents a systematic bias where the total number of materials after classification is not modeled. LEO-UP This represents a systematic bias where the material properties after classification are not modeled; in this case, VTEC LEO-BTM-ALL and VTEC LEO-UP-ALL Both are related to VTEC GNSS The same complete material TEC value.

[0030] In one embodiment, the image enhancement algorithm employs the Retinex method, including the following steps:

[0031] Introducing a semi-parametric Retinex model, ξ LEO-BTM ξ represents a systematic bias where the total number of materials after classification is not modeled. LEO-UP To treat systematic deviations in material properties that have not been modeled after classification as nonparametric parameters and separate them from random errors, the monitoring equation is expressed as follows:

[0032] L = BX + S + Δ;

[0033] In the formula, L is the monitoring vector; Δ is the error vector; B is the full-rank design matrix; P is a symmetric positive definite matrix, which is the weight of the monitoring value L; S = (s1, s2, ..., s n ) T This describes the unmodeled systematic bias between monitoring values ​​of different material types, i.e., the semi-parametric component;

[0034] To obtain unique solutions for both parametric and nonparametric components, a Retinex matrix and a smoothing factor are introduced into the adjustment criterion, namely:

[0035]

[0036] In the formula, R is a suitable rearrangement matrix, becoming the Retinex matrix; α is a given scalar factor, which, during the minimization process, is related to V and... It plays a balancing role between these factors and is called the smoothing factor.

[0037] In one embodiment, the estimated value of the unknown parameter X is:

[0038]

[0039] By selecting appropriate smoothing factors α and Retinex matrix R, estimates of the nonparametric components can be obtained. and the estimated values ​​of the parameter components In this way, when the Retinex matrix R is rearranged, the nonparametric components S and parametric components X, as well as random errors, are separated from the monitored values.

[0040] The Retinex matrix R is typically selected using the natural spline function method or the time series method; the smoothing factor α can be selected using methods such as the signal-to-noise ratio method, the L-curve method, and the cross-validation method.

[0041] In the above parameter estimation method, for the material model parameters among the unknown parameters, the following approach is adopted:

[0042]

[0043] In the formula, n and m are the order and degree of the spherical harmonic function, respectively. max The maximum order of the spherical harmonic function expansion; For the normalized Legendre function; λ represents the geomagnetic or geographic latitude of the material puncture point; s = λ - λ0 represents the diurnal longitude of the puncture point, where λ and λ0 are the puncture point and longitude, respectively. The coefficients of the spherical harmonic function to be estimated are denoted as .

[0044] Another object of the present invention is to provide a computer device including a memory and a processor, the memory storing a computer program, which, when executed by the processor, causes the processor to perform the cloud-connected video AI image enhancement method.

[0045] Another object of the present invention is to provide a computer-readable storage medium storing a computer program, which, when executed by a processor, causes the processor to perform the cloud-connected video AI image enhancement method.

[0046] Another object of the present invention is to provide an information data processing terminal, which, when executed on an electronic device, provides a user input interface to implement the cloud-connected video AI image enhancement method described above.

[0047] Combining all the above technical solutions, the advantages and positive effects of this invention are as follows:

[0048] First, addressing the technical problems existing in the prior art and the difficulty in solving them, this paper closely analyzes, in conjunction with the technical solution to be protected by this invention and the results and data obtained during the research and development process, how the technical solution of this invention solves the technical problems, and the inventive technical effects brought about by solving these problems. The specific description is as follows:

[0049] The image analysis described in this invention includes: fitting a multidimensional material model data volume, which involves first establishing a three-dimensional coordinate system for the material image using OpendTect software based on the quantity and type of imported material image data; importing the processed material image data into OpendTect software; confirming the spacing of the survey lines and the channel spacing of the material data during material acquisition, inputting parameters related to the spacing in the Manipulate module of OpendTect software, and determining the corresponding calculation function; completing the interpolation of the material image data through calculation commands to construct a three-dimensional spatial information map of the material; and obtaining a clear material image containing more image information.

[0050] The classification and organization described in this invention includes: projecting STECs from different monitored line-of-sight directions into VTECs; assuming that the full-time images monitored by the local image processing unit of the high-definition camera are represented as STECs. GNSS The still image of the local image processing unit of the high-definition camera is represented as STEC. LEO-BTM The dynamic image is represented as STEC. LEO-UP Then, when projected onto the direction of the traffic segment, they are represented as VTEC. GNSS VTEC LEO-BTM and VTEC LEO-UP To avoid large projection errors, selecting the elevation angle intersecting the road surface as much as possible can obtain material images in different states.

[0051] The image enhancement algorithm of this invention adopts the Retinex method, introducing a semi-parametric Retinex model, and using ξ LEO-BTM ξ represents a systematic bias where the total number of materials after classification is not modeled. LEO-UP Systematic biases in material properties that are not modeled after classification are treated as nonparametric parameters and separated from random errors, thereby enhancing cloud-connected video AI images.

[0052] Second, considering the technical solution as a whole or from the perspective of the product, the technical effects and advantages of the technical solution to be protected by this invention are specifically described as follows:

[0053] This invention enhances the clarity of standard definition and lower definition images, enabling them to meet the requirements of high definition video.

[0054] By changing the material and algorithm, the original unclear image was transformed into a high-definition image. Since highway video content is relatively simple compared to public security surveillance images, the image improvement is more significant, thereby enhancing the accuracy of subsequent intelligent analysis and event detection functions, resulting in a noticeable application effect.

[0055] This invention reduces the construction costs of surveillance equipment. The challenge lies in replacing a large number of standard-definition (SD) cameras with high-definition (HD) cameras, which requires not only replacing the cameras but also upgrading storage and transmission equipment, increasing bandwidth, and other related work, resulting in high upgrade costs. By directly improving image quality at the backend through traffic video cloud networking AI image enhancement methods, the cost of upgrading is significantly reduced, resulting in better economic benefits.

[0056] This invention has good compatibility. This method is for video sources that conform to the national standard GB28181 and various proprietary decoding protocols, covering mainstream video equipment suppliers and has a wide range of applications.

[0057] Third, as supporting evidence of the inventive step of the claims of this invention, it is also reflected in the following: By efficiently classifying and organizing the image library, the speed of automatic material AI patching is improved, making the images meet the requirements of real-time monitoring. In the market, the processing of standard definition images is often done by directly replacing the camera. However, replacing cameras involves modifying network transmission bandwidth, which often causes image transmission delays and loss. Processing images with AI at the monitoring backend is not only low-cost and fast to implement, but also meets the functional requirements of real-time event analysis in intelligent transportation. Attached Figure Description

[0058] The accompanying drawings, which are incorporated in and form a part of this specification, illustrate embodiments consistent with this disclosure and, together with the description, serve to explain the principles of this disclosure.

[0059] Figure 1 This is a schematic diagram of the cloud-connected video AI image enhancement system provided in an embodiment of the present invention;

[0060] Figure 2 This is a flowchart of the cloud-connected video AI image enhancement method provided in an embodiment of the present invention;

[0061] Figure 3 This is a schematic diagram of the cloud-connected video AI image enhancement method provided in an embodiment of the present invention;

[0062] In the diagram: 1. Video source data processing module; 2. Road material database; 3. Intelligent AI enhancement algorithm module; 4. Image output module. Detailed Implementation

[0063] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Many specific details are set forth in the following description to provide a thorough understanding of the present invention. However, the present invention can be practiced in many other ways different from those described herein, and those skilled in the art can make similar modifications without departing from the spirit of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0064] I. Explanation of the Implementation Example:

[0065] The cloud-connected video AI image enhancement method provided in this embodiment of the invention includes: using the established road material database 2, and through AI video recognition, intelligently and efficiently upgrading the original standard definition image to a high definition image, thereby obtaining a high definition image of the road without replacing the original camera, thereby reducing hardware investment and improving monitoring effect.

[0066] Example 1

[0067] like Figure 1 As shown, the cloud-connected video AI image enhancement system provided in this embodiment of the invention is based on road classification information after traffic videos are uploaded to the cloud. It includes a video source data processing module 1, a road material database 2, an intelligent AI enhancement algorithm module 3, and an image output module 4.

[0068] Video source data processing module 1,

[0069] The video source data consists of video data uploaded from various roads to the provincial or ministerial center. This video data forms the data foundation of this method; all video image processing and handling are performed on this module. The module also supports mainstream domestic image compression and processing methods, conforms to GB28181 and proprietary protocols, and enables the parsing and analysis of existing images to output compliant video data.

[0070] Road Material Database 2 mainly includes materials extracted from high-definition road cameras, organized materials, and materials intelligently generated after machine learning. These data constitute the road material database 2 for roads, vehicles, geology, etc., which can be used directly according to the needs of the algorithm.

[0071] The intelligent AI enhancement algorithm module 3 mainly includes image analysis, classification and organization, Retinex image enhancement algorithm, and material replacement and enhancement. Its core is to intelligently analyze the original standard definition images, sort out image attributes, and classify and organize them. Combined with image enhancement algorithms, it improves image quality through processes such as material replacement, image repair, contrast enhancement, and intelligent adjustment, ultimately achieving high-definition image quality.

[0072] The video output module 4 mainly outputs video images that conform to standards such as GB28181. These images can be used directly or stored by mainstream domestic monitoring equipment suppliers. It can also support video display on a wall and achieve high-definition video output.

[0073] Example 2

[0074] Figure 2 The cloud-connected video AI image enhancement method provided in this embodiment of the invention includes:

[0075] S101, Obtain the original standard definition image of the road material database 2, which includes road, vehicle, and geological components;

[0076] S102, Perform intelligent image analysis, sort out image attributes, and classify them;

[0077] S103 utilizes image enhancement algorithms to obtain clear images through material replacement, image repair, contrast enhancement, and intelligent adjustment.

[0078] Specifically, it includes the following:

[0079] In one embodiment, the image analysis includes:

[0080] Fitting a multidimensional material model data volume involves first establishing a three-dimensional coordinate system for the material image using OpendTect software, based on the number and type of the imported material image data.

[0081] Next, import the processed material image data into the OpendTect software;

[0082] Then confirm the spacing of the survey lines and the channel spacing of the material data during material acquisition. Input the spacing-related parameters in the Manipulate module of the OpendTect software and determine the corresponding calculation function.

[0083] Finally, the material image data is interpolated using calculation commands to construct a three-dimensional spatial information map of the material.

[0084] The image attribute analysis includes the analysis of coherence attributes, instantaneous attributes, velocity attributes, and range attributes.

[0085] In this embodiment of the invention, based on the constructed three-dimensional spatial information map, the cross-sectional view and attribute analysis type of attribute analysis are determined, and the corresponding attribute type is selected using the Attribute module in the image analysis function. First, coherent attributes are applied on the horizontal cross-sectional view. Coherent attributes quantify the similarity of material waveforms in the axial and vertical directions to obtain the three-dimensional spatial information of the material and obtain the preliminary suspected anomalies. Second, instantaneous attributes are applied on the longitudinal cross-sectional view. Instantaneous attributes highlight the slight changes in horizontal continuity. Then, velocity attributes are applied on the transverse cross-sectional view. Velocity attributes obtain a spectrum containing rich information.

[0086] The classification and organization include:

[0087] (1) Project STECs from different monitoring line-of-sight directions into VTECs; assume that the full-time images monitored by the local image processing unit of the high-definition camera are represented as STECs. GNSS The still image of the local image processing unit of the high-definition camera is represented as STEC. LEO-BTM The dynamic image is represented as STEC. LEO-UP Then, when projected onto the direction of the traffic segment, they are represented as VTEC. GNSS VTEC LEO-BTM and VTEC LEO-UP To avoid large projection errors, select the elevation angle intersecting the road surface as much as possible.

[0088]

[0089] In the formula, α and β represent STEC, respectively. GNSS and STEC LEO-BTM The distance of the traffic segment corresponding to the puncture point, γ represents STEC. LEO-UP The distance between high-definition cameras and traffic sections; among them, R is the average radius of the projection; H is the height of the projection; z1 and z2 represent STEC respectively. GNSS STEC LEO-BTM The distance of the traffic section at the corresponding high-definition camera location;

[0090] (2) The TEC values ​​of the dynamic and static images of the traffic segment were calculated by introducing the IRI model, and the calculated values ​​were VTEC and VTEC, respectively. IRI-BTM and VTEC IRI-UP Based on the material TEC upper and lower part ratio relationship represented by the IRI model, the preliminary classification of all-time image monitoring values ​​is calculated, as shown in the following formula:

[0091]

[0092] In the formula, VTEC LEO-BTM-ALL For the total quantity of materials, VTECLEO-UP-ALL For material properties, ξ LEO-BTM ξ represents a systematic bias where the total number of materials after classification is not modeled. LEO-UP This represents a systematic bias where the material properties after classification are not modeled; in this case, VTEC LEO-BTM-ALL and VTEC LEO-UP-ALL Both are related to VTEC GNSS The same complete material TEC value.

[0093] The image enhancement algorithm employs the Retinex method and includes the following steps:

[0094] Introducing a semi-parametric Retinex model, ξ LEO-BTM ξ represents a systematic bias where the total number of materials after classification is not modeled. LEO-UP To treat systematic deviations in material properties that have not been modeled after classification as nonparametric parameters and separate them from random errors, the monitoring equation is expressed as follows:

[0095] L = BX + S + Δ;

[0096] In the formula, L is the monitoring vector; Δ is the error vector; B is the full-rank design matrix; P is a symmetric positive definite matrix, which is the weight of the monitoring value L; S = (s1, s2, ..., s n ) T This describes the unmodeled systematic bias between monitoring values ​​of different material types, i.e., the semi-parametric component;

[0097] To obtain unique solutions for both parametric and nonparametric components, a Retinex matrix and a smoothing factor are introduced into the adjustment criterion, namely:

[0098]

[0099] In the formula, R is a suitable rearrangement matrix, becoming the Retinex matrix; α is a given scalar factor, which, during the minimization process, is related to V and... It plays a balancing role between these factors and is called the smoothing factor.

[0100] In one embodiment, the estimated value of the unknown parameter X is:

[0101]

[0102] By selecting appropriate smoothing factors α and Retinex matrix R, estimates of the nonparametric components can be obtained. and the estimated values ​​of the parameter components In this way, when the Retinex matrix R is rearranged, the nonparametric components S and parametric components X, as well as random errors, are separated from the monitored values.

[0103] The Retinex matrix R is typically selected using the natural spline function method or the time series method; the smoothing factor α can be selected using methods such as the signal-to-noise ratio method, the L-curve method, and the cross-validation method.

[0104] In the above parameter estimation method, for the material model parameters among the unknown parameters, the following approach is adopted:

[0105]

[0106] In the formula, n and m are the order and degree of the spherical harmonic function, respectively. max The maximum order of the spherical harmonic function expansion; For the normalized Legendre function; λ represents the geomagnetic or geographic latitude of the material puncture point; s = λ - λ0 represents the diurnal longitude of the puncture point, where λ and λ0 are the puncture point and longitude, respectively. The coefficients of the spherical harmonic function to be estimated are denoted as .

[0107] Example 3

[0108] like Figure 3 As shown, the cloud-connected video AI image enhancement method provided in this embodiment of the invention includes the following steps:

[0109] Step 1: Input the existing video image;

[0110] Step 2: Extract road material data and process video images.

[0111] Step 3: Integrate the existing data to establish road material database 2, and at the same time make the image data processed in step 2 conform to the national standard GB28181 image.

[0112] Step 4: Perform image analysis based on the road material database 2 obtained in Step 3 and the images conforming to the national standard GB28181;

[0113] Step 5: Classify and organize the images analyzed in Step 4;

[0114] Step 6: Perform image enhancement using the Retinex image enhancement algorithm;

[0115] Step 7: Replace and reinforce the materials;

[0116] Step 8: High-definition image output.

[0117] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0118] The information interaction and execution process between the above-mentioned devices / units are based on the same concept as the method embodiments of the present invention. For details on their specific functions and technical effects, please refer to the method embodiments section, and they will not be repeated here.

[0119] Those skilled in the art will understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the functions described above can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this invention. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0120] II. Application Examples:

[0121] Application Example 1

[0122] Taking road surveillance images as an example, a certain road has 87 fixed 720P bullet cameras and 8 newly replaced 4K overpass panoramic cameras. The total cost of replacing all the fixed bullet cameras is estimated at 920,000 yuan (including replacement of cameras, switches, and storage expansion).

[0123] For example, the traffic video cloud-based AI image enhancement method can be used as follows:

[0124] Based on the image processing computation requirements, four additional servers will be added for image data processing.

[0125] Using 4K panoramic cameras on site, information such as the basic material and address of the road is obtained, and a basic database of existing vehicles, roads, and addresses is established on the local server.

[0126] A 720P image input server enhances image quality through intelligent AI algorithms.

[0127] Input 4K high-definition images into the existing monitoring system's storage system.

[0128] It can save a lot of money.

[0129] Application Example 2

[0130] An application embodiment of the present invention provides a computer device comprising: at least one processor, a memory, and a computer program stored in the memory and executable on the at least one processor, wherein the processor executes the computer program to implement the steps in any of the above-described method embodiments.

[0131] Application Example 3

[0132] The present invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, can implement the steps in the above-described method embodiments.

[0133] Application Example 4

[0134] The present invention also provides an information data processing terminal, which is used to provide a user input interface to implement the steps in the above method embodiments when executed on an electronic device. The information data processing terminal is not limited to mobile phones, computers, and switches.

[0135] Application Example 5

[0136] The present invention also provides a server that, when executed on an electronic device, provides a user input interface to implement the steps described in the above method embodiments.

[0137] Application 6

[0138] The present invention provides a computer program product that, when run on an electronic device, enables the electronic device to implement the steps described in the various method embodiments above.

[0139] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of the present invention can be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include at least: any entity or device capable of carrying the computer program code to a photographing device / terminal device, a recording medium, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium. Examples include USB flash drives, portable hard drives, magnetic disks, or optical disks.

[0140] III. Evidence of the relevant effects of the embodiments:

[0141] Experimental results show that:

[0142] For example, upgrading 100 video feeds to high definition involves replacing cameras, switches, and increasing bandwidth. If a carrier network were leased, the bandwidth rental fee would be significantly increased, amounting to approximately 500,000 to 800,000 yuan. In contrast, AI image restoration only requires adding two servers, with a total cost not exceeding 130,000 yuan, offering a substantial cost advantage. Furthermore, camera replacement involves extensive outdoor construction, while AI image restoration only requires installing two servers in a data center.

[0143] AI images can simultaneously detect road events, such as illegal parking, littering, accidents, and weather conditions in real time, which can solve the problem of being unable to handle a large number of events due to unclear video footage.

[0144] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications, equivalent substitutions and improvements made by those skilled in the art within the scope of the technology disclosed in the present invention and within the spirit and principles of the present invention should be covered within the scope of protection of the present invention.

Claims

1. A cloud-connected video AI image enhancement system, characterized in that, The cloud-connected video AI image enhancement system includes: The video source data processing module (1) is used to parse and analyze the video data uploaded from each road through a private protocol and output video data that meets the requirements. The road material database (2) is used to extract materials from road cameras, organize materials, and intelligently create materials after machine learning, forming a road material database (2) containing road, vehicle, and geological information. The intelligent AI enhancement algorithm module (3) is used to perform image analysis on the images of roads, vehicles and geology in the road material database (2), sort out the image attributes and classify them, and enhance the images of roads, vehicles and geology through material replacement, image repair, contrast enhancement and intelligent allocation in conjunction with the image enhancement algorithm. The image output module (4) is used to output enhanced video images of roads, vehicles, and geological images; In the intelligent AI enhancement algorithm module (3), the classification and organization include: (1) Project STECs from different monitoring line-of-sight directions into VTECs; assuming that the full-time images monitored by the local image processing unit of the high-definition camera are represented as follows: The still image represented by the local image processing unit of the high-definition camera is as follows: Dynamic images are represented as Then, when projected onto the traffic segment direction, they are represented as follows: , and To avoid large projection errors, select the elevation angle intersecting the road surface as much as possible. ； In the formula, , They represent and The distance of the traffic segment corresponding to the puncture point express The distance between high-definition cameras and traffic sections; among them, , ; H is the average radius of the projection; H is the height of the projection. , They represent , The distance of the traffic section at the corresponding high-definition camera location; (2) The TEC values ​​of the dynamic and static images of the traffic segment were calculated by introducing the IRI model. The calculated values ​​are as follows: and Based on the material TEC upper and lower part ratio relationship represented by the IRI model, the preliminary classification of all-time image monitoring values ​​is calculated, as shown in the following formula: ； In the formula, For the total quantity of materials, For material properties, This is a systematic bias stemming from the fact that the total number of materials after classification was not modeled. This represents a systematic bias where the material properties after classification were not modeled; at this point, and All are with The same complete material TEC value; The image enhancement algorithm uses the Retinex method and includes the following steps: Introducing a semi-parametric Retinex model, This is a systematic bias stemming from the fact that the total number of materials after classification was not modeled. To treat systematic deviations in material properties that have not been modeled after classification as nonparametric parameters and separate them from random errors, the monitoring equation is expressed as follows: ； In the formula, For monitoring vectors; This is the error vector; Design a matrix to achieve full column rank; It is a symmetric positive definite square matrix, which is the monitored value. The right; This describes the unmodeled systematic bias between monitoring values ​​of different material types, i.e., the semi-parametric component; To obtain unique solutions for both parametric and nonparametric components, a Retinex matrix and a smoothing factor are introduced into the adjustment criterion, namely: ； In the formula, A properly arranged matrix is ​​called a Retinex matrix; For a given scalar factor, during the minimization process... and It plays a balancing role between these factors and is called the smoothing factor; The estimated value of the unknown parameter X is: ； By selecting an appropriate smoothing factor and Retinex matrix This yields the estimates of the nonparametric components. and the estimated values ​​of the parameter components Thus, when the Retinex matrix... During the processing, nonparametric components are... and parameter components And random errors are separated from the monitored values; For Retinex matrix The selection of smoothing factors is typically achieved using natural spline function methods or time series methods; The selection can be made using methods such as signal-to-noise ratio, L-curve method, and cross-validation. In the above parameter estimation method, for the material model parameters among the unknown parameters, the following approach is adopted: ； In the formula, n and m are the order and degree of the spherical harmonic function, respectively. The maximum order of the spherical harmonic function expansion; For the normalized Legendre function; The geomagnetic or geographical latitude at the point of material penetration; The fixed longitude of the puncture point. These are the puncture point and longitude, respectively. , The coefficients of the spherical harmonic function to be estimated are denoted as .

2. A cloud-connected video AI image enhancement method for a cloud-connected video AI image enhancement system according to claim 1, characterized in that, The cloud-connected video AI image enhancement method includes: Using the established road material database (2), the original standard definition image is intelligently upgraded to a high definition image through AI video recognition, and a high definition image of the road is obtained without replacing the original camera; In the obtained high-definition road images, the obtained road material database (2) is analyzed, the image attributes are sorted out and classified, and with the help of image enhancement algorithms, a clear image is obtained through material replacement, image repair, contrast enhancement and intelligent adjustment. The image analysis includes: Fitting a multidimensional material model data volume involves first establishing a three-dimensional coordinate system for the material image using OpendTect software, based on the number and type of the imported material image data. Next, import the processed material image data into the OpendTect software; Then confirm the spacing of the survey lines and the channel spacing of the material data during material acquisition. Input the spacing-related parameters in the Manipulate module of the OpendTect software and determine the corresponding calculation function. Finally, the material image data is interpolated using calculation commands to construct a three-dimensional spatial information map of the material. The image attribute analysis includes the analysis of coherence attributes, instantaneous attributes, velocity attributes, and range attributes; Based on the constructed 3D spatial information map, the cross-sectional view and attribute analysis type of the attribute analysis are determined, and the corresponding attribute type is selected using the Attribute module in the image analysis function. First, coherence attributes are applied to the horizontal cross-sectional view. Coherence attributes quantify the similarity of material waveforms in the axial and vertical directions to obtain the 3D spatial information of the material and obtain the preliminary suspected anomalies. Second, instantaneous attributes are applied to the longitudinal cross-sectional view. Instantaneous attributes highlight the slight changes in horizontal continuity. Then, velocity attributes are applied to the transverse cross-sectional view. Velocity attributes obtain a spectrum containing rich information.

3. A computer device, characterized in that, The computer device includes a memory and a processor. The memory stores a computer program, which, when executed by the processor, causes the processor to perform the cloud-connected video AI image enhancement method according to claim 2.

4. A computer-readable storage medium storing a computer program, which, when executed by a processor, causes the processor to perform the cloud-connected video AI image enhancement method of claim 2.

5. An information data processing terminal, characterized in that, The information data processing terminal is used to provide a user input interface when executed on an electronic device to implement the method described in claim 2 for cloud-connected video AI image enhancement.

Images