WO2021047211A1 - 基于神经网络的数据验证方法、装置、设备及存储介质 - Google Patents
基于神经网络的数据验证方法、装置、设备及存储介质 Download PDFInfo
- Publication number
- WO2021047211A1 WO2021047211A1 PCT/CN2020/093590 CN2020093590W WO2021047211A1 WO 2021047211 A1 WO2021047211 A1 WO 2021047211A1 CN 2020093590 W CN2020093590 W CN 2020093590W WO 2021047211 A1 WO2021047211 A1 WO 2021047211A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- repair plan
- picture
- damaged
- repair
- damage degree
- Prior art date
Links
- 238000013524 data verification Methods 0.000 title claims abstract description 59
- 238000000034 method Methods 0.000 title claims abstract description 55
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 36
- 230000008439 repair process Effects 0.000 claims abstract description 440
- 238000012423 maintenance Methods 0.000 claims abstract description 146
- 238000004458 analytical method Methods 0.000 claims abstract description 39
- 238000012795 verification Methods 0.000 claims abstract description 39
- 230000004044 response Effects 0.000 claims description 47
- 238000013527 convolutional neural network Methods 0.000 claims description 21
- 238000012545 processing Methods 0.000 claims description 18
- 238000004590 computer program Methods 0.000 claims description 14
- 230000009466 transformation Effects 0.000 claims description 9
- 230000006870 function Effects 0.000 claims description 8
- 238000003709 image segmentation Methods 0.000 claims description 7
- 230000008569 process Effects 0.000 claims description 6
- 238000010606 normalization Methods 0.000 claims description 5
- 238000012935 Averaging Methods 0.000 claims description 4
- 230000011218 segmentation Effects 0.000 claims description 3
- 238000013507 mapping Methods 0.000 description 7
- 238000004364 calculation method Methods 0.000 description 5
- 238000004891 communication Methods 0.000 description 5
- 238000010586 diagram Methods 0.000 description 4
- 239000002184 metal Substances 0.000 description 4
- 238000010422 painting Methods 0.000 description 4
- 238000011176 pooling Methods 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 3
- 239000011159 matrix material Substances 0.000 description 3
- 230000003287 optical effect Effects 0.000 description 3
- 230000000694 effects Effects 0.000 description 2
- 230000010365 information processing Effects 0.000 description 2
- 230000000717 retained effect Effects 0.000 description 2
- PXFBZOLANLWPMH-UHFFFAOYSA-N 16-Epiaffinine Natural products C1C(C2=CC=CC=C2N2)=C2C(=O)CC2C(=CC)CN(C)C1C2CO PXFBZOLANLWPMH-UHFFFAOYSA-N 0.000 description 1
- 206010039203 Road traffic accident Diseases 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004140 cleaning Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 239000000835 fiber Substances 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 239000000446 fuel Substances 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 238000012549 training Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9535—Search customisation based on user profiles and personalisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/20—Administration of product repair or maintenance
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/70—Denoising; Smoothing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
Definitions
- This application relates to the technical field of big data services, in particular to the technical field of complaints and early warning, and in particular to a method, device, computer equipment and storage medium based on neural network data verification.
- the purpose of the embodiments of this application is to propose a neural network-based data verification method, device, computer equipment, and storage medium for verifying the method of repairing damaged parts of a car, and warning users in advance based on the verification results. To reduce potential safety hazards.
- an embodiment of the present application provides a data verification method based on a neural network, which adopts the following technical solutions:
- a data verification method based on neural network for car repair including the following steps:
- the verification result is fed back to the user.
- an embodiment of the present application also provides a neural network-based data verification device, which adopts the following technical solutions:
- a data verification device based on neural network including:
- the acquisition module is used to acquire the vehicle maintenance data to be verified sent by the user, and extract the damaged picture and the repair plan of the vehicle from the vehicle maintenance data;
- the matching module is used for data matching between the repair plan and the standard repair plan in the repair plan database
- the standard plan processing module is used to obtain the historical maintenance data corresponding to the standard repair plan from the repair plan database when matching the standard repair plan corresponding to the repair plan in the repair plan database, and extract the historical maintenance data
- the damage degree of the damaged picture and the historical damaged picture are analyzed respectively through the preset component damage degree analysis model, and the damage degree of the damaged picture response is calculated
- the difference between the degree of damage in response to the historical damage picture, and verify whether the vehicle maintenance data to be verified is correct according to the difference. If the vehicle maintenance data is verified to be incorrect, then the standard
- the repair plan updates the repair plan in the vehicle maintenance data;
- the non-standard plan processing module is used for when there is no standard repair plan corresponding to the repair plan in the repair plan database, use the maintenance data verification model to call the corresponding target repair plan according to the degree of damage, and according to the target repair plan and the said repair plan.
- the repair plan verifies whether the vehicle repair data to be verified is correct, and if the vehicle repair data is verified to be incorrect, the repair plan in the repair data is updated with the target repair plan;
- the feedback module is used to feed back the verification result to the user.
- the embodiments of the present application also provide a computer device, which adopts the following technical solutions:
- a computer device includes a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the following steps when the processor executes the computer program:
- the verification result is fed back to the user.
- the embodiments of the present application also provide a computer-readable storage medium, which adopts the following technical solutions:
- the verification result is fed back to the user.
- This embodiment is beneficial to overcome subjective errors caused by manual judgment, ensure the reliability of the repair plan, and avoid potential safety hazards caused by improper repairs.
- Figure 1 is an exemplary system architecture diagram to which the present application can be applied;
- Figure 2 is a flowchart of an embodiment of the applied neural network-based data verification method
- FIG. 3 is a flowchart of a specific implementation of step 203 in FIG. 2;
- FIG. 4 is a flowchart of a specific implementation of step 204 in FIG. 2;
- Fig. 5 is a flowchart of another embodiment of the neural network-based data verification method of the present application.
- Fig. 6 is a schematic structural diagram of an embodiment of the neural network-based data verification device of the present application.
- Fig. 7 is a schematic structural diagram of an embodiment of a computer device of the present application.
- the system architecture 100 may include terminal devices 101, 102, 103, a network 104, and a server 105.
- the network 104 is used to provide a medium for communication links between the terminal devices 101, 102, 103 and the server 105.
- the network 104 may include various connection types, such as wired, wireless communication links, or fiber optic cables, and so on.
- the user can use the terminal devices 101, 102, and 103 to interact with the server 105 through the network 104 to receive or send messages and so on.
- Various communication client applications such as web browser applications, shopping applications, search applications, instant messaging tools, email clients, and social platform software, may be installed on the terminal devices 101, 102, and 103.
- the terminal devices 101, 102, 103 may be various electronic devices with a display screen and supporting web browsing, including but not limited to user equipment, network equipment, or a device formed by integrating user equipment and network equipment through a network.
- the user equipment includes, but is not limited to, any mobile electronic product that can interact with a user through a touch panel, such as a smart phone, a tablet computer, etc., and the mobile electronic product can use any operating system, such as an android operating system. , IOS operating system, etc.
- the network device includes an electronic device that can automatically perform numerical calculation and information processing in accordance with pre-set or stored instructions. Its hardware includes, but is not limited to, a microprocessor, an application specific integrated circuit (ASIC), and a programmable gate.
- ASIC application specific integrated circuit
- the network device includes, but is not limited to, a computer, a network host, a single network server, a set of multiple network servers, or a cloud composed of multiple servers; here, the cloud is composed of a large number of computers or network servers based on Cloud Computing (Cloud Computing) Among them, cloud computing is a type of distributed computing, a virtual supercomputer composed of a group of loosely coupled computer sets.
- the network includes, but is not limited to, the Internet, a wide area network, a metropolitan area network, a local area network, a VPN network, and a wireless ad-hoc network (Ad Hoc network).
- Ad Hoc network wireless ad-hoc network
- the server 105 may be a server, or a server cluster composed of several servers, or a cloud computing service center. It may also be a server that provides various services, such as a background server that provides support for the pages displayed on the terminal devices 101, 102, and 103.
- the neural network-based data verification method provided in the embodiments of the present application is generally executed by the server 105, and correspondingly, the neural network-based data verification device is generally set in the server 105.
- terminal devices, networks, and servers in FIG. 1 are merely illustrative. According to implementation needs, there can be any number of terminal devices, networks, and servers.
- the neural network-based data verification method used in automobile maintenance includes the following steps:
- Step 201 Obtain the vehicle maintenance data to be verified sent by the user, and extract the damaged picture of the vehicle and the repair plan from the vehicle maintenance data.
- the electronic device for example, the server shown in Figure 1 on which the neural network-based data verification method for automobile maintenance runs can obtain the data to be verified sent by the user through a wired connection or a wireless connection.
- Car repair data and extract the damaged picture and repair plan of the car from the car repair data.
- the above-mentioned wireless connection methods can include, but are not limited to, 3G/4G connection, WiFi connection, Bluetooth connection, WiMAX connection, Zigbee connection, UWB (ultra wideband) connection, and other wireless connection methods currently known or developed in the future .
- the damaged picture and the repair plan are stored in a relational database.
- the repair plan can be found by mapping through the damaged picture.
- the damaged image can be found by mapping through the repair plan.
- each damaged picture forms a mapping relationship with the repair plan, or multiple damaged pictures are related to the repair plan in the form of a collection of pictures.
- the scheme forms a mapping relationship.
- the repair plan includes damaged parts of the automobile and the corresponding repair methods.
- the repair plan includes, but is not limited to, replacement, sheet metal repair, machine repair, electronic repair, painting and other repair methods.
- Fuel injector replace Chassis Machine repair Left front door Sheet metal repair, painting
- step 202 it is judged whether there is a standard repair plan consistent with the repair plan in the repair plan database, if yes, step 203 is executed, otherwise, step 205 is executed.
- the standard repair plan includes reasonable repair plans that have been implemented for the same parts stored in the repair plan database and repair plans that have not been implemented but reasonably exist. Further, in this embodiment, the repair plan in the repair plan database is matched with the repair plan of the damaged car to find the standard repair plan corresponding to the repair plan.
- repair plan database eliminates unreasonable repair plans through data cleaning. For example, a repair plan that eliminates the degree of damage to the parts requires simple repairs but replaces the parts.
- Step 203 Obtain historical maintenance data corresponding to the standard repair plan, and extract historical damaged pictures in the historical maintenance data.
- the corresponding historical damaged picture is found from the relational database.
- the historical damaged picture refers to the damaged picture of the same component stored in the repair plan database.
- the historical damaged pictures include one or more damaged pictures of the parts corresponding to the standard repair plan, and each of the damaged pictures forms a mapping relationship with the standard repair plan. Or a plurality of the damaged pictures form a mapping relationship with the standard repair plan in the form of a picture collection.
- Step 204 Analyze the damage degree of the damaged picture and the historical damaged picture through the neural network-based component damage degree analysis model, and calculate the damage degree of the damaged picture response and the damage degree of the historical damaged picture response. And verify whether the vehicle maintenance data to be verified is correct according to the difference.
- the damage degree of the damaged picture and the historical damaged picture reaction is digitized by the component damage degree analysis model, and the damage degree of the damaged picture reaction is calculated. Determine whether the difference between the damage degree of the damaged picture response and the damage degree of the historical damaged picture response is within the threshold range, when If the difference is within the threshold range, it is determined that the repair plan is reasonable and the verification result of the vehicle maintenance data to be verified is correct; otherwise, it is determined that the repair plan is unreasonable and the verification result of the vehicle maintenance data to be verified Is incorrect.
- the damage degree analysis model of the parts includes:
- Input layer used to input the damaged picture and the historical damaged picture into the component damage degree analysis model.
- Hidden layer used to input the damaged picture and the historical damaged picture input by the input layer, and perform image segmentation and removal on the undamaged picture, the damaged picture, and the historical damaged picture.
- the normalization processing and filtering of the mean image abstract the local features of the normalized undamaged picture, the damaged picture, and the historical damaged picture.
- Output layer on the basis of the local features calculated by the hidden layer, reassemble the undamaged picture, the damaged picture, and the historical damaged picture through a weight matrix. And based on the undamaged picture, analyze the damage degree of the damaged picture and the historical damaged picture and digitize it to calculate the difference between the two damage levels, and determine whether the repair plan is based on the difference. Reasonable. If the repair plan is reasonable, the verification result of the vehicle maintenance data to be verified is correct; otherwise, the verification result of the vehicle maintenance data to be verified is incorrect.
- Step 205 Analyze the damage degree of the damaged picture of the car through the maintenance data verification model, and call the corresponding target repair plan according to the damage degree, and judge whether the repair plan is reasonable according to the target repair plan. If the repair plan is reasonable, Then the verification result of the vehicle maintenance data to be verified is correct; otherwise, the verification result of the vehicle maintenance data to be verified is incorrect.
- the damage degree analysis is performed on the damaged picture through the maintenance data verification model, and the corresponding damage degree is called according to the damage degree.
- Target repair plan and match the repair plan with the target repair plan. When the target repair plan matches the repair plan, determine that the repair plan is reasonable; otherwise, determine that the repair plan is unreasonable .
- the maintenance data verification model includes the component damage degree analysis model, and the component damage degree analysis model is used to analyze the damage degree of the component reaction to obtain the damage picture response.
- the damage degree is compared with the preset damage degree threshold, and then the target repair plan is called according to the comparison result through the maintenance data verification model.
- the maintenance data verification model described in this embodiment simplifies the target repair plan into "repair” and "replacement.” When the damage degree of the damaged picture reaction is greater than the damage degree threshold, it is judged that the parts in the damaged picture should adopt the target repair plan of "replacement". When the damage degree of the damaged picture reaction is less than the damage degree threshold, it is judged that the parts in the damaged picture should adopt the target repair plan of "repair”.
- the target repair plan and the repair plan are compared, and when the two repair plans are consistent, the repair plan is determined to be reasonable, otherwise, the repair plan is determined to be unreasonable.
- the repair methods such as “sheet metal repair”, “mechanical repair”, “electronic repair”, and “painting” recorded in the repair plan are different from those in the target repair plan.
- the repair plan is the above-mentioned repair method, the repair method will be replaced by “replacement” during the comparison.
- Step 206 If the verification result of the vehicle maintenance data to be verified is incorrect, update the repair plan in the maintenance data with the standard repair plan or the target repair plan.
- the repair plan in the maintenance data is retained; when the verification result of the vehicle maintenance data to be verified is incorrect, Based on the unreasonable original repair plan, delete the original repair plan in the vehicle maintenance data, and use the standard repair plan or the target repair plan as a reasonable plan to replace the original repair plan, and complete the update of the automobile repair data , To ensure the correctness of the vehicle maintenance data.
- Step 207 Feed back the verification result to the user.
- the verification result is output. If the verification result is that the vehicle maintenance data is correct, the vehicle maintenance data with the original repair plan is retained; if the verification result is that the vehicle maintenance data is incorrect, the output verification result includes the updated vehicle maintenance data Car maintenance data. Furthermore, when verifying that the vehicle maintenance data is incorrect, reliable maintenance data is provided to ensure the quality of vehicle maintenance.
- the corresponding standard repair plan is matched with the repair plan in the vehicle maintenance data, and then the corresponding historical damaged picture is obtained, and then the damaged picture and the damaged picture are respectively analyzed by the component damage degree analysis model. Analyze the damage degree of historical damage pictures, judge whether the repair plan is reasonable according to the difference between the two damage degrees, or use the maintenance data verification model to call the corresponding target repair plan according to the damage degree, and then judge according to the target repair plan Whether the repair plan is reasonable, complete the verification of the vehicle repair data, and if the verification result of the vehicle repair data is incorrect, update the repair plan in the repair data with the standard repair plan or the target repair plan, which can not only verify the vehicle repair data It can also provide reliable maintenance data when the vehicle maintenance data is incorrect, so as to overcome the subjective error caused by manual evaluation, ensure the correctness of the vehicle maintenance data, and avoid potential safety hazards caused by improper maintenance programs. .
- Step 202 Determine whether there is a standard repair plan consistent with the repair plan in the repair plan database, which specifically includes the following steps:
- Step 301 Perform keyword comparison between the repair plan and the repair plan database to obtain the classification of the keywords. If the comparison is successful, go to step 302; otherwise, go to step 303.
- Step 302 Match the corresponding standard repair plan according to the classification.
- Step 303 Determine that there is no standard repair plan consistent with the repair plan in the repair plan database.
- a keyword list is set in the repair plan database, the keyword list classifies keywords, and a set of keywords belonging to the same category is mapped to the same standard repair plan.
- the classification corresponding to the keyword is found, and the standard repair plan consistent with the repair plan is retrieved in the repair plan database according to the classification. Otherwise, there is no standard repair plan corresponding to the repair plan in the repair plan database, and it is necessary to analyze the damage degree of the damaged picture through the repair data verification model, and call the corresponding target according to the damage degree
- the repair plan is to compare the target repair plan with the repair plan, and determine whether the repair plan is reasonable according to the comparison result, and then obtain a verification result for whether the vehicle repair data is correct.
- the first keyword is the repair method of the damaged component in the repair plan, and the first keyword can be directly extracted from the repair method.
- step 301 the keyword comparison between the repair plan and the repair plan database to obtain the classification of the keywords includes the following steps:
- Step 3011 Extract the first keyword in the repair plan.
- Step 3012 Initiate a search to the keyword list using the first keyword as a filter condition, and determine whether the first keyword corresponds to any second keyword in the keyword list, if yes, perform step 3013, otherwise, perform step 3014.
- Step 3013 Obtain the classification corresponding to the first keyword.
- Step 3014 Input the first keyword and the keyword list into the deep semantic matching model, search for a second keyword semantically matching the first keyword from the keyword list, and obtain a classification according to the second keyword.
- the first keyword in the repair plan is first extracted, and the first keyword is used as the filter condition in the repair plan.
- Search in the repair plan database and then determine whether the first keyword matches any second keyword in the keyword list, if so, obtain the category corresponding to the keyword, and then map the category to the standard repair plan The relationship matches the corresponding standard repair plan.
- the first keyword is not in the keyword list, input the first keyword and the keyword list data into the deep semantic matching model, and search for the first keyword from the keyword list.
- the second keyword that matches the semantics of the keyword, and the standard repair plan is obtained according to the category corresponding to the second keyword.
- the deep semantic matching model includes:
- the input layer inputs the first keyword in the repair plan and the keyword list in the form of a character vector (one-hot).
- the presentation layer is composed of at least one convolutional neural network.
- the convolutional neural network includes: a convolutional layer, a pooling layer, and a fully connected layer. Wherein, the convolutional layer is used to extract context features.
- the pooling layer is used to find global context features for the first keyword in the repair solution and the keyword list.
- the fully connected layer is used to convert the first keyword in the repair solution and the high-latitude semantic vector of the keyword list into a low-latitude semantic vector.
- the matching layer uses the cosine distance between the low-latitude semantic vector of the first keyword in the repair plan and the low-latitude semantic vector of the keyword list to represent the first key in the repair plan The semantic similarity between the word and each keyword in the keyword list. Then, the semantic similarity is transformed into a posterior probability through the softmax function. It is determined by the posterior probability whether the first keyword in the repair plan matches the semantics of any second keyword in the keyword list.
- the first keyword may be stored in the keyword list to expand the keyword list, thereby reducing the dependence on the deep semantic matching model and improving the matching efficiency.
- Step 204 Analyze the damage degree of the damaged picture and the historical damaged picture respectively through the component damage degree analysis model, and calculate the damage degree of the damaged picture response and the historical damaged picture response According to the difference of the degree of damage, and verifying whether the vehicle maintenance data to be verified is correct according to the difference, it specifically includes the following steps:
- Step 501 Obtain an undamaged picture of the damaged component, and process the damaged picture, the historically damaged picture, and the undamaged picture to obtain a picture in a standard format;
- Step 502 Input the processed undamaged pictures, damaged pictures, and historical damaged pictures into the convolutional neural network in the component damage degree analysis model, respectively;
- Step 503 Based on the undamaged pictures of the damaged parts, analyze and identify the damage degree of the damaged picture response of the damaged parts and the damage degree of the historical damaged picture response through the convolutional neural network, and calculate two Difference between
- Step 504 When the difference is within the preset threshold range, it is determined that the repair plan is reasonable and the verification result of the vehicle maintenance data to be verified is correct; otherwise, it is determined that the repair plan is unreasonable and the vehicle maintenance data to be verified The verification result of is incorrect.
- the image processing includes image segmentation, de-averaging, and image normalization processing to reduce image noise and enhance image effects.
- image segmentation is performed on the undamaged picture, the damaged picture, and the historically damaged picture to make the undamaged picture, the damaged picture, and the historically damaged picture
- the parts in are separated from the background of the picture.
- the undamaged pictures, the damaged pictures, and the historical damaged pictures after segmentation are de-averaged, and the dimensions of the undamaged pictures, the damaged pictures, and the historical damaged pictures are calculated. The characteristics of are all centered to zero.
- this embodiment uses the moments that are invariant to affine transformation among the undamaged pictures, the damaged pictures, and the historical damaged pictures to determine the parameters of the transformation function, that is, the invariant moments of the image are used. Find a set of parameters so that it can eliminate the influence of other transformation functions on image transformation, and then use the transformation function to transform the undamaged picture, the damaged picture, and the historically damaged picture into pictures in a standard format.
- the component damage degree analysis model analyzes the damage degree of the damaged picture and the historical damaged picture based on the normal picture (undamaged picture) of the component, and obtains a value of the damage degree In order to calculate the difference between the degree of damage between the two, so as to judge whether the repair plan is reasonable according to the difference, and obtain the verification result of the vehicle maintenance data.
- the component damage degree analysis model analyzes the damage degree of the damaged picture and the historical damaged picture through a convolutional neural network, and respectively judges the similarity between the two and the undamaged picture of the component degree.
- the damage degree analysis model of parts includes:
- a data input layer where the input layer is used to implement step 501;
- the convolution calculation layer is used to filter the normalized damaged picture and the historical damaged picture separately, and abstract the normalized damaged picture and The local features of the historically damaged pictures;
- Pooling layer the pooling layer is used to reduce the resolution of the local features to reduce the amount of calculation, and to enhance the robustness of the convolutional neural network;
- the fully connected layer is used to re-assemble the local features of the damaged picture into a complete damaged picture through a weight matrix, and re-pass the local features of the historical damaged picture through weights
- the matrix is assembled into a complete picture of the historical damage. Based on the undamaged picture, analyze the damage degree of the damaged picture and the historical damaged picture and digitize it to calculate the difference between the two damage levels, and judge the repair plan to be reasonable based on the difference. And verify whether the vehicle maintenance data is correct.
- the convolutional neural network is trained by inputting damaged pictures and undamaged pictures of different components, so that the convolutional neural network can accurately identify the affected parts of the damaged picture.
- the degree of damage and the degree of damage to the historically damaged picture response.
- the damage degree of the components in the damaged picture is greater, it is reflected in the convolutional neural network that the degree of similarity between the damaged picture and the undamaged picture is lower.
- the degree of damage to the components of the historically damaged picture is greater, it is reflected in the convolutional neural network that the degree of similarity between the historically damaged picture and the undamaged picture is lower.
- the computer program can be stored in a computer readable storage medium, and the program can be stored in a computer readable storage medium. When executed, it may include the procedures of the above-mentioned method embodiments.
- the aforementioned storage medium may be a non-volatile storage medium such as a magnetic disk, an optical disc, a read-only memory (Read-Only Memory, ROM), or a random access memory (Random Access Memory, RAM), etc.
- this application provides an embodiment of a data verification device based on a neural network.
- the device embodiment corresponds to the method embodiment shown in FIG. 2.
- the device can be applied to various electronic devices.
- the neural network-based data verification device 600 in this embodiment includes:
- the acquiring module 601 is used to acquire the vehicle maintenance data to be verified sent by the user, and extract the damaged picture and the repair plan of the vehicle from the vehicle maintenance data;
- the matching module 602 is used for data matching between the repair plan and the standard repair plan in the repair plan database;
- the standard plan processing module 603 is used to obtain the historical maintenance data corresponding to the standard repair plan from the repair plan database when matching the standard repair plan corresponding to the repair plan in the repair plan database, and extract the historical repair data from the repair plan database. Analyze the damage degree of the damaged picture and the historical damaged picture respectively through the preset component damage degree analysis model, and calculate the damage degree and the damage degree of the damaged picture response. The difference in the degree of damage reflected in the historical damaged picture, and verify whether the vehicle maintenance data to be verified is correct according to the difference, and if the vehicle maintenance data is verified to be incorrect, then repair it according to the standard Plan to update the repair plan in the vehicle repair data;
- the non-standard plan processing module 604 is used for when there is no standard repair plan corresponding to the repair plan in the repair plan database, use the maintenance data verification model to call the corresponding target repair plan according to the degree of damage, and according to the target repair plan and the repair plan.
- the repair plan verifies whether the vehicle repair data to be verified is correct, and if the verification of the vehicle repair data is incorrect, the repair plan in the vehicle repair data is updated with the target repair plan;
- the feedback module 605 is configured to output the verification result to the user.
- the repair plan in the repair data is matched to the corresponding standard repair plan, and then the corresponding historical damage picture is obtained, and then the damage degree analysis model is used to analyze the damage respectively.
- the picture and the historical damaged picture are analyzed for the degree of damage, and the vehicle maintenance data is judged according to the difference in the degree of damage, or the maintenance data verification model is used to call the corresponding target repair plan according to the degree of damage, and Verify whether the vehicle maintenance data to be verified is correct according to the target repair plan and the repair plan, and when verifying that the vehicle maintenance data to be verified is incorrect, provide a more reliable repair plan to ensure that the vehicle is repaired
- the correctness of the data not only helps overcome the subjective error caused by manual judgment, and provides users with early warning, but also provides reliable repair plans to customers when the vehicle repair data is incorrect, avoiding potential safety hazards caused by improper repairs .
- FIG. 7 is a block diagram of the basic structure of the computer device in this embodiment.
- the computer device 7 includes a memory 71, a processor 72, and a network interface 73 that are connected to each other in communication via a system bus. It should be noted that the figure only shows the computer device 7 with components 71-73, but it should be understood that it is not required to implement all the components shown, and more or fewer components may be implemented instead. Among them, those skilled in the art can understand that the computer device here is a device that can automatically perform numerical calculation and/or information processing in accordance with pre-set or stored instructions.
- Its hardware includes, but is not limited to, a microprocessor, a dedicated Integrated Circuit (Application Specific Integrated Circuit, ASIC), Programmable Gate Array (Field-Programmable Gate Array, FPGA), Digital Processor (Digital Signal Processor, DSP), embedded equipment, etc.
- ASIC Application Specific Integrated Circuit
- ASIC Application Specific Integrated Circuit
- FPGA Field-Programmable Gate Array
- DSP Digital Processor
- the computer device may be a computing device such as a desktop computer, a notebook, a palmtop computer, and a cloud server.
- the computer device can interact with the user through a keyboard, a mouse, a remote control, a touch panel, or a voice control device.
- the memory 71 includes at least one type of readable storage medium.
- the readable storage medium includes flash memory, hard disk, multimedia card, card-type memory (for example, SD or DX memory, etc.), random access memory (RAM), and static memory.
- the memory 71 may be an internal storage unit of the computer device 7, such as a hard disk or memory of the computer device 7.
- the memory 71 may also be an external storage device of the computer device 7, such as a plug-in hard disk equipped on the computer device 7, a smart media card (SMC), a secure digital (Secure Digital, SD) card, flash card (Flash Card), etc.
- the memory 71 may also include both the internal storage unit of the computer device 7 and the external storage device thereof.
- the memory 71 is generally used to store an operating system and various application software installed in the computer device 7, such as program code of a neural network-based data verification method for auto parts maintenance.
- the memory 71 can also be used to temporarily store various types of data that have been output or will be output.
- the processor 72 may be a central processing unit (Central Processing Unit, CPU), a controller, a microcontroller, a microprocessor, or other data processing chips in some embodiments.
- the processor 72 is generally used to control the overall operation of the computer device 7.
- the processor 72 is configured to run the program code or process data stored in the memory 71, for example, run the program code of the neural network-based data verification method for automobile maintenance.
- the network interface 73 may include a wireless network interface or a wired network interface, and the network interface 73 is generally used to establish a communication connection between the computer device 7 and other electronic devices.
- This application also provides another implementation manner, that is, to provide a computer-readable storage medium that stores a neural network-based data verification program for automobile maintenance, and the computer-readable storage medium stores a neural network-based data verification program for automobile maintenance.
- the data verification program of the neural network may be executed by at least one processor, so that the at least one processor executes the steps of the above-mentioned neural network-based data verification method.
- the computer-readable storage medium may be a non-volatile storage medium or a volatile storage medium.
- the method of the above embodiments can be implemented by means of software plus the necessary general hardware platform. Of course, it can also be implemented by hardware, but in many cases the former is better. ⁇
- the technical solution of this application essentially or the part that contributes to the existing technology can be embodied in the form of a software product, and the computer software product is stored in a storage medium (such as ROM/RAM, magnetic disk, The optical disc) includes several instructions to enable a terminal device (which can be a mobile phone, a computer, a server, an air conditioner, or a network device, etc.) to execute the method described in each embodiment of the present application.
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Business, Economics & Management (AREA)
- Databases & Information Systems (AREA)
- Data Mining & Analysis (AREA)
- Human Resources & Organizations (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Software Systems (AREA)
- Biomedical Technology (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Economics (AREA)
- Entrepreneurship & Innovation (AREA)
- Marketing (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Strategic Management (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Multimedia (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
一种基于神经网络的数据的验证方法，用于汽车维修，包括步骤：获取用户发送的待验证的汽车维修数据，并提取汽车维修数据中的汽车受损图片以及修理方案，基于修理方案从修理方案数据库中匹配出相应的标准修理方案及其对应的历史受损图片，通过零部件受损程度分析模型分别对受损图片和历史受损图片进行受损程度分析，计算二者反应的受损程度的差值，根据差值验证汽车维修数据。当修理方案数据库中不存在相应的标准修理方案时，通过维修数据验证模型根据受损程度调取目标修理方案，根据目标修理方案验证汽车维修数据。若验证结果为不正确，则以标准修理方案或目标修理方案替换修理方案，提供可靠的汽车维修数据。
Description
本申请要求于2019年9月11日提交中国专利局、申请号为201910860773.X，发明名称为“基于神经网络的数据验证方法、装置、设备及存储介质”的中国专利申请的优先权，其全部内容通过引用结合在本申请中。
本申请涉及大数据服务技术领域，具体涉及投诉预警技术领域，尤其涉及一种基于神经网络数据验证方法、装置、计算机设备和存储介质。
随着我国生产力的发展，经济水平的提高，汽车保有量的急剧增长，轿车快速地进入普通百姓家庭，城市道路上的车辆密度越来越大，车辆之间发生交通事故也越来越频繁。目前常用的汽车维修方法大多是车主将待维修车辆开到维修厂，维修厂的工作人员通过对待维修车辆进行检查，对汽车零配件损坏的修换依据于修理厂商/4S店根据相关的经验或者规章确定出的维修方案，进而对车辆进行维修。然而，发明人意识到，汽车技术复杂，零配件数量众多，零配件称谓不规范，维修厂的工作人员的技术水平良莠不齐，这种采用人工判断的方式对汽车零配件的损耗进行判断，并拟定修理方案的方式无法确保维修方式的准确性，往往存在判断误差、修理方案可靠性偏低与修理效果不理想的问题，而不恰当的维修容易带来潜在的安全隐患。
发明内容
本申请实施例的目的在于提出一种基于神经网络的数据验证方法、装置、计算机设备以及存储介质，用于对修复汽车的受损零部件的方式进行验证，根据验证结果提前对用户进行预警，以减少安全隐患。
为了解决上述技术问题，本申请实施例提供一种基于神经网络的数据验证方法，采用了如下所述的技术方案：
一种基于神经网络数据验证方法，用于汽车维修，包括下述步骤：
获取用户发送的待验证的汽车维修数据，并提取所述汽车维修数据中的汽车受损图片以及修理方案；
将修理方案与修理方案数据库中的标准修理方案进行数据匹配，判断所述修理方案数据库中是否存在与所述修理方案一致的标准修理方案；
若存在，则从所述修理方案数据库中获取所述标准修理方案对应的历史维修数据，并提取所述历史维修数据中的历史受损图片，再通过预设的零部件受损程度分析模型分别对所述受损图片和所述历史受损图片进行受损程度分析，计算所述受损图片反应的受损程度与所述历史受损图片反应的受损程度的差值，并根据所述差值验证待验证的所述汽车维修数据是否正确；
若不存在，则通过维修数据验证模型分析所述汽车受损图片的受损程度，并根据受损程度调取对应的目标修理方案，根据目标修理方案与所述修理方案验证待验证的所述汽车维修数据是否正确；
若验证所述汽车维修数据为不正确，则以所述标准修理方案或者所述目标修理方案更新所述维修数据中的所述修理方案；
向所述用户反馈验证结果。
为了解决上述技术问题，本申请实施例还提供一种基于神经网络的数据验证装置，采用了如下所述的技术方案：
一种基于神经网络的数据验证装置，包括：
获取模块，用于获取用户发送的待验证的汽车维修数据，并从所述汽车维修数据中提取所述汽车的受损图片以及修理方案；
匹配模块，用于修理方案与修理方案数据库中的标准修理方案进行数据匹配；
标准方案处理模块，用于在修理方案数据库中匹配出与修理方案对应的标准修理方案时，从所述修理方案数据库中获取所述标准修理方案对应的历史维修数据，并提取所述历史维修数据中的历史受损图片，并通过预设的零部件受损程度分析模型分别对所述受损图片和所述历史受损图片进行受损程度分析，计算所述受损图片反应的受损程度与所述历史受损图片反应的受损程度的差值，并根据所述差值验证待验证的所述汽车维修数据是否正确，若验证所述汽车维修数据为不正确，则以所述标准修理方案更新所述汽车维修数据中的所述修理方案；
非标方案处理模块，用于在修理方案数据库中不存在与修理方案对应的标准修理方案时，通过维修数据验证模型根据受损程度调取对应的目标修理方案，并根据目标修理方案与所述修理方案验证待验证的所述汽车维修数据是否正确，若验证所述汽车维修数据为不正确，则以所述目标修理方案更新所述维修数据中的所述修理方案；
反馈模块，用于向所述用户反馈验证结果。
为了解决上述技术问题，本申请实施例还提供一种计算机设备，采用了如下所述的技术方案：
一种计算机设备，包括存储器和处理器，所述存储器中存储有计算机程序，所述处理器执行所述计算机程序时实现以下步骤：
获取用户发送的待验证的汽车维修数据，并提取所述汽车维修数据中的汽车受损图片以及修理方案；
将修理方案与修理方案数据库中的标准修理方案进行数据匹配，判断所述修理方案数据库中是否存在与所述修理方案一致的标准修理方案；
若存在，则从所述修理方案数据库中获取所述标准修理方案对应的历史维修数据，并提取所述历史维修数据中的历史受损图片，再通过预设的零部件受损程度分析模型分别对所述受损图片和所述历史受损图片进行受损程度分析，计算所述受损图片反应的受损程度与所述历史受损图片反应的受损程度的差值，并根据所述差值验证待验证的所述汽车维修数据是否正确；
若不存在，则通过维修数据验证模型分析所述汽车受损图片的受损程度，并根据受损程度调取对应的目标修理方案，根据目标修理方案与所述修理方案验证待验证的所述汽车维修数据是否正确；
若验证所述汽车维修数据为不正确，则以所述标准修理方案或者所述目标修理方案更新所述维修数据中的所述修理方案；
向所述用户反馈验证结果。
为了解决上述技术问题，本申请实施例还提供一种计算机可读存储介质，采用了如下所述的技术方案：
一种计算机可读存储介质，所述计算机可读存储介质上存储有计算机程序，所述计算机程序被处理器执行时实现以下步骤：
获取用户发送的待验证的汽车维修数据，并提取所述汽车维修数据中的汽车受损图片以及修理方案；
将修理方案与修理方案数据库中的标准修理方案进行数据匹配，判断所述修理方案数据库中是否存在与所述修理方案一致的标准修理方案；
若存在，则从所述修理方案数据库中获取所述标准修理方案对应的历史维修数据，并提取所述历史维修数据中的历史受损图片，再通过预设的零部件受损程度分析模型分别对所述受损图片和所述历史受损图片进行受损程度分析，计算所述受损图片反应的受损程度与所述历史受损图片反应的受损程度的差值，并根据所述差值验证待验证的所述汽车维修 数据是否正确；
若不存在，则通过维修数据验证模型分析所述汽车受损图片的受损程度，并根据受损程度调取对应的目标修理方案，根据目标修理方案与所述修理方案验证待验证的所述汽车维修数据是否正确；
若验证所述汽车维修数据为不正确，则以所述标准修理方案或者所述目标修理方案更新所述维修数据中的所述修理方案；
向所述用户反馈验证结果。
本实施例利于克服人工判断造成的主观误差，确保修理方案的可靠性，避免不恰当的维修导致的安全隐患。
图1是本申请可以应用于其中的示例性系统架构图；
图2是申请的基于神经网络的数据验证方法的一个实施例的流程图；
图3是图2中步骤203的一种具体实施方式的流程图；
图4是图2中步骤204的一种具体实施方式的流程图；
图5是本申请的基于神经网络的数据验证方法的另一个实施例的流程图；
图6是本申请的基于神经网络的数据验证装置的一个实施例的结构示意图；
图7是本申请的计算机设备的一个实施例的结构示意图。
本申请的说明书和权利要求书及上述附图说明中的术语“包括”和“具有”以及它们的任何变形，意图在于覆盖不排他的包含。本申请的说明书和权利要求书或上述附图中的术语“第一”、“第二”等是用于区别不同对象，而不是用于描述特定顺序。
本申请的技术方案可应用于大数据技术领域，涉及的数据可存储于数据库中，或者可以通过区块链分布式存储。本领域技术人员显式地和隐式地理解的是，本文所描述的实施例可以与其它实施例相结合。
如图1所示，系统架构100可以包括终端设备101、102、103，网络104和服务器105。网络104用以在终端设备101、102、103和服务器105之间提供通信链路的介质。网络104可以包括各种连接类型，例如有线、无线通信链路或者光纤电缆等等。
用户可以使用终端设备101、102、103通过网络104与服务器105交互，以接收或发送消息等。终端设备101、102、103上可以安装有各种通讯客户端应用，例如网页浏览器应用、购物类应用、搜索类应用、即时通信工具、邮箱客户端、社交平台软件等。
终端设备101、102、103可以是具有显示屏并且支持网页浏览的各种电子设备，包括但不限于用户设备、网络设备或用户设备与网络设备通过网络相集成所构成的设备。所述用户设备其包括但不限于任何一种可与用户通过触摸板进行人机交互的移动电子产品，例如智能手机、平板电脑等，所述移动电子产品可以采用任意操作系统，如android操作系统、IOS操作系统等。其中，所述网络设备包括一种能够按照事先设定或存储的指令，自动进行数值计算和信息处理的电子设备，其硬件包括但不限于微处理器、专用集成电路(ASIC)、可编程门阵列(FPGA)、数字处理器(DSP)、嵌入式设备等。所述网络设备其包括但不限于计算机、网络主机、单个网络服务器、多个网络服务器集或多个服务器构成的云；在此，云由基于云计算(Cloud Computing)的大量计算机或网络服务器构成，其中，云计算是分布式计算的一种，由一群松散耦合的计算机集组成的一个虚拟超级计算机。所述网络包括但不限于互联网、广域网、城域网、局域网、VPN网络、无线自组织网络(Ad Hoc网络)等。当然，本领域技术人员应能理解上述终端设备仅为举例，其他现有的或今后可能出现的终端设备如可适用于本申请，也应包含在本申请保护范围以内，并在此以引用方式包含于此。
服务器105可以是一台服务器，或者由若干台服务器组成的服务器集群，或者是一个云计算服务中心。其也可以是提供各种服务的服务器，例如对终端设备101、102、103上显示的页面提供支持的后台服务器。
需要说明的是，本申请实施例所提供的基于神经网络的数据验证方法一般由服务器105执行，相应地，基于神经网络的数据验证装置一般设置于服务器105中。
应该理解，图1中的终端设备、网络和服务器的数目仅仅是示意性的。根据实现需要，可以具有任意数目的终端设备、网络和服务器。
继续参考图2，示出了本申请的基于神经网络的数据验证方法的一个实施例的流程图。所述基于神经网络的数据验证方法，用于汽车维修，包括以下步骤：
步骤201，获取用户发送的待验证的汽车维修数据，并从汽车维修数据中提取汽车的受损图片以及修理方案。
在本实施例中，用于汽车维修的基于神经网络的数据验证方法运行于其上的电子设备(例如图1所示的服务器)可以通过有线连接方式或者无线连接方式获取用户发送的待验证的汽车维修数据，并从所述汽车维修数据中提取所述汽车的受损图片以及修理方案。需要指出的是，上述无线连接方式可以包括但不限于3G/4G连接、WiFi连接、蓝牙连接、WiMAX连接、Zigbee连接、UWB(ultra wideband)连接、以及其它现在已知或将来开发的无线连接方式。
需要说明的是，所述受损图片和所述修理方案以关系型数据库的方式进行存储。通过所述受损图片可以映射寻找到所述修理方案。同样的，通过所述修理方案可以映射寻找到所述受损图片。当所述零部件包括多张所述受损图片时，每一张所述受损图片均与所述修理方案形成映射关系，或者多张所述受损图片以图片集合的形式与所述修理方案形成映射关系。
进一步地，如下表1所示，所述修理方案中包括汽车受损零件及其对应的修理方式。根据所述受损零部件的受损图片反应的定损程度，所述修理方案包括但不限于更换、钣金修理、机修、电子修理、喷漆等修理方式。
表1
| 零部件 | 修理方式 |
| 喷油嘴 | 更换 |
| 底盘 | 机修 |
| 左前车门 | 钣金修理、喷漆 |
步骤202，判修理方案数据库中是否存在与修理方案一致的标准修理方案，若是，执行步骤203，否则，执行步骤205。
在本实施例中，所述标准修理方案包括存储在修理方案数据库中的同样的零部件已经实施过的合理的修理方案以及没有被实施过但合理存在的修理方案。进一步地，本实施例通过将所述修理方案数据库中的修理方案与受损汽车的修理方案进行数据匹配以查找所述修理方案对应的标准修理方案。
需要说明的是，所述修理方案数据库通过数据清洗排除不合理的修理方案。例如，排除零部件的受损程度只需简单修理但却进行了零部件替换的修理方案。
步骤203：获取所述标准修理方案对应的历史维修数据，并提取所述历史维修数据中的历史受损图片。
在本实施例中，根据所述历史受损图片和所述标准修理方案的映射关系，从所述关系型数据库中查找到对应的所述历史受损图片。
需要说明的是，所述历史受损图片指的是存储在修理方案数据库中的同样的零部件的受损图片。具体的，所述历史受损图片包括所述零部件的一张或者多张对应所述标准修理 方案的受损图片，每一张所述受损图片均与所述标准修理方案形成映射关系，或者多张所述受损图片以图片集合的形式与所述标准修理方案形成映射关系。
步骤204：通过基于神经网络的零部件受损程度分析模型分别对受损图片和历史受损图片进行受损程度分析，计算所述受损图片反应的受损程度与历史受损图片反应的受损程度的差值，并根据差值验证待验证的所述汽车维修数据是否正确。
在本实施例中，通过所述零部件受损程度分析模型分别将所述受损图片和所述历史受损图片反应的受损程度数值化，并计算所述受损图片反应的受损程度与所述历史受损图片反应的受损程度的差值，判断所述受损图片反应的受损程度与所述历史受损图片反应的受损程度的差值是否在阈值范围内，当所述差值在阈值范围内，判定所述修理方案合理，待验证的所述汽车维修数据的验证结果为正确，否则，判定所述修理方案不合理，待验证的所述汽车维修数据的验证结果为不正确。
具体的，所述零部件受损程度分析模型包括：
输入层：用于将所述受损图片和所述历史受损图片输入所述零部件受损程度分析模型中。
隐藏层：用于将所述输入层输入的所述受损图片和所述历史受损图片，对所述未受损图片、所述受损图片和所述历史受损图片进行图像分割、去均值图像归一化处理和滤波等处理，抽象出归一化处理后的所述未受损图片、所述受损图片和所述历史受损图片的局部特征。
输出层：在所述隐藏层计算得出的局部特征的基础上，通过权值矩阵重新组装所述未受损图片、受损图片和所述历史受损图片。并基于未受损图片，对所述受损图片和所述历史受损图片进行受损程度分析并数值化以计算二者受损程度的差值，针对所述差值判断所述修理方案是否合理，若所述修理方案合理，则待验证的所述汽车维修数据的验证结果为正确，否则，待验证的所述汽车维修数据的验证结果为不正确。
步骤205：通过维修数据验证模型分析所述汽车受损图片的受损程度，并根据受损程度调取对应的目标修理方案，根据目标修理方案判修理方案是否合理，若所述修理方案合理，则待验证的所述汽车维修数据的验证结果为正确，否则，待验证的所述汽车维修数据的验证结果为不正确。
在本实施例中，当所述修理方案数据库中不存在所述标准修理方案时，通过所述维修数据验证模型对所述受损图片进行受损程度分析，并根据受损程度调取对应的目标修理方案，并对所述修理方案和所述目标修理方案进行匹配，当所述目标修理方案与所述修理方案相匹配时，判定所述修理方案合理，否则，判断所述修理方案不合理。
具体的，所述维修数据验证模型包括所述零部件受损程度分析模型，通过所述零部件受损程度分析模型对零部件反应的受损程度进行分析，获得所述受损图片反应的受损程度，将其受损程度与预设的受损程度阈值进行比较，再通过所述维修数据验证模型根据比较结果调取目标修理方案。具体的，本实施例所述维修数据验证模型将所述目标修理方案简化为“修理”和“更换”。当所述受损图片反应的受损程度大于受损程度阈值时，判断所述受损图片中的零部件应采用“更换”的目标修理方案。当所述受损图片反应的受损程度小于受损程度阈值时，判断所述受损图片中的零部件应采用“修理”的目标修理方案。
进一步地，将所述目标修理方案和所述修理方案进行比对，当二者修理方案一致时，判定所述修理方案合理，否则，判定所述修理方案不合理。
需要说明的是，在本申请的一些实施方式中，对所述修理方案中记录的“钣金修理”、“机修”、“电子修理”、“喷漆”等修理方式与所述目标修理方案中的“对应”对应，当修理方案为上述修理方式时，在比对时将其修理方式用“代替”代替。
步骤206：若待验证的所述汽车维修数据的验证结果为不正确，则以所述标准修理方案或者所述目标修理方案更新所述维修数据中的所述修理方案。
在本实施例中，当待验证的所述汽车维修数据的验证结果为正确时，将所述维修数据中的修理方案保留；当待验证的所述汽车维修数据的验证结果为不正确时，基于原修理方案不合理，将所述汽车维修数据中原有的修理方案删除，并以所述标准修理方案或者所述目标修理方案作为合理方案，替代原修理方案，完成所述汽车维修数据的更新，确保所述汽车维修数据的正确性。
步骤207：向所述用户反馈验证结果。
在本实施例中，在所述汽车维修数据的验证完成后，输出所述验证结果。若所述验证结果为所述汽车维修数据正确，则保留具有原修理方案的汽车维修数据；若所述验证结果为所述汽车维修数据不正确，则输出的验证结果中包括更新后的所述汽车维修数据。进而在验证所述汽车维修数据为不正确时，提供可靠的维修数据，以确保汽车维修的质量。
本实施例通过汽车维修数据中的修理方案匹配出与之对应的标准修理方案，进而获取对应的历史受损图片，然后再通过零部件受损程度分析模型分别对所述受损图片和所述历史受损图片进行受损程度分析，根据二者受损程度的差别判断所述修理方案是否合理，或者通过维修数据验证模型根据受损程度调取对应的目标修理方案，进而根据目标修理方案判断所述修理方案是否合理，完成汽车维修数据的验证，并且，若汽车维修数据的验证结果为不正确时，以标准修理方案或者目标修理方案更新维修数据中的修理方案，不仅能验证汽车维修数据的正确性，还能在汽车维修数据不正确的情况下提供可靠的维修数据，实现克服人工评判造成的主观误差、确保汽车维修数据的正确性，避免不恰当的维修方案带来潜在的安全隐患。
请参阅图3，其示出步骤202的一个实施例的流程图。步骤202，判修理方案数据库中是否存在与修理方案一致的标准修理方案，具体包括以下步骤：
步骤301：将修理方案与修理方案数据库进行关键词比对，以获取关键词的分类，若比对成功，执行步骤302，否则，执行步骤303。
步骤302：根据分类匹配出对应的标准修理方案。
步骤303：确定修理方案数据库中不存在与所述修理方案一致的标准修理方案。
在本实施例中，所述修理方案数据库中设置关键词列表，所述关键词列表对关键词进行分类，且属于同一类的关键词集合映射给同一标准修理方案。通过所述关键词比对，查找出关键词对应的分类，以根据分类在所述修理方案数据库中检索出与所述修理方案一致的标准修理方案。否则，所述修理方案数据库中不存在所述修理方案对应的标准修理方案，需要通过所述维修数据验证模型对所述受损图片进行受损程度分析，并根据受损程度调取对应的目标修理方案，对所述目标修理方案和所述修理方案进行比对，根据比对结果判断所述修理方案是否合理，进而得到针对所述汽车维修数据是否正确的验证结果。
具体的，所述第一关键词为所述修理方案中所述受损零部件的修理方式，可直接从所述修理方式中提取所述第一关键词。
需要说明的是，在本申请的一些实施方式中，当所述修理方案中并没有记录“钣金修理”、“机修”、“电子修理”、“喷漆”等关键词，并且所述修理方案也不是“更换”时，引入关键词“修理”作为所述修理方案的关键词。
请参阅图4，图中示出步骤301的一个实施方式流程图，步骤301，将修理方案与修理方案数据库进行关键词比对，以获取关键词的分类包括以下步骤：
步骤3011：提取修理方案中的第一关键词。
步骤3012：以第一关键词为过滤条件向关键词列表发起检索，判断第一关键词是否对应关键词列表中的任意第二关键词，若是，执行步骤3013，否者，执行步骤3014。
步骤3013：获取第一关键词对应的分类。
步骤3014：将第一关键词和关键词列表输入到深度语义匹配模型中，从关键词列表中查找与第一关键词语义相匹配的第二关键词，并根据第二关键词获取分类。
在本实施例中，在对所述修理方案与修理方案数据库进行关键词比对时，首先提取所述修理方案中的第一关键词，并以所述第一关键词为过滤条件在所述修理方案数据库中检索，然后判断所述第一关键词是否与所述关键词列表中的任意第二关键词匹配，若是，获取该关键词对应的类别，然后根据该类别与标准修理方案的映射关系匹配出对应的所述标准修理方案。当所述第一关键词不在所述关键词列表时，将所述第一关键词和所述关键词列表数据输入深度语义匹配模型中，从所述关键词列表中与查找与所述第一关键词语义相匹配的第二关键词，并根据所述第二关键词对应的类别获取所述标准修理方案。
具体的，所述深度语义匹配模型包括：
输入层，所述输入层将所述修理方案中的第一关键词和所述关键词列表以字符向量(one-hot)的形式输入。
表示层，所述表示层由至少一个卷积神经网络组成。所述卷积神经网络包括：卷积层、池化层以及全连接层。其中，所述卷积层用于提取上下文特征。所述池化层用于为所述修理方案中的第一关键词和所述关键词列表找到全局的上下文特征。所述全连接层用于将所述修理方案中的第一关键词和所述关键词列表的高纬度的语义向量转化成低纬度的语义向量。
匹配层，所述匹配层用所述修理方案中的第一关键词的低纬度的语义向量与所述关键词列表各低纬度的语义向量的cosine距离来表示所述修理方案中的第一关键词与所述关键词列表各关键词的语义相似度。然后通过softmax函数将所述语义相似度转化为一个后验概率。通过所述后验概率判断所述修理方案中的第一关键词是否与所述关键词列表中的任意第二关键词的语义相匹配。
需要说明的是，在执行步骤3014后，可将所述第一关键词存储在关键词列表中，以扩展所述关键词列表，从而减少对深度语义匹配模型的依赖，提高匹配效率。
请参阅图5，示出步骤204的一个实施例的流程图。步骤204：通过零部件受损程度分析模型分别对所述受损图片和所述历史受损图片进行受损程度分析，计算所述受损图片反应的受损程度与所述历史受损图片反应的受损程度的差值，并根据差值验证待验证的所述汽车维修数据是否正确，具体包括以下步骤：
步骤501：获取所述受损部件的未受损图片，将所述受损图片、所述历史受损图片以及所述未受损图片进行图片处理，以获取标准形式的图片；
步骤502：将处理后的未受损图片、受损图片和历史受损图片分别输入到零部件受损程度分析模型中的卷积神经网络中；
步骤503：基于受损零部件的未受损图片，通过卷积神经网络分别分析识别出受损零部件的受损图片反应的受损程度以及历史受损图片反应的受损程度，并计算二者的差值；
步骤504：当差值在预设的阈值范围内时，判定修理方案合理，待验证的所述汽车维修数据的验证结果为正确，否则，判定修理方案不合理，待验证的所述汽车维修数据的验证结果为不正确。
在本实施例中，在步骤501中，所述图片处理包括图像分割、去均值和图像归一化处理，以降低图像噪声，增强图像效果。具体的，本实施例通过对所述未受损图片、受损图片和所述历史受损图片进行图像分割，以使所述未受损图片、所述受损图片和所述历史受损图片中的零部件从图片背景中分割出来。然后对分割后的所述未受损图片、所述受损图片和所述历史受损图片进行去均值，将所述未受损图片、所述受损图片和所述历史受损图片各个维度的特征都中心化为零。对去均值后的所述未受损图片、所述受损图片和所述历 史受损图片进行归一化处理，以加快训练卷积神经网络时的收敛性。具体的，本实施例利用所述未受损图片、所述受损图片和所述历史受损图片中对仿射变换具有不变性的矩来确定变换函数的参数，即利用图像的不变矩寻找一组参数使其能够消除其他变换函数对图像变换的影响，再利用所述变换函数把所述未受损图片、所述受损图片和所述历史受损图片变换为标准形式的图片。
在步骤502中，所述零部件受损程度分析模型基于零部件的正常图片(未受损图片)对所述受损图片和所述历史受损图片进行受损程度分析，并受损程度数值化，以计算二者受损程度的差值，从而根据所述差值判所述修理方案是否合理，得到所述汽车维修数据的验证结果。具体的，所述零部件受损程度分析模型通过卷积神经网络对所述受损图片和所述历史受损图片进行受损程度分析，分别判断二者与零部件的未受损图片的相似程度。
在本实施例中，零部件受损程度分析模型包括：
数据输入层，所述输入层用于实现步骤501；
卷积计算层，所述卷积计算层用于对归一化处理后的所述受损图片和所述历史受损图片分别进行滤波，抽象出归一化处理后的所述受损图片和所述历史受损图片的局部特征；
池化层，所述池化层用于降低所述局部特征的分辨率以减少计算量，并增强所述卷积神经网络的robustness(鲁棒性)；
全连接层，所述全连接层用于将所述受损图片的局部特征重新通过权值矩阵组装成完整的所述受损图片，以及将所述历史受损图片的局部特征重新通过权值矩阵组装成完整的所述历史受损图片。并基于未受损图片，对所述受损图片和所述历史受损图片进行受损程度分析并数值化以计算二者受损程度的差值，针对所述差值判断所述修理方案合理性，进而验证所述汽车维修数据是否正确。
需要说明的是，本实施例通过输入不同零部件的受损图片和未受损图片对所述卷积神经网络进行训练，使得所述卷积神经网络能够准确识别所述受损图片反应的受损程度以及所述历史受损图片反应的受损程度。当所述受损图片中的零部件受损程度越大时，反应在所述卷积神经网络中为所述受损图片与所述未受损图片的相似程度越低。同样的，当所述历史受损图片的零部件受损程度越大时，反应在所述卷积神经网络中为所述历史受损图片与所述未受损图片的相似程度越低。
本领域普通技术人员可以理解实现上述实施例方法中的全部或部分流程，是可以通过计算机程序来指令相关的硬件来完成，该计算机程序可存储于一计算机可读取存储介质中，该程序在执行时，可包括如上述各方法的实施例的流程。其中，前述的存储介质可为磁碟、光盘、只读存储记忆体(Read-Only Memory，ROM)等非易失性存储介质，或随机存储记忆体(Random Access Memory，RAM)等。
应该理解的是，虽然附图的流程图中的各个步骤按照箭头的指示依次显示，但是这些步骤并不是必然按照箭头指示的顺序依次执行。除非本文中有明确的说明，这些步骤的执行并没有严格的顺序限制，其可以以其他的顺序执行。而且，附图的流程图中的至少一部分步骤可以包括多个子步骤或者多个阶段，这些子步骤或者阶段并不必然是在同一时刻执行完成，而是可以在不同的时刻执行，其执行顺序也不必然是依次进行，而是可以与其他步骤或者其他步骤的子步骤或者阶段的至少一部分轮流或者交替地执行。
进一步参考图6，作为对上述图2所示方法的实现，本申请提供了一种基于神经网络的数据验证装置的一个实施例，该装置实施例与图2所示的方法实施例相对应，该装置具体可以应用于各种电子设备中。
如图6所示，本实施例所述的基于神经网络的数据验证装置600包括：
获取模块601，用于获取用户发送的待验证的汽车维修数据，并从所述汽车维修数据中提取所述汽车的受损图片以及修理方案；
匹配模块602，用于修理方案与修理方案数据库中的标准修理方案进行数据匹配；
标准方案处理模块603，用于在修理方案数据库中匹配出与修理方案对应的标准修理方案时，从所述修理方案数据库中获取所述标准修理方案对应的历史维修数据，并提取历史维修数据中的历史受损图片，再通过预设的零部件受损程度分析模型分别对所述受损图片和所述历史受损图片进行受损程度分析，计算所述受损图片反应的受损程度与所述历史受损图片反应的受损程度的差值，并根据所述差值验证待验证的所述汽车维修数据是否正确，若验证所述汽车维修数据为不正确，则以所述标准修理方案更新所述汽车维修数据中的所述修理方案；
非标方案处理模块604，用于在修理方案数据库中不存在与修理方案对应的标准修理方案时，通过维修数据验证模型根据受损程度调取对应的目标修理方案，并根据目标修理方案与所述修理方案验证待验证的汽车维修数据是否正确，若验证所述汽车维修数据为不正确，则以所述目标修理方案更新所述汽车维修数据中的所述修理方案；
反馈模块605，用于向所述用户输出验证结果。
在本实施例中，所述通过维修数据中的修理方案匹配出与之对应的标准修理方案，进而获取对应的历史受损图片，然后再通过零部件受损程度分析模型分别对所述受损图片和所述历史受损图片进行受损程度分析，根据二者受损程度的差别判断所述汽车维修数据是否正确，或者通过维修数据验证模型根据受损程度调取对应的目标修理方案，并根据目标修理方案与所述修理方案验证待验证的所述汽车维修数据是否正确，并且在验证待验证的所述汽车维修数据为不正确时，提供更为可靠的修理方案，确保所述汽车维修数据的正确性，不仅有利于克服人工判断造成的主观误差，为用户提供预警，还能在汽车维修数据不正确的情况下向客户可靠的修理方案，避免不恰当的维修带来的潜在安全隐患。
为解决上述技术问题，本申请实施例还提供计算机设备。具体请参阅图7，图7为本实施例计算机设备基本结构框图。
所述计算机设备7包括通过系统总线相互通信连接存储器71、处理器72、网络接口73。需要指出的是，图中仅示出了具有组件71-73的计算机设备7，但是应理解的是，并不要求实施所有示出的组件，可以替代的实施更多或者更少的组件。其中，本技术领域技术人员可以理解，这里的计算机设备是一种能够按照事先设定或存储的指令，自动进行数值计算和/或信息处理的设备，其硬件包括但不限于微处理器、专用集成电路(Application Specific Integrated Circuit，ASIC)、可编程门阵列(Field－Programmable Gate Array，FPGA)、数字处理器(Digital Signal Processor，DSP)、嵌入式设备等。
所述计算机设备可以是桌上型计算机、笔记本、掌上电脑及云端服务器等计算设备。所述计算机设备可以与用户通过键盘、鼠标、遥控器、触摸板或声控设备等方式进行人机交互。
所述存储器71至少包括一种类型的可读存储介质，所述可读存储介质包括闪存、硬盘、多媒体卡、卡型存储器(例如，SD或DX存储器等)、随机访问存储器(RAM)、静态随机访问存储器(SRAM)、只读存储器(ROM)、电可擦除可编程只读存储器(EEPROM)、可编程只读存储器(PROM)、磁性存储器、磁盘、光盘等。在一些实施例中，所述存储器71可以是所述计算机设备7的内部存储单元，例如该计算机设备7的硬盘或内存。在另一些实施例中，所述存储器71也可以是所述计算机设备7的外部存储设备，例如该计算机设备7上配备的插接式硬盘，智能存储卡(Smart Media Card,SMC)，安全数字(Secure Digital,SD)卡，闪存卡(Flash Card)等。当然，所述存储器71还可以既包括所述计算机设备7的内部存储单元也包括其外部存储设备。本实施例中，所述存储器71通常用于存储安装于所述计算机设备7的操作系统和各类应用软件，例如汽车零部件维修的基于神经网络的数据验证方法的程序代码等。此外，所述存储器71还可以用于暂时地存储已经输出或者将要输出的各 类数据。
所述处理器72在一些实施例中可以是中央处理器(Central Processing Unit，CPU)、控制器、微控制器、微处理器、或其他数据处理芯片。该处理器72通常用于控制所述计算机设备7的总体操作。本实施例中，所述处理器72用于运行所述存储器71中存储的程序代码或者处理数据，例如运行用于汽车维修的所述基于神经网络的数据验证方法的程序代码。
所述网络接口73可包括无线网络接口或有线网络接口，该网络接口73通常用于在所述计算机设备7与其他电子设备之间建立通信连接。
本申请还提供了另一种实施方式，即提供一种计算机可读存储介质，所述计算机可读存储介质存储有用于汽车维修的基于神经网络的数据验证程序，用于汽车维修的所述基于神经网络的数据验证程序可被至少一个处理器执行，以使所述至少一个处理器执行如上述的基于神经网络的数据验证方法的步骤。可选的，该计算机可读存储介质可以是非易失性的存储介质，也可以是易失性的存储介质。
通过以上的实施方式的描述，本领域的技术人员可以清楚地了解到上述实施例方法可借助软件加必需的通用硬件平台的方式来实现，当然也可以通过硬件，但很多情况下前者是更佳的实施方式。基于这样的理解，本申请的技术方案本质上或者说对现有技术做出贡献的部分可以以软件产品的形式体现出来，该计算机软件产品存储在一个存储介质(如ROM/RAM、磁碟、光盘)中，包括若干指令用以使得一台终端设备(可以是手机，计算机，服务器，空调器，或者网络设备等)执行本申请各个实施例所述的方法。
显然，以上所描述的实施例仅仅是本申请一部分实施例，而不是全部的实施例，附图中给出了本申请的较佳实施例，但并不限制本申请的专利范围。本申请可以以许多不同的形式来实现，相反地，提供这些实施例的目的是使对本申请的公开内容的理解更加透彻全面。尽管参照前述实施例对本申请进行了详细的说明，对于本领域的技术人员来而言，其依然可以对前述各具体实施方式所记载的技术方案进行修改，或者对其中部分技术特征进行等效替换。凡是利用本申请说明书及附图内容所做的等效结构，直接或间接运用在其他相关的技术领域，均同理在本申请专利保护范围之内。
Claims (20)
- 一种基于神经网络的数据验证方法，用于汽车维修，其中，包括下述步骤：接收用户发送的待验证的汽车维修数据，并提取所述汽车维修数据中的汽车受损图片以及修理方案；将所述修理方案与修理方案数据库中的标准修理方案进行数据匹配，判断所述修理方案数据库中是否存在与所述修理方案一致的标准修理方案；若存在，则从所述修理方案数据库中获取所述标准修理方案对应的历史维修数据，并提取所述历史维修数据中的历史受损图片，再通过预设的零部件受损程度分析模型分别对所述受损图片和所述历史受损图片进行受损程度分析，计算所述受损图片反应的受损程度与所述历史受损图片反应的受损程度的差值，并根据所述差值验证待验证的所述汽车维修数据是否正确；若不存在，则通过维修数据验证模型分析所述汽车受损图片的受损程度，并根据受损程度调取对应的目标修理方案，根据所述目标修理方案与所述修理方案验证待验证的所述汽车维修数据是否正确；若验证所述维修数据为不正确，则以所述标准修理方案或者所述目标修理方案更新所述维修数据中的所述修理方案；向所述用户反馈验证结果。
- 根据权利要求1所述的基于神经网络的数据验证方法，其中，所述修理方案数据库中设置关键词列表，所述关键词列表对关键词进行分类，且属于同一类的关键词集合映射给同一标准修理方案；所述将修理方案与修理方案数据库中的标准修理方案进行数据匹配，判断所述修理方案数据库中是否存在与所述修理方案一致的标准修理方案的步骤具体包括：将所述修理方案与所述修理方案数据库进行关键词比对，以获取所述关键词的分类；若比对成功，根据所述类别获取所述类别匹配对应的所述标准修理方案；若比对不成功，则确定修理方案数据库中不存在与所述修理方案一致的标准修理方案。
- 根据权利要求2所述的基于神经网络的数据验证方法，其中，所述将所述修理方案与所述修理方案数据库进行关键词比对，以获取所述关键词的分类包括以下步骤：提取所述修理方案中的第一关键词；以所述第一关键词为过滤条件在所述关键词列表中检索，判断所述第一关键词是否与所述关键词列表中的任意第二关键词匹配；若是，获取所述第一关键词对应的分类；若否，则将所述第一关键词和所述关键词列表输入到深度语义匹配模型中，从所述关键词列表中查找与所述第一关键词语义相匹配的第二关键词，并根据所述第二关键词获取分类；根据所述分类获取对应的标准修理方案。
- 根据权利要求1所述的基于神经网络的数据验证方法，其中，所述通过零部件受损程度分析模型分别对所述受损图片和所述历史受损图片进行受损程度分析，计算所述受损图片反应的受损程度与所述历史受损图片反应的受损程度的差值，并根据所述差值验证待验证的所述汽车维修数据是否正确的步骤包括：获取受损零部件的未受损图片，将所述受损图片、所述历史受损图片以及所述未受损图片进行图片处理，以获取标准形式的图片；将处理后的所述未受损图片、所述受损图片和所述历史受损图片分别输入到所述零部件受损程度分析模型中的卷积神经网络中；基于所述未受损图片，通过所述卷积神经网络分别分析识别出所述零部件的受损图片反应的受损程度以及所述历史受损图片反应的受损程度，并计算二者的差值；当所述差值在预设的阈值范围内时，判定所述修理方案通过验证。
- 根据权利要求4所述的基于神经网络的数据验证方法，其中，所述基于所述未受损图片，通过所述卷积神经网络分别分析识别出所述受损图片反应的受损程度以及所述历史受损图片反应的受损程度，并计算二者的差值的步骤包括：通过所述卷积神经网络分析所述受损图片与所述未受损图片的相似程度以及所述历史受损图片与所述未受损图片的相似程度，基于所述相似程度计算所述受损图片反应的受损程度以及所述历史受损图片反应的受损程度；将所述受损图片反应的受损程度和所述历史受损图片反应的受损程度数值化并进行差值计算。
- 根据权利要求4所述的基于神经网络的数据验证方法，其中，所述图片处理包括图像分割、去均值和图像归一化处理；所述获取所述零部件的未受损图片，将所述受损图片、所述历史受损图片以及所述未受损图片进行图片处理，以获取标准形式的图片的步骤包括：对所述未受损图片、所述受损图片和所述历史受损图片进行图像分割，以使所述受损图片和所述历史受损图片中的零部件从图片背景中分割出来；对分割后的所述未受损图片、所述受损图片和所述历史受损图片进行去均值，将所述未受损图片、所述受损图片和所述历史受损图片各个维度的特征都中心化为零；对去均值后的所述未受损图片、所述受损图片和所述历史受损图片进行归一化，利用图像的不变矩寻找并确定变换函数的参数，再利用所述变换函数把所述未受损图片、所述受损图片和所述历史受损图片变换为标准形式的图片。
- 根据权利要求1所述的基于神经网络的数据验证方法，其中，所述维修数据验证模型包括所述零部件受损程度分析模型，所述通过维修数据验证模型根据受损程度调取对应的目标修理方案，并根据所述目标修理方案与所述修理方案验证待验证的所述汽车维修数据是否正确的步骤包括：通过所述零部件受损程度分析模型分别对所述受损图片反应的受损程度进行分析，得出所述受损图片反应的受损程度；将得出的所述受损程度与预设的受损程度阈值进行比较；所述维修数据验证模型根据比较结果调取目标修理方案；将所述目标修理方案和所述修理方案进行比对，根据比对结果判断所述修理方案是否通过验证。
- 一种基于神经网络的数据验证装置，其中，包括：获取模块，用于获取用户发送的待验证的汽车维修数据，并提取所述汽车维修数据中的汽车受损图片以及修理方案；匹配模块，用于修理方案与修理方案数据库中的标准修理方案进行数据匹配，判断所述修理方案数据库中是否存在与所述修理方案一致的标准修理方案；标准方案处理模块，用于在修理方案数据库中匹配出与修理方案对应的标准修理方案时，从所述修理方案数据库中获取所述标准修理方案对应的历史维修数据，并提取所述历史维修数据中的历史受损图片，再通过预设的零部件受损程度分析模型分别对所述受损图片和所述历史受损图片进行受损程度分析，计算所述受损图片反应的受损程度与所述历史受损图片反应的受损程度的差值，并根据所述差值验证待验证的所述汽车维修数据是否正确，若验证所述汽车维修数据为不正确，则以所述标准修理方案更新所述汽车维修数据中的所述修理方案；非标方案处理模块，用于在修理方案数据库中不存在与修理方案对应的标准修理方案时，通过维修数据验证模型根据受损程度调取对应的目标修理方案，并根据所述目标修理方案与所述修理方案验证待验证的所述汽车维修数据是否正确，若验证所述汽车维修数据 为不正确，则以所述目标修理方案更新所述维修数据中的所述修理方案；反馈模块，用于向所述用户反馈验证结果。
- 一种计算机设备，其中，包括存储器和处理器，所述存储器中存储有计算机程序，所述处理器执行所述计算机程序时实现以下步骤：接收用户发送的待验证的汽车维修数据，并提取所述汽车维修数据中的汽车受损图片以及修理方案；将所述修理方案与修理方案数据库中的标准修理方案进行数据匹配，判断所述修理方案数据库中是否存在与所述修理方案一致的标准修理方案；若存在，则从所述修理方案数据库中获取所述标准修理方案对应的历史维修数据，并提取所述历史维修数据中的历史受损图片，再通过预设的零部件受损程度分析模型分别对所述受损图片和所述历史受损图片进行受损程度分析，计算所述受损图片反应的受损程度与所述历史受损图片反应的受损程度的差值，并根据所述差值验证待验证的所述汽车维修数据是否正确；若不存在，则通过维修数据验证模型分析所述汽车受损图片的受损程度，并根据受损程度调取对应的目标修理方案，根据所述目标修理方案与所述修理方案验证待验证的所述汽车维修数据是否正确；若验证所述维修数据为不正确，则以所述标准修理方案或者所述目标修理方案更新所述维修数据中的所述修理方案；向所述用户反馈验证结果。
- 根据权利要求9所述的计算机设备，其中，所述修理方案数据库中设置关键词列表，所述关键词列表对关键词进行分类，且属于同一类的关键词集合映射给同一标准修理方案；所述处理器执行所述将修理方案与修理方案数据库中的标准修理方案进行数据匹配，判断所述修理方案数据库中是否存在与所述修理方案一致的标准修理方案时，具体执行以下步骤：将所述修理方案与所述修理方案数据库进行关键词比对，以获取所述关键词的分类；若比对成功，根据所述类别获取所述类别匹配对应的所述标准修理方案；若比对不成功，则确定修理方案数据库中不存在与所述修理方案一致的标准修理方案。
- 根据权利要求10所述的计算机设备，其中，所述处理器执行所述将所述修理方案与所述修理方案数据库进行关键词比对，以获取所述关键词的分类时，具体执行以下步骤：提取所述修理方案中的第一关键词；以所述第一关键词为过滤条件在所述关键词列表中检索，判断所述第一关键词是否与所述关键词列表中的任意第二关键词匹配；若是，获取所述第一关键词对应的分类；若否，则将所述第一关键词和所述关键词列表输入到深度语义匹配模型中，从所述关键词列表中查找与所述第一关键词语义相匹配的第二关键词，并根据所述第二关键词获取分类；根据所述分类获取对应的标准修理方案。
- 根据权利要求9所述的计算机设备，其中，所述处理器执行所述通过零部件受损程度分析模型分别对所述受损图片和所述历史受损图片进行受损程度分析，计算所述受损图片反应的受损程度与所述历史受损图片反应的受损程度的差值，并根据所述差值验证待验证的所述汽车维修数据是否正确时，具体执行以下步骤：获取受损零部件的未受损图片，将所述受损图片、所述历史受损图片以及所述未受损图片进行图片处理，以获取标准形式的图片；将处理后的所述未受损图片、所述受损图片和所述历史受损图片分别输入到所述零部 件受损程度分析模型中的卷积神经网络中；基于所述未受损图片，通过所述卷积神经网络分别分析识别出所述零部件的受损图片反应的受损程度以及所述历史受损图片反应的受损程度，并计算二者的差值；当所述差值在预设的阈值范围内时，判定所述修理方案通过验证。
- 根据权利要求12所述的计算机设备，其中，所述图片处理包括图像分割、去均值和图像归一化处理；所述处理器执行所述获取所述零部件的未受损图片，将所述受损图片、所述历史受损图片以及所述未受损图片进行图片处理，以获取标准形式的图片时，具体执行以下步骤：对所述未受损图片、所述受损图片和所述历史受损图片进行图像分割，以使所述受损图片和所述历史受损图片中的零部件从图片背景中分割出来；对分割后的所述未受损图片、所述受损图片和所述历史受损图片进行去均值，将所述未受损图片、所述受损图片和所述历史受损图片各个维度的特征都中心化为零；对去均值后的所述未受损图片、所述受损图片和所述历史受损图片进行归一化，利用图像的不变矩寻找并确定变换函数的参数，再利用所述变换函数把所述未受损图片、所述受损图片和所述历史受损图片变换为标准形式的图片。
- 根据权利要求12所述的基于神经网络的数据验证方法，其中，所述基于所述未受损图片，通过所述卷积神经网络分别分析识别出所述受损图片反应的受损程度以及所述历史受损图片反应的受损程度，并计算二者的差值的步骤包括：通过所述卷积神经网络分析所述受损图片与所述未受损图片的相似程度以及所述历史受损图片与所述未受损图片的相似程度，基于所述相似程度计算所述受损图片反应的受损程度以及所述历史受损图片反应的受损程度；将所述受损图片反应的受损程度和所述历史受损图片反应的受损程度数值化并进行差值计算。
- 根据权利要求9所述的基于神经网络的数据验证方法，其中，所述维修数据验证模型包括所述零部件受损程度分析模型，所述通过维修数据验证模型根据受损程度调取对应的目标修理方案，并根据所述目标修理方案与所述修理方案验证待验证的所述汽车维修数据是否正确的步骤包括：通过所述零部件受损程度分析模型分别对所述受损图片反应的受损程度进行分析，得出所述受损图片反应的受损程度；将得出的所述受损程度与预设的受损程度阈值进行比较；所述维修数据验证模型根据比较结果调取目标修理方案；将所述目标修理方案和所述修理方案进行比对，根据比对结果判断所述修理方案是否通过验证。
- 一种计算机可读存储介质，其中，所述计算机可读存储介质上存储有计算机程序，所述计算机程序被处理器执行时实现以下步骤：接收用户发送的待验证的汽车维修数据，并提取所述汽车维修数据中的汽车受损图片以及修理方案；将所述修理方案与修理方案数据库中的标准修理方案进行数据匹配，判断所述修理方案数据库中是否存在与所述修理方案一致的标准修理方案；若存在，则从所述修理方案数据库中获取所述标准修理方案对应的历史维修数据，并提取所述历史维修数据中的历史受损图片，再通过预设的零部件受损程度分析模型分别对所述受损图片和所述历史受损图片进行受损程度分析，计算所述受损图片反应的受损程度与所述历史受损图片反应的受损程度的差值，并根据所述差值验证待验证的所述汽车维修数据是否正确；若不存在，则通过维修数据验证模型分析所述汽车受损图片的受损程度，并根据受损程度调取对应的目标修理方案，根据所述目标修理方案与所述修理方案验证待验证的所述汽车维修数据是否正确；若验证所述维修数据为不正确，则以所述标准修理方案或者所述目标修理方案更新所述维修数据中的所述修理方案；向所述用户反馈验证结果。
- 根据权利要求16所述的计算机可读存储介质，其中，所述修理方案数据库中设置关键词列表，所述关键词列表对关键词进行分类，且属于同一类的关键词集合映射给同一标准修理方案；所述将修理方案与修理方案数据库中的标准修理方案进行数据匹配，判断所述修理方案数据库中是否存在与所述修理方案一致的标准修理方案时，所述计算机程序被处理器执行时具体实现以下步骤：将所述修理方案与所述修理方案数据库进行关键词比对，以获取所述关键词的分类；若比对成功，根据所述类别获取所述类别匹配对应的所述标准修理方案；若比对不成功，则确定修理方案数据库中不存在与所述修理方案一致的标准修理方案。
- 根据权利要求17所述的计算机可读存储介质，其中，所述将所述修理方案与所述修理方案数据库进行关键词比对，以获取所述关键词的分类时，所述计算机程序被处理器执行时具体实现以下步骤：提取所述修理方案中的第一关键词；以所述第一关键词为过滤条件在所述关键词列表中检索，判断所述第一关键词是否与所述关键词列表中的任意第二关键词匹配；若是，获取所述第一关键词对应的分类；若否，则将所述第一关键词和所述关键词列表输入到深度语义匹配模型中，从所述关键词列表中查找与所述第一关键词语义相匹配的第二关键词，并根据所述第二关键词获取分类；根据所述分类获取对应的标准修理方案。
- 根据权利要求16所述的计算机可读存储介质，其中，所述通过零部件受损程度分析模型分别对所述受损图片和所述历史受损图片进行受损程度分析，计算所述受损图片反应的受损程度与所述历史受损图片反应的受损程度的差值，并根据所述差值验证待验证的所述汽车维修数据是否正确时，所述计算机程序被处理器执行时具体实现以下步骤：获取受损零部件的未受损图片，将所述受损图片、所述历史受损图片以及所述未受损图片进行图片处理，以获取标准形式的图片；将处理后的所述未受损图片、所述受损图片和所述历史受损图片分别输入到所述零部件受损程度分析模型中的卷积神经网络中；基于所述未受损图片，通过所述卷积神经网络分别分析识别出所述零部件的受损图片反应的受损程度以及所述历史受损图片反应的受损程度，并计算二者的差值；当所述差值在预设的阈值范围内时，判定所述修理方案通过验证。
- 根据权利要求16所述的计算机可读存储介质，其中，所述维修数据验证模型包括所述零部件受损程度分析模型，所述通过维修数据验证模型根据受损程度调取对应的目标修理方案，并根据所述目标修理方案与所述修理方案验证待验证的所述汽车维修数据是否正确时，所述计算机程序被处理器执行时具体实现以下步骤：通过所述零部件受损程度分析模型分别对所述受损图片反应的受损程度进行分析，得出所述受损图片反应的受损程度；将得出的所述受损程度与预设的受损程度阈值进行比较；所述维修数据验证模型根据比较结果调取目标修理方案；将所述目标修理方案和所述修理方案进行比对，根据比对结果判断所述修理方案是否通过验证。
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910860773.XA CN110781381B (zh) | 2019-09-11 | 2019-09-11 | 基于神经网络的数据验证方法、装置、设备及存储介质 |
| CN201910860773.X | 2019-09-11 |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| WO2021047211A1 true WO2021047211A1 (zh) | 2021-03-18 |
Family
ID=69383496
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| PCT/CN2020/093590 WO2021047211A1 (zh) | 2019-09-11 | 2020-05-30 | 基于神经网络的数据验证方法、装置、设备及存储介质 |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN110781381B (zh) |
| WO (1) | WO2021047211A1 (zh) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110781381B (zh) * | 2019-09-11 | 2023-09-01 | 平安科技（深圳）有限公司 | 基于神经网络的数据验证方法、装置、设备及存储介质 |
| CN111274158A (zh) * | 2020-02-27 | 2020-06-12 | 北京首汽智行科技有限公司 | 一种数据验证方法 |
| CN111444340B (zh) * | 2020-03-10 | 2023-08-11 | 腾讯科技（深圳）有限公司 | 文本分类方法、装置、设备及存储介质 |
| CN112200543A (zh) * | 2020-10-28 | 2021-01-08 | 中国华电科工集团有限公司 | 新能源工程建设的管理方法及管理系统 |
| CN112348799B (zh) * | 2020-11-11 | 2021-07-13 | 德联易控科技(北京)有限公司 | 车辆定损方法、装置、终端设备及存储介质 |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5099436A (en) * | 1988-11-03 | 1992-03-24 | Allied-Signal Inc. | Methods and apparatus for performing system fault diagnosis |
| CN104850114A (zh) * | 2014-12-19 | 2015-08-19 | 北汽福田汽车股份有限公司 | 一种车辆故障分析方法及系统 |
| CN105955243A (zh) * | 2016-06-08 | 2016-09-21 | 中国商用飞机有限责任公司 | 一种基于故障案例的飞机诊断维修方法及其装置 |
| CN108563214A (zh) * | 2018-04-25 | 2018-09-21 | 深圳市道通科技股份有限公司 | 车辆诊断方法、装置及设备 |
| CN108985466A (zh) * | 2018-06-19 | 2018-12-11 | 深圳市元征科技股份有限公司 | 一种车辆维修方法、装置及服务器 |
| CN110781381A (zh) * | 2019-09-11 | 2020-02-11 | 平安科技（深圳）有限公司 | 基于神经网络的数据验证方法、装置、设备及存储介质 |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106203644A (zh) * | 2016-08-09 | 2016-12-07 | 深圳市永兴元科技有限公司 | 车辆定损方法和装置 |
| CN107766805A (zh) * | 2017-09-29 | 2018-03-06 | 阿里巴巴集团控股有限公司 | 提升车辆定损图像识别结果的方法、装置及服务器 |
| CN108983748B (zh) * | 2018-06-27 | 2021-06-11 | 深圳市轱辘车联数据技术有限公司 | 一种车辆故障检测方法及终端设备 |
-
2019
- 2019-09-11 CN CN201910860773.XA patent/CN110781381B/zh active Active
-
2020
- 2020-05-30 WO PCT/CN2020/093590 patent/WO2021047211A1/zh active Application Filing
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5099436A (en) * | 1988-11-03 | 1992-03-24 | Allied-Signal Inc. | Methods and apparatus for performing system fault diagnosis |
| CN104850114A (zh) * | 2014-12-19 | 2015-08-19 | 北汽福田汽车股份有限公司 | 一种车辆故障分析方法及系统 |
| CN105955243A (zh) * | 2016-06-08 | 2016-09-21 | 中国商用飞机有限责任公司 | 一种基于故障案例的飞机诊断维修方法及其装置 |
| CN108563214A (zh) * | 2018-04-25 | 2018-09-21 | 深圳市道通科技股份有限公司 | 车辆诊断方法、装置及设备 |
| CN108985466A (zh) * | 2018-06-19 | 2018-12-11 | 深圳市元征科技股份有限公司 | 一种车辆维修方法、装置及服务器 |
| CN110781381A (zh) * | 2019-09-11 | 2020-02-11 | 平安科技（深圳）有限公司 | 基于神经网络的数据验证方法、装置、设备及存储介质 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110781381B (zh) | 2023-09-01 |
| CN110781381A (zh) | 2020-02-11 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| WO2021047211A1 (zh) | 基于神经网络的数据验证方法、装置、设备及存储介质 | |
| WO2021051517A1 (zh) | 基于卷积神经网络的信息检索方法、及其相关设备 | |
| CA3067326A1 (en) | Machine-learning system for servicing queries for digital content | |
| CN107730389A (zh) | 电子装置、保险产品推荐方法及计算机可读存储介质 | |
| US9116964B2 (en) | System, method and computer program product for decoding vehicle identification numbers | |
| US20220005126A1 (en) | Virtual assistant for recommendations on whether to arbitrate claims | |
| WO2021003803A1 (zh) | 数据处理方法、装置、存储介质及电子设备 | |
| CN112101520A (zh) | 风险评估模型训练方法、业务风险评估方法及其他设备 | |
| CN112765003B (zh) | 一种基于app行为日志的风险预测方法 | |
| WO2022062968A1 (zh) | 一种自训练方法、系统、装置、电子设备及存储介质 | |
| CN107229614A (zh) | 用于分类数据的方法和装置 | |
| CN118132750A (zh) | 电力行业客户服务数据的处理方法及装置 | |
| CN110276013A (zh) | 一种维修技师的推荐方法、装置及存储介质 | |
| CN117786656A (zh) | 一种api识别方法、装置、电子设备及存储介质 | |
| CN111611099A (zh) | 失联用户信息的修复方法、装置、设备及存储介质 | |
| CN117195319A (zh) | 保函文件电子件的验真方法、装置、电子设备和介质 | |
| US11822562B2 (en) | Unstructured text processing for geographical location determination system | |
| CN110852893A (zh) | 基于海量数据的风险识别方法、系统、设备及存储介质 | |
| CN110728582A (zh) | 信息处理的方法、装置、存储介质和处理器 | |
| CN114842198A (zh) | 车辆智能定损方法、装置、设备及存储介质 | |
| CN115050042A (zh) | 一种理赔资料录入方法、装置、计算机设备及存储介质 | |
| CN113902531A (zh) | 投标文件自动生成系统、方法、电子设备和存储介质 | |
| CN109829713B (zh) | 一种基于知识与数据共同驱动的移动支付方式识别方法 | |
| CN113486086A (zh) | 一种基于特征工程的数据挖掘方法及系统 | |
| CN111475657A (zh) | 一种显示设备、显示系统及实体对齐方法 |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| 121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 20863455 Country of ref document: EP Kind code of ref document: A1 |
|
| NENP | Non-entry into the national phase |
Ref country code: DE |
|
| 122 | Ep: pct application non-entry in european phase |
Ref document number: 20863455 Country of ref document: EP Kind code of ref document: A1 |