Picture auditing method and device, electronic equipment and storage medium

Abstract

The invention relates to a picture auditing method and device, electronic equipment and a storage medium. The picture auditing method comprises the following steps: acquiring a picture to be audited; the number of the pictures to be audited is at least two; calculating the similarity between the to-be-audited pictures by adopting a preset algorithm; wherein the preset algorithms comprise at least two similarity calculation algorithms; and determining whether the to-be-audited pictures are similar pictures or not according to the plurality of similarities among the to-be-audited pictures. In this way, the similarity between the to-be-audited pictures can be calculated through multiple algorithms, and whether the to-be-audited pictures are similar pictures or not is determined through the similarity. By adopting the picture auditing method provided by the embodiment of the invention, the picture auditing efficiency can be improved, and the accuracy and integrity of a picture auditing result can be improved.

Description

Technical Field

[0001] The present disclosure relates to the field of image processing technology, and in particular to a picture auditing method, device, electronic device and storage medium. Background Art

[0002] In the related technology, the audit of maintenance work photos is an important basis for evaluating the execution quality of maintenance personnel. Maintenance work photos refer to the check-in photos, equipment photos, etc. uploaded by maintenance personnel during the execution of their work as proof of work execution. Since maintenance work photos may be adjusted through filters, cropping, image rotation, etc., the same set of photos can be used as proof of other work orders, but the work is not actually performed. Therefore, it is necessary to audit the maintenance work photos, which are usually handled manually. However, due to the large number of maintenance work photos, manual audits are prone to misjudgment due to fatigue and other reasons, resulting in low audit efficiency and low accuracy and completeness of audit results. Summary of the invention

[0003] The present disclosure provides a method, device, electronic device and storage medium for image auditing, so as to at least solve the technical problems of low auditing efficiency, low accuracy and completeness of auditing results in related technologies. The technical solution of the present disclosure is as follows:

[0004] According to a first aspect of an embodiment of the present disclosure, there is provided a method for image auditing, comprising:

[0005] Obtaining pictures to be audited; the pictures to be audited are at least two;

[0006] Using a preset algorithm to calculate the similarity between the images to be audited; wherein the preset algorithm includes at least two similarity calculation algorithms;

[0007] According to the multiple similarities between the pictures to be audited, it is determined whether the pictures to be audited are similar pictures.

[0008] In a possible implementation manner, before using a preset algorithm to calculate the similarity between the to-be-audited pictures, the method further includes:

[0009] Determine the preset target information in each of the images to be audited by a preset AI recognition model; wherein the preset target information includes preset targets and the number of each of the preset targets;

[0010] In the case where the preset target information between the pictures to be audited is different, determining that the pictures to be audited are dissimilar pictures;

[0011] The using of a preset algorithm to calculate the similarity between the images to be audited includes:

[0012] When the preset target information between the pictures to be audited is the same, a preset algorithm is used to calculate the similarity between the pictures to be audited.

[0013] In a possible implementation, the similarity calculation algorithm includes a mean hash algorithm, a difference hash algorithm, and a perceptual hash algorithm;

[0014] The similarity between the images to be audited is calculated using a preset algorithm, including:

[0015] The mean hash algorithm, the difference hash algorithm, and the perceptual hash algorithm are respectively used to calculate the first similarity, the second similarity, and the third similarity between the pictures to be audited;

[0016] The step of determining whether the pictures to be audited are similar pictures according to the multiple similarities between the pictures to be audited includes:

[0017] According to the first similarity, the second similarity, and the third similarity, it is determined whether the image to be audited is a similar image.

[0018] In a possible implementation manner, determining whether the image to be audited is a similar image according to the first similarity, the second similarity, and the third similarity includes:

[0019] Selecting the minimum similarity among the first similarity, the second similarity, and the third similarity;

[0020] When the minimum similarity is less than a first preset similarity threshold, the image to be audited is determined to be a dissimilar image.

[0021] In a possible implementation, the image audit method further includes:

[0022] In a case where the minimum similarity is greater than or equal to the first preset similarity threshold, determining whether the minimum similarity is greater than a second preset similarity threshold;

[0023] When the minimum similarity is greater than the second preset similarity threshold, the image to be audited is determined to be a similar image.

[0024] In a possible implementation, the similarity calculation algorithm further includes a scale-invariant feature transformation algorithm;

[0025] The image audit method further includes:

[0026] When the minimum similarity is less than the second preset similarity threshold, the scale-invariant feature transformation algorithm is used to calculate the fourth similarity between the pictures to be audited;

[0027] In the case where the fourth similarity is greater than the second preset similarity threshold, determining that the image to be audited is a similar image;

[0028] When the fourth similarity is less than or equal to the second preset similarity threshold, the image to be audited is determined to be a dissimilar image.

[0029] According to a second aspect of an embodiment of the present disclosure, there is provided a picture auditing device, comprising:

[0030] A picture receiving module, used to receive pictures to be audited; the pictures to be audited are at least two;

[0031] A calculation module, used to calculate the similarity between the images to be audited using a preset algorithm; wherein the preset algorithm includes at least two similarity calculation algorithms;

[0032] The audit module is used to determine whether the pictures to be audited are similar pictures according to the multiple similarities between the pictures to be audited.

[0033] According to a third aspect of an embodiment of the present disclosure, there is provided an electronic device, including:

[0034] processor;

[0035] a memory for storing instructions executable by the processor;

[0036] The processor is configured to execute the instructions to implement the image audit method as described in any one of the first aspects.

[0037] According to a fourth aspect of an embodiment of the present disclosure, a storage medium is provided. When instructions in the storage medium are executed by a processor of an electronic device, the electronic device is enabled to execute the image audit method as described in any one of the first aspects.

[0038] According to a fifth aspect of an embodiment of the present disclosure, there is provided a computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the image audit method as described in any one of the first aspects is implemented.

[0039] The technical solution provided by the embodiments of the present disclosure brings at least the following beneficial effects:

[0040] In the disclosed embodiment, pictures to be audited are obtained; there are at least two pictures to be audited; a preset algorithm is used to calculate the similarity between the pictures to be audited; wherein the preset algorithm includes at least two similarity calculation algorithms; and based on the multiple similarities between the pictures to be audited, it is determined whether the pictures to be audited are similar pictures. In this way, the similarities between the pictures to be audited can be calculated by a variety of algorithms, and whether the pictures to be audited are similar pictures can be determined by the similarities. In this way, automatic audit processing of maintenance operation pictures can be achieved through a variety of similarity algorithms. Compared with manual audit processing, this can not only effectively avoid the judgment errors that are prone to occur in manual audits, but also effectively improve the efficiency of picture audits and improve the accuracy and completeness of picture audit results; it can also be applicable to situations where the number of maintenance operation photos is large, thereby improving the applicability of picture audit methods.

[0041] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] The drawings herein are incorporated into and constitute a part of the specification, illustrate embodiments consistent with the present disclosure, and together with the description are used to explain the principles of the present disclosure, and do not constitute improper limitations on the present disclosure.

[0043] Figure 1 It is a flow chart of a method for image auditing shown in an embodiment of the present disclosure.

[0044] Figure 2 It is a flow chart of a method for image auditing shown in an embodiment of the present disclosure.

[0045] Figure 3 It is a block diagram of an execution system of a picture audit method shown in an embodiment of the present disclosure.

[0046] Figure 4 It is a block diagram of a picture auditing device shown in an embodiment of the present disclosure.

[0047] Figure 5 It is a block diagram of an electronic device shown in an embodiment of the present disclosure. DETAILED DESCRIPTION

[0048] In order to enable ordinary persons in the art to better understand the technical solutions of the present disclosure, the technical solutions in the embodiments of the present disclosure will be clearly and completely described below in conjunction with the accompanying drawings.

[0049] It should be noted that the terms "first", "second", etc. in the specification and claims of the present disclosure and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present disclosure described herein can be implemented in an order other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present disclosure. Instead, they are merely examples of devices and methods consistent with some aspects of the present disclosure as detailed in the appended claims.

[0050] It should be noted that the information (including but not limited to user device information, user personal information, etc.), data (including but not limited to data used for analysis, stored data, displayed data, etc.) and signals involved in this application are all authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant laws, regulations and standards of relevant countries and regions.

[0051] As can be seen from the background technology, in the related technology, the audit of maintenance work photos is an important basis for evaluating the execution quality of maintenance personnel. Through the audit results, the local and municipal companies can judge whether the maintenance personnel have completed the relevant maintenance work on time and in quality, and whether they have solved various problems reported by customers in a timely and accurate manner. Maintenance work photos refer to the photos of check-in and equipment uploaded by maintenance personnel through mobile phone APP (application) as proof of work execution during the execution of work. However, there is a hidden danger of fraud in the process of uploading work photos: maintenance personnel adjust the same set of photos by filters, cropping, picture rotation, etc., as proof of other work orders, but in fact they have not performed the work. Maintenance administrators need to spend a lot of energy to audit the work photos. At present, it mainly relies on manual audit processing, and the check-in photos of the work orders are randomly checked in the maintenance management system every month to determine whether there are similar or identical photos. At present, the monthly work order data of the maintenance system reaches more than 300,000, with an average of 10 photos per work order, and the total number of work order pictures per month exceeds 3 million. Manual review is time-consuming and costly. At the same time, fake photos may appear in work orders in different regions and at different times, or even the same fake photo may appear in work orders across months. All of the above situations bring great difficulties to manual auditing: long-term uninterrupted auditing can easily lead to misjudgment due to fatigue; the division of labor between different people cannot identify all the photos that have appeared, resulting in low audit accuracy and completeness. In other words, the current solution has at least the following problems:

[0052] 1. High labor costs, time-consuming manual audits, and low audit efficiency. 2. Low audit accuracy. There are three main types of fake work photos: using filters, cropping, rotation, etc. to adjust the same photo; taking photos of the same scene from different angles (the same scene refers to the same maintenance personnel, the same clothing, and the same location); fake photos appearing on work orders of different work types and at different times. Manual audits are prone to misjudgment due to fatigue. At the same time, the number of fake photos is large and the time span may be large, making it difficult for auditors to identify them. 3. Low identification integrity. Due to the huge number of existing work orders, it is impossible to manually audit all work orders, and only spot checks can be used for auditing, which may result in omissions; manual audits are also limited by personal ability and division of labor and cooperation: the same photo may be normal within the work order range of auditor A, but it may be a fake photo within the work order range of auditor B, so the number of fake photos manually audited is far less than the number of actual fake photos.

[0053] Based on this, the disclosed embodiment provides a method, device, electronic device and storage medium for image auditing, which can intelligently identify the operation photos based on the YOLO (You Only Look Once) target detection algorithm, and combine the image similarity algorithm combination to realize automatic auditing of the maintenance operation photos (referred to as pictures). The image auditing method is dominated by improving the efficiency and accuracy of the auditing of maintenance operation photos. The key content in each business scene picture is quickly identified through the YOLO target detection algorithm, and the mean hash algorithm, the difference hash algorithm, the perceptual hash algorithm, the scale-invariant feature transform algorithm (SIFT) and other algorithms are combined to realize the full amount of maintenance operation photo auditing and improve the accuracy of the image auditing. By integrating the automatic auditing tool in the maintenance management system, the operation pictures can be audited in time, the processing time of the image auditing can be reduced, and the efficiency of the image auditing can be improved; at the same time, the present invention only needs to spend a small amount of manpower to manually intervene in the pictures that the automatic auditing tool cannot recognize, which greatly reduces the manpower cost, thereby achieving the purpose of improving the image auditing effect and customer satisfaction.

[0054] The image audit method, device, electronic device and storage medium provided by the embodiments of the present disclosure are described in detail below with reference to the accompanying drawings.

[0055] Figure 1 1 is a flow chart of a method for image auditing shown in an embodiment of the present disclosure. The method for image auditing can be applied to a maintenance management system, which can be installed in a server to execute the method for image auditing. Figure 1 As shown, the image audit method may include the following steps.

[0056] S101, obtaining the image to be audited.

[0057] Among them, there are at least two pictures to be audited.

[0058] In the embodiment of the present disclosure, when executing the image audit method, the images to be audited may be obtained first, for example, two images to be audited may be obtained. The images to be audited may include real-time uploaded maintenance audit images, or may include previously uploaded historical maintenance audit images.

[0059] S102, using a preset algorithm to calculate the similarity between the images to be audited.

[0060] The preset algorithm includes at least two similarity calculation algorithms.

[0061] In the disclosed embodiment, after obtaining the pictures to be audited, a preset algorithm can be used to calculate the similarity between the two pictures to be audited. The preset algorithm can be a preset similarity calculation method, such as a mean hash algorithm, a difference hash algorithm, a perceptual hash algorithm, and the like. In this way, by calculating the similarity between the pictures to be audited using the preset algorithm, at least two similarities between the pictures to be audited can be obtained.

[0062] S103, determining whether the pictures to be audited are similar pictures according to multiple similarities between the pictures to be audited.

[0063] In the disclosed embodiment, after at least two similarities between the pictures to be audited are calculated using a preset algorithm, it can be determined whether the pictures to be audited are similar pictures based on the multiple similarities between the pictures to be audited. For example, when the similarity between the pictures to be audited is within a certain range, it can be determined that the pictures to be audited are similar pictures. Taking the case where the pictures to be audited include maintenance work picture 1 and maintenance work picture 2 as an example, similarity 1 and similarity 2 between maintenance work picture 1 and maintenance work picture 2 can be calculated respectively by similarity calculation algorithm 1 and similarity calculation algorithm 2. It is then determined whether similarity 1 and similarity 2 meet the preset similarity conditions, for example, whether one or both of similarity 1 and similarity 2 are within the preset similarity range. If the preset similarity conditions are met, it can be considered that the pictures to be audited are similar pictures; otherwise, it can be considered that the pictures to be audited are dissimilar pictures.

[0064] In the disclosed embodiment, pictures to be audited are obtained; there are at least two pictures to be audited; a preset algorithm is used to calculate the similarity between the pictures to be audited; wherein the preset algorithm includes at least two similarity calculation algorithms; and based on the multiple similarities between the pictures to be audited, it is determined whether the pictures to be audited are similar pictures. In this way, the similarities between the pictures to be audited can be calculated by a variety of algorithms, and whether the pictures to be audited are similar pictures can be determined by the similarities. In this way, automatic audit processing of maintenance operation pictures can be achieved through a variety of similarity algorithms. Compared with manual audit processing, this can not only effectively avoid the judgment errors that are prone to occur in manual audits, but also effectively improve the efficiency of picture audits and improve the accuracy and completeness of picture audit results; it can also be applicable to situations where the number of maintenance operation photos is large, thereby improving the applicability of picture audit methods.

[0065] In a possible implementation manner, before using a preset algorithm to calculate the similarity between the images to be audited, the method further includes:

[0066] Determine the preset target information in each image to be audited through a preset AI recognition model; wherein the preset target information includes the preset targets and the number of each preset target;

[0067] In the case where the preset target information between the images to be audited is different, determining the images to be audited as dissimilar images;

[0068] The preset algorithm is used to calculate the similarity between the images to be audited, including:

[0069] When the preset target information between the images to be audited is the same, a preset algorithm is used to calculate the similarity between the images to be audited.

[0070] In the disclosed embodiment, it is also possible to first perform a preliminary comparison of the image information of the image to be audited, and only when the image information in the image to be audited is different, determine whether the image to be audited is a similar image by calculating the similarity. Exemplarily, the image information in each image to be audited can be firstly identified by a preset AI recognition model, that is, the preset target information, and the preset target information can include the preset target and the number of preset targets. Among them, the preset AI (Artificial Intelligence) recognition model can be a pre-trained model for identifying the preset target information in the image to be audited, for example, the preset AI recognition model can be obtained by training the YOLO deep learning target detection algorithm. Among them, the preset target can be formulated according to the business scenario, and can be a key target, such as asset management tags, resource equipment (such as base stations, towers and antenna feeds, baseband processing units BBU, radio frequency remote modules RRU, etc.), scene identification, maintenance personnel, etc., and specific examples can be seen in Table 1. Different AI intelligent recognition models can be constructed according to business scenarios and key target lists.

[0071] Table 1

[0072]

[0073] The target detection model, i.e., the preset AI recognition model, can be trained according to the key target types of different business scenarios. The historical maintenance operation photos of the production environment are classified according to the business scenarios to form training data sets for different preset AI recognition models. Then, the key targets in various maintenance operation photos are annotated, for example, the key target of "maintenance personnel wearing safety helmets" is marked. Then, the preset AI recognition model is trained using the YOLO deep learning target detection algorithm and the labeled maintenance operation photo training set, and the model parameters are adjusted according to the results of each recognition, so that the loss function of the trained model for identifying the key target type part meets the set threshold and the key target type recognition accuracy meets the requirements. After the training of the preset AI recognition model is completed, the weight parameter file of the preset AI recognition model for different business scenarios is obtained, which supports the recognition of key target types such as asset management tags, resource equipment, and emergency vehicles. For the audit pictures, i.e., the maintenance operation photos to be audited, for example, the safety helmets and maintenance personnel wearing safety helmets can be extracted, and the recognition result can be, for example, that a total of 2 maintenance personnel wearing safety helmets are recognized.

[0074] After determining the preset targets and the number of each preset target contained in each picture to be audited through the preset AI recognition model, it is possible to determine whether the picture information between the two pictures to be audited is the same, that is, to determine whether the preset targets and the number of each preset target contained in the two pictures to be audited are the same. If the preset targets and the number of each preset target contained in the two pictures to be audited are not the same, they are definitely not similar photos. On the contrary, if the preset targets and the number of each preset target contained in the two pictures to be audited are the same, that is, the preset target information between the pictures to be audited is the same, it is necessary to further audit and confirm whether the two pictures are similar pictures by calculating the similarity. In this way, the preset targets and the number of each preset target in the maintenance operation photos can be identified by the AI ​​recognition model first, and the two pictures can be quickly confirmed whether they are similar photos: if the number of key targets is inconsistent, they are definitely not similar photos; if the number of key targets is consistent, further audit is required to confirm whether the two are similar photos.

[0075] It is understandable that after the preset AI recognition model has finished recognizing, the recognition results can be manually sampled, and re-labeled and learned based on the images that failed to be recognized, and the newly added images that passed the audit can be used as incremental samples to supplement the training data set of the preset AI recognition model. The newly added audit result images have a positive feedback effect on the training of the preset AI recognition model. Through dynamic data training, the AI ​​recognition model can be continuously adjusted and optimized to make the recognition results more accurate.

[0076] In a possible implementation, the similarity calculation algorithm includes a mean hash algorithm, a difference hash algorithm, and a perceptual hash algorithm;

[0077] The preset algorithm is used to calculate the similarity between the images to be audited, including:

[0078] The mean hash algorithm, difference hash algorithm and perceptual hash algorithm are used respectively to calculate the first similarity, second similarity and third similarity between the images to be audited;

[0079] Determine whether the images to be audited are similar images based on multiple similarities between the images to be audited, including:

[0080] According to the first similarity, the second similarity and the third similarity, it is determined whether the image to be audited is a similar image.

[0081] In the disclosed embodiment, the similarity calculation algorithm may include a mean hash algorithm, a difference hash algorithm, a perceptual hash algorithm, or may also include other similarity calculation algorithms. When a preset algorithm is used to calculate the similarity between the pictures to be audited, a mean hash algorithm may be used to calculate the similarity between the pictures to be audited (i.e., the first similarity), a difference hash algorithm may be used to calculate the similarity between the pictures to be audited (i.e., the second similarity), and a perceptual hash algorithm may be used to calculate the similarity between the pictures to be audited (i.e., the third similarity). Then, based on the calculated first similarity, second similarity, and third similarity, it may be determined whether the pictures to be audited are similar pictures. It is understandable that the value range of the similarity may be [0,1], the closer to 0, the lower the similarity, and the closer to 1, the higher the similarity.

[0082] Exemplarily, an evaluation of the image similarity calculation algorithm can be performed in advance. For example, the original image of the image and different images, images that have been cropped, filtered, rotated, subtitled, etc. can be used as algorithm inputs to calculate the similarity between each image and the original image, and record the similarity and performance (such as time consumption). A specific example can be shown in Table 2. It can be seen that for the four types of images of cropping, filtering, original images, and rotation, the greater the similarity, the higher the accuracy of the algorithm. For different images, the smaller the similarity, the higher the accuracy of the algorithm. For non-rotated scenes, the performance and accuracy of the three algorithms, the mean hash algorithm, the difference hash algorithm, and the perceptual hash algorithm, are all high. The accuracy of the SIFT algorithm is higher in the rotation scene, but the performance of the SIFT algorithm is poor. Therefore, the mean hash algorithm, the difference hash algorithm, and the perceptual hash algorithm are used preferentially.

[0083] Table 2

[0084]

[0085]

[0086] In a further possible implementation, determining whether the image to be audited is a similar image according to the first similarity, the second similarity, and the third similarity includes:

[0087] Selecting the minimum similarity among the first similarity, the second similarity, and the third similarity;

[0088] When the minimum similarity is less than the first preset similarity threshold, the image to be audited is determined to be a dissimilar image.

[0089] In the disclosed embodiment, when determining whether the image to be audited is a similar image based on the first similarity, the second similarity, and the third similarity, the minimum value of the three similarities can be selected first, that is, the minimum similarity among the first similarity, the second similarity, and the third similarity. Then, it can be determined whether the minimum similarity is less than a first preset similarity threshold, and the first similarity threshold can be set to 0.45, for example. If the minimum similarity among the first similarity, the second similarity, and the third similarity is less than the first preset similarity threshold, it can be determined that the images to be audited are dissimilar images. In this way, the similarity is calculated using three similarities with higher performance, and similar images are judged based on the minimum similarity, which can improve the performance and accuracy of the image audit method.

[0090] In a further possible implementation, the image audit method further includes:

[0091] When the minimum similarity is greater than or equal to the first preset similarity threshold, determining whether the minimum similarity is greater than a second preset similarity threshold;

[0092] When the minimum similarity is greater than the second preset similarity threshold, the image to be audited is determined to be a similar image.

[0093] In an embodiment of the present disclosure, if the minimum similarity among the first similarity, the second similarity, and the third similarity is greater than or equal to the first preset similarity threshold, the second preset similarity threshold can be obtained, and the second preset similarity threshold can be set to 0.75, for example. The minimum similarity among the first similarity, the second similarity, and the third similarity is compared with the second preset similarity threshold to determine whether the minimum similarity is greater than the second preset similarity threshold. If the minimum similarity is greater than the second preset similarity threshold, it can be determined that the image to be audited is a similar image. In this way, the image to be audited is determined to be a similar image only when the minimum similarity among the three similarities is greater than the second preset similarity threshold, so that the accuracy of the audit result can be further improved.

[0094] In a further possible implementation, the similarity calculation algorithm further includes a scale-invariant feature transformation algorithm. The image audit method further includes:

[0095] When the minimum similarity is less than the second preset similarity threshold, a scale-invariant feature transformation algorithm is used to calculate a fourth similarity between the images to be audited;

[0096] When the fourth similarity is greater than the second preset similarity threshold, determining that the image to be audited is a similar image;

[0097] When the fourth similarity is less than or equal to the second preset similarity threshold, the image to be audited is determined to be a dissimilar image.

[0098] In the disclosed embodiment, if the minimum similarity is less than the second preset similarity threshold, the scale-invariant feature transform (SIFT) can be used to calculate the similarity between the pictures to be audited again, that is, the fourth similarity. The fourth similarity is then compared with the second preset similarity threshold to determine whether the fourth similarity is greater than the second preset similarity threshold. If the fourth similarity is greater than the second preset similarity threshold, it can be determined that the picture to be audited is a similar picture. Conversely, if the fourth similarity is less than or equal to the second preset similarity threshold, it can be determined that the picture to be audited is a dissimilar picture. In this way, on the basis of the mean hash algorithm, the difference hash algorithm, and the perceptual hash algorithm, the scale-invariant feature transform SIFT algorithm is further combined to realize picture auditing, thereby further reducing the picture audit processing time and improving the picture audit efficiency and accuracy.

[0099] To make the method provided by the embodiment of the present disclosure clearer, the following Figure 2 The image audit method provided by the embodiment of the present disclosure is described. Figure 2 As shown, the image audit method may include the following processing:

[0100] First, confirm the preset AI recognition model according to the work order business scenario, identify and calculate the preset targets and the number of preset targets in the two pictures to be audited. If the preset targets or the number of preset targets in the two pictures to be audited are inconsistent, the two pictures to be audited are judged to be non-similar photos. If the preset targets and the number of preset targets in the two pictures to be audited are consistent, continue to use the preset algorithm (hash algorithm) combination to calculate the similarity of the pictures. The similarity value range is [0,1]. The closer to 0, the lower the similarity, and the closer to 1, the higher the similarity. For the two input pictures to be audited, first use three high-performance hash algorithms (mean hash algorithm, difference hash algorithm, perceptual hash algorithm) to calculate the similarity, and take the minimum similarity calculated by the three algorithms as the comparison result. If the minimum similarity is low (for example, less than the first preset similarity threshold of 0.45), the two pictures to be audited are judged to be dissimilar pictures, and the calculation ends. If the minimum similarity is high (for example, greater than the second preset similarity threshold of 0.75), the two pictures to be audited are judged to be similar pictures, and the calculation ends. If the minimum similarity is less than 0.75, the scale-invariant feature transform (SIFT) algorithm can be used to calculate the fourth similarity of the two images to be audited. If the fourth similarity is high (for example, greater than the second preset similarity threshold of 0.75), the two images to be audited are judged to be similar images and the calculation ends; otherwise, the two images to be audited are judged to be dissimilar images and the calculation ends. In this way, both the performance and accuracy of the solution are taken into account.

[0101] Based on this, the picture audit method provided by the embodiment of the present disclosure can be based on the method and system of AI intelligent auditing of maintenance operation photos, according to the work order business scenario, build a preset AI recognition model to calculate the preset target and quantity of the pictures to be audited, judge the similarity of the pictures to be audited, and analyze and evaluate from multiple scenes such as cropping, filtering, rotation, watermarking, etc. through a variety of hash algorithms combined with algorithms such as scale-invariant feature transformation, and then judge the similarity of the pictures to realize automatic auditing of maintenance operation photos. In this way, replacing manual auditing with automatic auditing tools can greatly reduce the time-consuming manual auditing and the large consumption of human resources, solve the problem of large workload of manual auditing and the problem of missed judgment of cross-maintenance work order auditing, reduce the management risk of maintenance, improve the management level of maintenance, thereby reducing the audit processing time, improving the audit efficiency and reducing labor costs. Moreover, the auditor does not need to actively audit, but only needs to confirm a small number of suspicious work order sign-in photos to complete the full amount of work order sign-in photos audit, fully improve the audit integrity, more effectively standardize the work of maintenance personnel, and ultimately achieve the purpose of improving customer satisfaction.

[0102] Exemplary, the execution system of the image audit method is as follows: Figure 3As shown in Figure 1, the system can include an interface layer, a business layer, and an algorithm layer. The business layer calls the algorithm layer to implement the degree of similarity calculation and provides business capabilities to the interface layer. As an example, the system optimizes performance by limiting the range of images to be compared. Since similar photos are generally only generated from work orders of the same business scenario within the same maintenance team, we limit the range of pairwise comparisons instead of calculating 100,000*100,000 similarities. For one image, we calculate it according to the maintenance team + work order type + business scenario + archived work order range within 7 days. The maintenance team is a team responsible for the maintenance of a region, generally composed of about 5 professional maintenance personnel. The work order type is the business category managed by the work order, such as wireless work orders, transmission work orders, etc. The business scenario is the specific work classification within the business category, as shown in Table 1. The number of images submitted by a maintenance team for the same business scenario per day is generally no more than 20, the maximum calculation amount is 140*100,000, and the average calculation time does not exceed 1 second. Using 5 32-core servers can achieve full work order auditing.

[0103] Based on the same inventive concept, the embodiment of the present disclosure also provides a picture auditing device, such as Figure 4 As shown, Figure 4 is a block diagram of a picture auditing device shown in an embodiment of the present disclosure. Figure 4 , the image auditing device 400 may include:

[0104] The picture receiving module 410 is used to receive pictures to be audited; the pictures to be audited are at least two pictures;

[0105] The calculation module 420 is used to calculate the similarity between the images to be audited by using a preset algorithm; wherein the preset algorithm includes at least two similarity calculation algorithms;

[0106] The audit module 430 is used to determine whether the pictures to be audited are similar pictures according to the multiple similarities between the pictures to be audited.

[0107] In a possible implementation manner, the image auditing device 400 further includes:

[0108] A target recognition module, used to determine the preset target information in each of the images to be audited through a preset AI recognition model; wherein the preset target information includes preset targets and the number of each preset target;

[0109] A first determination module, configured to determine that the pictures to be audited are dissimilar pictures when the preset target information between the pictures to be audited is different;

[0110] The calculation module 420 is used to:

[0111] When the preset target information between the pictures to be audited is the same, a preset algorithm is used to calculate the similarity between the pictures to be audited.

[0112] In a possible implementation, the similarity calculation algorithm includes a mean hash algorithm, a difference hash algorithm, and a perceptual hash algorithm;

[0113] The calculation module 420 is used to:

[0114] The mean hash algorithm, the difference hash algorithm, and the perceptual hash algorithm are respectively used to calculate the first similarity, the second similarity, and the third similarity between the pictures to be audited;

[0115] The audit module 430 is used to:

[0116] According to the first similarity, the second similarity, and the third similarity, it is determined whether the image to be audited is a similar image.

[0117] In a possible implementation, the audit module 430 includes:

[0118] A selection unit, configured to select the minimum similarity among the first similarity, the second similarity, and the third similarity;

[0119] The audit unit is used to determine that the image to be audited is a dissimilar image when the minimum similarity is less than a first preset similarity threshold.

[0120] In a possible implementation manner, the image auditing device 400 further includes:

[0121] A similarity determination module, configured to determine whether the minimum similarity is greater than a second preset similarity threshold when the minimum similarity is greater than or equal to the first preset similarity threshold;

[0122] The second determination module is used to determine that the image to be audited is a similar image when the minimum similarity is greater than the second preset similarity threshold.

[0123] In a possible implementation, the similarity calculation algorithm further includes a scale-invariant feature transformation algorithm;

[0124] The image auditing device 400 further includes:

[0125] A similarity calculation module, configured to calculate a fourth similarity between the images to be audited by using the scale-invariant feature transformation algorithm when the minimum similarity is less than the second preset similarity threshold;

[0126] A third determination module, configured to determine that the image to be audited is a similar image when the fourth similarity is greater than the second preset similarity threshold;

[0127] The fourth determination module is used to determine that the image to be audited is a dissimilar image when the fourth similarity is less than or equal to the second preset similarity threshold.

[0128] Regarding the device in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated here.

[0129] According to an embodiment of the present disclosure, the present disclosure also provides an electronic device, a storage medium and a computer program product.

[0130] Figure 5 A schematic block diagram of an example electronic device 500 that can be used to implement an embodiment of the present disclosure is shown. The electronic device 500 is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or required herein.

[0131] like Figure 5 As shown, the electronic device 500 includes a computing unit 501, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 502 or a computer program loaded from a storage unit 508 into a random access memory (RAM) 503. In the RAM 503, various programs and data required for the operation of the device 500 can also be stored. The computing unit 501, the ROM 502, and the RAM 503 are connected to each other via a bus 504. An input / output (I / O) interface 505 is also connected to the bus 504.

[0132] Multiple components in the electronic device 500 are connected to the I / O interface 505, including: an input unit 506, such as a keyboard, a mouse, etc.; an output unit 507, such as various types of displays, speakers, etc.; a storage unit 508, such as a disk, an optical disk, etc.; and a communication unit 509, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 509 allows the electronic device 500 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0133] The computing unit 501 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 501 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 501 performs the various methods and processes described above, such as the image audit method. For example, in some embodiments, the image audit method may be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as a storage unit 508. In some embodiments, part or all of the computer program may be loaded and / or installed on the electronic device 500 via the ROM 502 and / or the communication unit 509. When the computer program is loaded into the RAM 503 and executed by the computing unit 501, one or more steps of the image audit method described above may be performed. Alternatively, in other embodiments, the computing unit 501 may be configured to perform the image audit method in any other appropriate manner (e.g., by means of firmware).

[0134] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0135] The program code of the computer program product for implementing the method of the present disclosure can be written in any combination of one or more programming languages. These program codes can be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that the program code, when executed by the processor or controller, enables the functions / operations specified in the flow chart and / or block diagram to be implemented. The program code can be executed entirely on the machine, partially on the machine, partially on the machine as a stand-alone software package and partially on a remote machine, or entirely on a remote machine or server.

[0136] In the context of the present disclosure, a storage medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A storage medium may be a machine-readable signal medium or a machine-readable storage medium. Storage media may include, but are not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or equipment, or any suitable combination of the foregoing. More specific examples of storage media may include electrical connections based on one or more lines, portable computer disks, hard disks, random access memories (RAM), read-only memories (ROM), erasable programmable read-only memories (EPROM or flash memory), optical fibers, portable compact disk read-only memories (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0137] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0138] The systems and techniques described herein can be implemented in a computing system that includes backend components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes frontend components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such backend components, middleware components, or frontend components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), the Internet, and a blockchain network.

[0139] A computer system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The relationship between the client and the server is generated by computer programs running on the corresponding computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and VPS services ("Virtual Private Server", or "VPS" for short). The server may also be a server of a distributed system, or a server combined with a blockchain.

[0140] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this disclosure can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved, and this document does not limit this.

[0141] The above specific implementations do not constitute a limitation on the protection scope of the present disclosure. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present disclosure shall be included in the protection scope of the present disclosure.

Claims

1. A method for image auditing, It is characterized in that include: Obtaining pictures to be audited; the pictures to be audited are at least two; Using a preset algorithm to calculate the similarity between the images to be audited; wherein the preset algorithm includes at least two similarity calculation algorithms; According to the multiple similarities between the pictures to be audited, it is determined whether the pictures to be audited are similar pictures.

2. The image audit method according to claim 1, It is characterized in that Before using a preset algorithm to calculate the similarity between the images to be audited, the method further includes: Determine the preset target information in each of the images to be audited by a preset AI recognition model; wherein the preset target information includes preset targets and the number of each of the preset targets; In the case where the preset target information between the pictures to be audited is different, determining that the pictures to be audited are dissimilar pictures; The using of a preset algorithm to calculate the similarity between the images to be audited includes: When the preset target information between the pictures to be audited is the same, a preset algorithm is used to calculate the similarity between the pictures to be audited.

3. The image audit method according to claim 1 or 2, It is characterized in that The similarity calculation algorithm includes a mean hash algorithm, a difference hash algorithm, and a perceptual hash algorithm; The using of a preset algorithm to calculate the similarity between the images to be audited includes: The mean hash algorithm, the difference hash algorithm, and the perceptual hash algorithm are respectively used to calculate the first similarity, the second similarity, and the third similarity between the pictures to be audited; The step of determining whether the pictures to be audited are similar pictures according to the multiple similarities between the pictures to be audited includes: According to the first similarity, the second similarity, and the third similarity, it is determined whether the image to be audited is a similar image.

4. The image audit method according to claim 3, It is characterized in that The determining whether the image to be audited is a similar image according to the first similarity, the second similarity, and the third similarity includes: Selecting the minimum similarity among the first similarity, the second similarity, and the third similarity; When the minimum similarity is less than a first preset similarity threshold, the image to be audited is determined to be a dissimilar image.

5. The image audit method according to claim 4, It is characterized in that The image audit method further includes: In a case where the minimum similarity is greater than or equal to the first preset similarity threshold, determining whether the minimum similarity is greater than a second preset similarity threshold; When the minimum similarity is greater than the second preset similarity threshold, the image to be audited is determined to be a similar image.

6. The image audit method according to claim 5, It is characterized in that The similarity calculation algorithm also includes a scale-invariant feature transformation algorithm; The image audit method further includes: When the minimum similarity is less than the second preset similarity threshold, the scale-invariant feature transformation algorithm is used to calculate the fourth similarity between the pictures to be audited; In the case where the fourth similarity is greater than the second preset similarity threshold, determining that the image to be audited is a similar image; When the fourth similarity is less than or equal to the second preset similarity threshold, the image to be audited is determined to be a dissimilar image.

7. A picture auditing device, It is characterized in that include: A picture receiving module, used to receive pictures to be audited; The number of images to be audited is at least two; A calculation module, used to calculate the similarity between the images to be audited using a preset algorithm; wherein the preset algorithm includes at least two similarity calculation algorithms; The audit module is used to determine whether the pictures to be audited are similar pictures according to the multiple similarities between the pictures to be audited.

8. An electronic device, It is characterized in that include: processor; a memory for storing instructions executable by the processor; The processor is configured to execute the instructions to implement the image audit method as described in any one of claims 1 to 6.

9. A storage medium, when instructions in the storage medium are executed by a processor of an electronic device, the electronic device is enabled to execute the image audit method as described in any one of claims 1 to 6.

10. A computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the image audit method according to any one of claims 1 to 6 is implemented.

Images