A method and system for processing animation images based on big data

Abstract

The present invention relates to the field of animation image processing, and specifically to a big data-based animation image processing method and system, comprising an animation image conversion event acquisition module, a target conversion event identification module, a data set distinction set generation module, an effective data set analysis module and a matching storage module; the animation image conversion event acquisition module is used to acquire animation image conversion events generated by applying AI processing technology; the target conversion event identification module is used to identify target conversion events in the animation image conversion events corresponding to each type of image to be converted; the data set distinction set generation module is used to analyze the data set distinction sets between output animation images with user tendency deviations and output animation images without user tendency deviations in the target conversion events based on the data set to be analyzed; and the effective data set analysis module is used to analyze the effective data set of the output corresponding type of animation images generated by applying AI processing technology.

Description

Technical Field

[0001] The present invention relates to the technical field of animation image processing, and in particular to a method and system for animation image processing based on big data. Background Art

[0002] With the continuous development of artificial intelligence technology, more and more media and industries have begun to apply AI technology for production and creation. Among them, the field of animation images has gradually begun to use AI to automatically generate animation images and videos to improve production efficiency and save costs. AI automatic generation of animation images is usually based on deep learning technology. It mainly generates new animation content by analyzing and learning from a large amount of existing training data. It mainly includes two major steps. One is the establishment of a data set, and the other is model training based on the constructed data set. Therefore, in an animation image processing event, the rational construction of the data set is particularly important.

[0003] In the prior art, users use animation image processing software to convert real images into animation images. In this process, there will be differences in the animation images generated after multiple conversions of the same real image. The reason for this difference is the difference in data sets during the AI ​​conversion process, and the different images generated have different preferences for some users. When the animation images generated after multiple times still do not meet the user's preferences, the user's sense of use and favorability based on the animation processing software will decrease, thereby gradually reducing the frequency of application use. Therefore, for software developers, how to effectively capture user preferences to improve the efficiency, applicability and market enthusiasm of animation image conversion processing is worthy of further study. Summary of the invention

[0004] The purpose of the present invention is to provide a method and system for processing animation images based on big data to solve the problems raised in the above-mentioned background technology.

[0005] In order to solve the above technical problems, the present invention provides the following technical solutions: a method for processing cartoon images based on big data, comprising the following analysis steps:

[0006] Step S100: Obtaining an animation image conversion event generated by applying AI processing technology, the animation image conversion event includes an image to be converted and an output animation image, classifying the images to be converted based on image features, and identifying target conversion events in the animation image conversion event corresponding to each type of image to be converted; image features refer to features with the largest element area ratio in the image extraction process;

[0007] Step S200: the target conversion event records the animation image conversion process, extracts the data set collected each time by the AI ​​processing technology corresponding to the animation image conversion process as the data set to be analyzed, and analyzes the data set difference set between the output animation image with user tendency deviation and the output animation image without user tendency deviation in the target conversion event based on the data set to be analyzed;

[0008] Step S300: extracting a different set of data sets corresponding to target conversion event records of the same type of to-be-converted images, where the same type of to-be-converted images refer to to-be-converted images of the same type but with different image similarities, combining the to-be-analyzed data sets generated by the animation image conversion events corresponding to the different types of to-be-converted images, and analyzing and outputting a valid data set of the corresponding type of animation image generated by applying AI processing technology;

[0009] Step S400: Matching a valid data set corresponding to the image type of the image to be converted using AI processing technology in real time, and preferentially storing the image data in the valid data set into a data set required to be constructed when the AI ​​processing technology performs animation processing on the image.

[0010] Because the screening range of the data set in the AI ​​processing process is very large, there will be differences in the screening of different data with the same features, so the same image will generate different animation images after the same steps of animation processing; if it is possible to effectively understand the user's preference for animation image generation based on user feedback, and obtain the effective image data of the data set required for the same type of images after animation processing, it can not only improve the direction of the AI ​​analysis system for data set screening, but also enhance the user's positive use experience of AI animation image processing, without having to go through multiple conversion processes to achieve image selection with high user satisfaction.

[0011] Furthermore, step S100 includes the following specific steps:

[0012] Step S110: in each type of images to be converted, mark the same image to be converted to generate m animation image conversion events for outputting animation images and construct an investigation conversion event package, m≥2; the investigation conversion event package consists of one image to be converted and m1 different output animation images, 2≤m1≤m;

[0013] Step S120: Sort the output anime images in the investigation conversion event package in the order of output time; sequentially obtain the user operation content and the user operation duration T after each output anime image is generated and displayed; when the user operation content is download, save or screenshot, the output eigenvalue is A1; when the user operation content is other operation items except download, save and screenshot, the output eigenvalue is A2, and A2 < A1; use the formula: P = k1*A + k2*T to calculate the user preference deviation index P of each output anime image in the investigation conversion event package, where A = {A1, A2}; k1 represents the reference coefficient of the eigenvalue, and k2 represents the reference coefficient of the user operation duration.

[0014] Step S130: Obtain the user preference deviation index P recorded by all output anime images in the investigation conversion event package, and use the formula: L = {[∑(P - P0) 2 / m1} 1 / 2 to calculate the discrete index L corresponding to the investigation conversion event package; where P0 represents the average value of the user preference deviation index recorded in the investigation conversion event package, set the discrete index threshold L0, and when L > L0, output the corresponding investigation conversion event package as the target event package, and the anime image conversion event in the target event package is the target conversion event.

[0015] A discrete index greater than the discrete index threshold indicates that there are significant differences in the user preference deviation indices of the output anime images recorded in the investigation conversion event package, and to a certain extent, the user preference deviation index reflects the user's preference for the generated anime images; by analyzing the conversion event package with a large preference deviation, this application can quickly and effectively lock in the AI analysis data set that causes the user preference difference. It avoids the situation where the preference is not obvious due to a small discrete index or the user is unclear about the purpose of the image conversion event, and cannot intuitively reflect the impact of the difference in the algorithm's selected data set on the user's preference.

[0016] Further, step S200 includes the following specific steps:

[0017] Step S210: Obtain the average value P0 of the user preference deviation index recorded in the target event package, output the target conversion event corresponding to a user preference deviation index greater than the average value as the target conversion event with a user preference deviation and mark it as the first conversion event, and output the target conversion event corresponding to a user preference deviation index less than or equal to the average value as the target conversion event without a user preference deviation and mark it as the second conversion event;

[0018] Step S220: Extract the data set C to be analyzed corresponding to the i-th first conversion event 1i , C 1i = {b1, b2,..., b n1} i , b1, b2,..., bn1 represents the 1st, 2nd, ...n1th image data of the dataset constructed in the first transformation event, and the dataset to be analyzed C corresponding to the jth second transformation event 2j , C 2j ={d1,d2,...,d n2} j ,d1,d2,...,d n2 represents the 1st, 2nd, ...n2th image data of the constructed data set in the second transformation event, i≤w1, j≤w2; w1 represents the total number of the first transformation events, and w2 represents the total number of the second transformation events;

[0019] When w1 and w2 are 1, calculate the first data set difference set Z1, Z1 = C 1i -(C 1i ∩C 2j );

[0020] When w1 or w2 is greater than 1, the transformation events with different numbers of transformation events corresponding to the data sets to be analyzed are intersected to obtain the common data set C. 01 , when there is a total data set C 01 When the data set is the data set corresponding to the first conversion event, use the formula: Z2=C 01 -(C 01 ∩C), when there is a total data set C 01 When the data set is the data set corresponding to the second conversion event, the formula is: Z2 = C-(C 01 ∩C), calculate the second data set distinction set Z2, C represents the data set to be analyzed corresponding to the number of transformation events;

[0021] When w1 and w2 are greater than 1, the intersection of the corresponding data sets to be analyzed for the two types of conversion events is obtained to obtain the common data set C. 01 , C 02 , using the formula: Z3=C 01 -(C 01 ∩C 02 ), calculate the third data set distinction set Z3.

[0022] Furthermore, step S300 includes the following specific steps:

[0023] Step S310: extracting the same type of to-be-converted images with different image similarities as target conversion images and forming a conversion image set, traversing each target conversion image in the conversion image set to calculate the similarity R between any two target conversion images, and sorting the corresponding two target conversion images in the form of data packets from large to small according to the similarity values ​​based on the similarity, and generating a first sequence; extracting the data set difference set in which each data packet in the first sequence records the target conversion image, and calculating the similarity index E of the data set difference set, E=e1 / e0, wherein e1 represents the number of identical image data in the data packet record of the two data set difference sets, and e0 represents the minimum value of image data in the data packet record of the two data set difference sets;

[0024] Step S320: taking the similarity R and the similarity index E corresponding to each data packet as input, calculating the linear correlation coefficient between the two, if the similarity R is correlated with the similarity index E, then taking each data set difference set of the same type of image to be converted as the preliminary selected data set corresponding to the target conversion event; if the similarity R is not correlated with the similarity index E, then outputting the intersection of the difference sets of each data set in the conversion image set as the preliminary selected data set corresponding to the type of image to be converted;

[0025] Step S330: Mark the data sets to be analyzed generated by the animation image conversion events corresponding to different types of images to be converted as reference data sets, intersect the reference data set with the preliminary selected data set to obtain a first data set, and screen out the first data set based on the preliminary selected data set to generate effective data sets for various types of animation images generated by applying AI processing technology. The first data set represents a data set that cannot reflect the features of the corresponding images of the same type, and after elimination, the precise positioning of the analysis of the same type of data sets can be increased.

[0026] Furthermore, step S400 includes the following specific steps:

[0027] When the similarity R is related to the similarity index E, the image type that matches the image feature corresponding to the real-time image to be converted is obtained, and the image with the same image type in the historical records and the highest similarity to the real-time image to be converted is retrieved as the reference image; the valid data set of the target conversion event record of the reference image is extracted as the valid data set of the real-time conversion image;

[0028] When the similarity R is not related to the similarity index E, the image type corresponding to the real-time image to be converted is obtained, and the valid data set recorded by the image type is extracted as the valid data set of the real-time converted image.

[0029] A cartoon image processing system based on big data, comprising a cartoon image conversion event acquisition module, a target conversion event identification module, a data set distinction set generation module, a valid data set analysis module and a matching storage module;

[0030] The animation image conversion event acquisition module is used to obtain animation image conversion events generated by applying AI processing technology;

[0031] The target conversion event identification module is used to identify the target conversion event in the animation image conversion event corresponding to each type of image to be converted;

[0032] The data set difference set generation module is used to analyze the data set difference set between the output animation image with user tendency deviation and the output animation image without user tendency deviation in the target conversion event based on the data set to be analyzed;

[0033] The effective data set analysis module is used to analyze and output the effective data set of the corresponding type of animation images generated by applying AI processing technology;

[0034] The matching storage module is used to match the effective data set corresponding to the image type of the image to be converted using AI processing technology in real time, and preferentially stores the image data in the effective data set into the data set required to be constructed when the AI ​​processing technology performs animation processing on the image.

[0035] Furthermore, the target conversion event identification module includes a conversion event package construction unit, a user tendency deviation index calculation unit, a discrete index calculation unit and a target conversion event output unit;

[0036] The inspection conversion event package construction unit is used to mark the animation image conversion event of the same image to be converted in each type of image to be converted to generate an output animation image and construct an inspection conversion event package;

[0037] The user tendency deviation index calculation unit is used to sort the output animation images in the inspection conversion event package according to the output time sequence; sequentially obtain the user operation content and user operation duration of each output animation image after the generation and display, and calculate the user tendency deviation index of each output animation image in the inspection conversion event package;

[0038] The discrete index calculation unit is used to obtain the user tendency deviation index of all output animation image records in the inspection conversion event package, and calculate the discrete index corresponding to the inspection conversion event package;

[0039] The target conversion event output unit is used to set a discrete index threshold. When the discrete index is greater than the discrete index threshold, the corresponding conversion event package is output as a target event package, and the animation image conversion event in the target event package is the target conversion event.

[0040] Further, the data set distinction set generation module includes a conversion event classification unit and a data set distinction set classification calculation unit;

[0041] The conversion event classification unit is used to obtain the average value of the user tendency deviation index of the target event record, output the target conversion event corresponding to the user tendency deviation index that is greater than the average value as the target conversion event with user tendency deviation and mark it as the first conversion event, and output the target conversion event corresponding to the user tendency deviation index that is less than or equal to the average value as the target conversion event without user tendency deviation and mark it as the second conversion event;

[0042] The data set difference set classification calculation unit is used to calculate the first data set difference set, the second data set difference set and the third data set difference set based on the first conversion event and the second conversion event number output.

[0043] Further, the effective data set analysis module includes a conversion image set construction unit, a first sequence generation unit, a similarity index calculation unit, a linear analysis unit and an effective data set output unit;

[0044] The conversion image set construction unit is used to extract the to-be-converted images of the same type but with different image similarities as target conversion images and form a conversion image set;

[0045] The first sequence generating unit is used to sort the corresponding two target conversion images in the form of data packets from large to small according to the similarity values ​​based on the similarity, and generate a first sequence;

[0046] The similarity index calculation unit is used to calculate the similarity index of the data set distinction set;

[0047] The linear analysis unit is used to take the similarity and similarity index corresponding to each data packet as input, calculate the linear correlation coefficient between the two to determine the relevant results, and preliminarily select the data set based on the result output;

[0048] The effective data set output unit is used to mark the data sets to be analyzed generated by the animation image conversion events corresponding to different types of images to be converted as control data sets, intersecting the control data set with the preliminary selected data set to obtain the first data set, and filtering the first data set on the basis of the preliminary selected data set to generate effective data sets for various types of animation images generated by applying AI processing technology.

[0049] Compared with the prior art, the beneficial effects achieved by the present invention are as follows: the present invention first performs original classification of the image to be converted, and reduces the error of feature analysis from different categories; and obtains multiple image conversion events that can be verified as user preferences by extracting and analyzing the animation conversion events, extracts the data set required in the conversion process based on the event, analyzes the differences in data sets in the same type of conversion events, and thus extracts the difference image data that affects the generation of preferences; and compares and analyzes the difference image data with different types of conversion events to finally output a valid data set, so that the valid data set can be given priority in the data set in real-time conversion, thereby improving user satisfaction with the image generated by the conversion of animation data; and improving the efficiency of user-based image conversion applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0051] Figure 1 It is a structural schematic diagram of a big data-based animation image processing system of the present invention. DETAILED DESCRIPTION

[0052] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0053] See also Figure 1 The present invention provides a technical solution: a method for processing cartoon images based on big data, comprising the following analysis steps:

[0054] Step S100: Obtaining an animation image conversion event generated by applying AI processing technology, the animation image conversion event includes an image to be converted and an output animation image, classifying the images to be converted based on image features, and identifying target conversion events in the animation image conversion event corresponding to each type of image to be converted; image features refer to features with the largest area ratio of extracted elements in the image extraction process; such as image features of people, animals, landscapes, etc.;

[0055] Step S200: the target conversion event records the animation image conversion process, extracts the data set collected each time by the AI ​​processing technology corresponding to the animation image conversion process as the data set to be analyzed, and analyzes the data set difference set between the output animation image with user tendency deviation and the output animation image without user tendency deviation in the target conversion event based on the data set to be analyzed;

[0056] Step S300: Extract the dataset difference set of the target transformation event records corresponding to the to-be-transformed images of the same type. The to-be-transformed images of the same type refer to the to-be-transformed images of the same type with different image similarities. Combine the to-be-analyzed dataset generated by the anime image transformation events corresponding to the to-be-transformed images of different types, and analyze and output the effective dataset for generating anime images of the corresponding type using AI processing technology;

[0057] Step S400: Match the effective dataset corresponding to the image type of the to-be-transformed image in real-time using AI processing technology, and preferentially store the image data in the effective dataset into the dataset required for the AI processing technology to perform anime processing on the image.

[0058] Because the screening range of the dataset in the AI processing process is very large, and there will be differences in the screening of the same features but different data, the same image will generate different anime images after the same steps of anime processing; if it is possible to effectively know the user's preference tendency for anime image generation based on user feedback, and obtain the effective image data of the dataset required for anime processing of the same type of images, it can not only improve the directionality of the AI analysis system for dataset screening, but also enhance the user's positive usage experience of AI anime image processing, without having to go through multiple transformation processes to select an image with a high user satisfaction.

[0059] Step S100 includes the following specific steps:

[0060] Step S110: Mark the anime image transformation events of generating m output anime images for the same to-be-transformed image in each type of to-be-transformed image and construct an inspection transformation event package, where m≥2; the inspection transformation event package consists of a to-be-transformed image and m1 different output anime images, and 2≤m1≤m;

[0061] Step S120: Sort the output anime images in the inspection transformation event package in the order of output time; sequentially obtain the user operation content and user operation duration T after each output anime image is generated and displayed; when the user operation content is download, save or screenshot, the output feature value is A1; when the user operation content is other operation items except download, save and screenshot, the output feature value is A2, and A2 < A1; use the formula: P = k1*A + k2*T to calculate the user preference deviation index P of each output anime image in the inspection transformation event package, where A = {A1, A2}; k1 represents the reference coefficient of the feature value, and k2 represents the reference coefficient of the user operation duration;

[0062] Step S130: Obtain the user preference deviation index P recorded for all output anime images in the inspection transformation event package, and use the formula: L = {[∑(P - P0) 2 / m1}1 / 2 , calculate the discrete index L corresponding to the inspection conversion event package; where P0 represents the average value of the user tendency deviation index recorded in the inspection conversion event package, set the discrete index threshold L0, and when L>L0, output the corresponding inspection conversion event package as the target event package, and the animation image conversion event in the target event package is the target conversion event.

[0063] If the discrete index is greater than the discrete index threshold, it means that there is a large difference in the user tendency deviation index of the output animation image recorded in the conversion event package, and the user tendency deviation index reflects the user's preference for the generated animation image to a certain extent; this application can quickly and effectively lock the AI ​​analysis data set that causes the difference in user preferences by analyzing the conversion event package with a large preference deviation. Avoiding the situation where the preference is not obvious due to a small discrete index or the user is unclear about the purpose of the image conversion event cannot intuitively reflect the impact of the difference in the algorithm's selection of data sets on user preferences.

[0064] Step S200 includes the following specific steps:

[0065] Step S210: Obtain the average value P0 of the user tendency deviation index recorded in the target event package, output the target conversion event corresponding to the user tendency deviation index that is greater than the average value as the target conversion event with user tendency deviation and mark it as the first conversion event, and output the target conversion event corresponding to the user tendency deviation index that is less than or equal to the average value as the target conversion event without user tendency deviation and mark it as the second conversion event;

[0066] Step S220: Extract the to-be-analyzed data set C corresponding to the i-th first conversion event 1i , C 1i ={b1,b2,...,b n1} i ,b1,b2,...,b n1 represents the 1st, 2nd, ...n1th image data of the dataset constructed in the first transformation event, and the dataset to be analyzed C corresponding to the jth second transformation event 2j , C 2j ={d1,d2,...,d n2} j ,d1,d2,...,d n2 represents the 1st, 2nd, ...n2th image data of the constructed data set in the second transformation event, i≤w1, j≤w2; w1 represents the total number of the first transformation events, and w2 represents the total number of the second transformation events;

[0067] When w1 and w2 are 1, calculate the first data set difference set Z1, Z1 = C 1i -(C 1i ∩C 2j );

[0068] When w1 or w2 is greater than 1, the transformation events with different numbers of transformation events corresponding to the data sets to be analyzed are intersected to obtain the common data set C. 01 , when there is a total data set C 01 When the data set is the data set corresponding to the first conversion event, use the formula: Z2=C 01 -(C 01 ∩C), when there is a total data set C 01 When the data set is the data set corresponding to the second conversion event, the formula is: Z2 = C-(C 01 ∩C), calculate the second data set difference set Z2, C represents the data set to be analyzed corresponding to the number of conversion events; C can be C 11 Can also be C 21 ;

[0069] When w1 and w2 are greater than 1, the intersection of the corresponding data sets to be analyzed for the two types of conversion events is obtained to obtain the common data set C. 01 , C 02 , using the formula: Z3=C 01 -(C 01 ∩C 02 ), calculate the third data set distinction set Z3.

[0070] Step S300 includes the following specific steps:

[0071] Step S310: extracting the same type of to-be-converted images with different image similarities as target conversion images and forming a conversion image set, traversing each target conversion image in the conversion image set to calculate the similarity R between any two target conversion images, and sorting the corresponding two target conversion images in the form of data packets from large to small according to the similarity values ​​based on the similarity, and generating a first sequence; extracting the data set difference set in which each data packet in the first sequence records the target conversion image, and calculating the similarity index E of the data set difference set, E=e1 / e0, wherein e1 represents the number of identical image data in the data packet record of the two data set difference sets, and e0 represents the minimum value of image data in the data packet record of the two data set difference sets;

[0072] Step S320: taking the similarity R and the similarity index E corresponding to each data packet as input, calculating the linear correlation coefficient between the two, if the similarity R is correlated with the similarity index E, generally positively correlated; then each data set difference set of the same type of image to be converted is used as the preliminary selected data set corresponding to the target conversion event; if the similarity R is not correlated with the similarity index E, then the intersection of the difference sets of each data set in the output conversion image set is the preliminary selected data set corresponding to the type of image to be converted;

[0073] As shown in the examples:

[0074] The same type of images are recorded as image 1, image 2 and image 3 to be converted;

[0075] The similarity R between image 1 and image 2 is 70%, the similarity R between image 1 and image 3 is 30%, and the similarity R between image 2 and image 3 is 65%;

[0076] Then the first sequence is generated by sorting as follows: (image 1, image 3|30%), (image 2, image 3|65%), (image 1, image 2|70%); (image 1, image 3|30%) is a data packet;

[0077] And the similarity indexes of the different sets of data sets corresponding to each data packet in the first sequence are 50%, 68%, and 80% respectively;

[0078] The linear correlation coefficient is calculated from the above data. The formula of the linear correlation coefficient is as follows:

[0079] r=(R,E)=Cov(R,E) / [Var(R)Var(E)] 1 / 2 ;

[0080] It can be known that the two are positively correlated; therefore, among the same type of images to be converted, the data set distinction sets of image 1, image 2, and image 3 can be used as the preliminary selection data set.

[0081] Step S330: Mark the data sets to be analyzed generated by the animation image conversion events corresponding to different types of images to be converted as reference data sets, intersect the reference data set with the preliminary selected data set to obtain a first data set, and screen out the first data set based on the preliminary selected data set to generate effective data sets for various types of animation images generated by applying AI processing technology. The first data set represents a data set that cannot reflect the features of the corresponding images of the same type, and after elimination, the precise positioning of the analysis of the same type of data sets can be increased.

[0082] Step S400 includes the following specific steps:

[0083] When the similarity R is related to the similarity index E, the image type that matches the image feature corresponding to the real-time image to be converted is obtained, and the image with the same image type in the historical records and the highest similarity to the real-time image to be converted is retrieved as the reference image; the valid data set of the target conversion event record of the reference image is extracted as the valid data set of the real-time conversion image;

[0084] When the similarity R is not related to the similarity index E, the image type corresponding to the real-time image to be converted is obtained, and the valid data set recorded by the image type is extracted as the valid data set of the real-time converted image.

[0085] A cartoon image processing system based on big data, comprising a cartoon image conversion event acquisition module, a target conversion event identification module, a data set distinction set generation module, a valid data set analysis module and a matching storage module;

[0086] The animation image conversion event acquisition module is used to obtain animation image conversion events generated by applying AI processing technology;

[0087] The target conversion event identification module is used to identify the target conversion event in the animation image conversion event corresponding to each type of image to be converted;

[0088] The data set difference set generation module is used to analyze the data set difference set between the output animation image with user tendency deviation and the output animation image without user tendency deviation in the target conversion event based on the data set to be analyzed;

[0089] The effective data set analysis module is used to analyze and output the effective data set of the corresponding type of animation images generated by applying AI processing technology;

[0090] The matching storage module is used to match the effective data set corresponding to the image type of the image to be converted using AI processing technology in real time, and preferentially stores the image data in the effective data set into the data set required to be constructed when the AI ​​processing technology performs animation processing on the image.

[0091] The target conversion event analysis module includes a conversion event package construction unit, a user tendency deviation index calculation unit, a discrete index calculation unit and a target conversion event output unit;

[0092] The inspection conversion event package construction unit is used to mark the animation image conversion event of the same image to be converted in each type of image to be converted to generate an output animation image and construct an inspection conversion event package;

[0093] The user tendency deviation index calculation unit is used to sort the output animation images in the inspection conversion event package according to the output time sequence; sequentially obtain the user operation content and user operation duration of each output animation image after the generation and display, and calculate the user tendency deviation index of each output animation image in the inspection conversion event package;

[0094] The discrete index calculation unit is used to obtain the user tendency deviation index of all output animation image records in the inspection conversion event package, and calculate the discrete index corresponding to the inspection conversion event package;

[0095] The target conversion event output unit is used to set a discrete index threshold. When the discrete index is greater than the discrete index threshold, the corresponding conversion event package is output as a target event package, and the animation image conversion event in the target event package is the target conversion event.

[0096] The data set distinction set generation module includes a conversion event classification unit and a data set distinction set classification calculation unit;

[0097] The conversion event classification unit is used to obtain the average value of the user tendency deviation index of the target event record, output the target conversion event corresponding to the user tendency deviation index that is greater than the average value as the target conversion event with user tendency deviation and mark it as the first conversion event, and output the target conversion event corresponding to the user tendency deviation index that is less than or equal to the average value as the target conversion event without user tendency deviation and mark it as the second conversion event;

[0098] The data set difference set classification calculation unit is used to calculate the first data set difference set, the second data set difference set and the third data set difference set based on the first conversion event and the second conversion event number output.

[0099] The effective data set analysis module includes a conversion image set construction unit, a first sequence generation unit, a similarity index calculation unit, a linear analysis unit and an effective data set output unit;

[0100] The conversion image set construction unit is used to extract the to-be-converted images of the same type but with different image similarities as target conversion images and form a conversion image set;

[0101] The first sequence generating unit is used to sort the corresponding two target conversion images in the form of data packets from large to small according to the similarity values ​​based on the similarity, and generate a first sequence;

[0102] The similarity index calculation unit is used to calculate the similarity index of the data set distinction set;

[0103] The linear analysis unit is used to take the similarity and similarity index corresponding to each data packet as input, calculate the linear correlation coefficient between the two to determine the relevant results, and preliminarily select the data set based on the result output;

[0104] The effective data set output unit is used to mark the data sets to be analyzed generated by the animation image conversion events corresponding to different types of images to be converted as control data sets, intersecting the control data set with the preliminary selected data set to obtain the first data set, and filtering the first data set on the basis of the preliminary selected data set to generate effective data sets for various types of animation images generated by applying AI processing technology.

[0105] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0106] Finally, it should be noted that the above is only an effective embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or replace some of the technical features therein by equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for processing cartoon images based on big data, characterized in that: Including the following analysis steps: Step S100: Obtain the anime image conversion event generated by applying AI processing technology. The anime image conversion event includes the image to be converted and the output anime image. Classify the image to be converted based on image features, and distinguish the target conversion event in the anime image conversion event corresponding to each type of image to be converted. The image feature refers to the feature with the largest proportion of the extraction element area during the image extraction process. Step S200: The target conversion event records the anime image conversion process. Extract the dataset collected each time by the AI processing technology corresponding to the anime image conversion process as the dataset to be analyzed. Analyze the dataset difference set between the output anime images with user preference biases and the output anime images without user preference biases in the target conversion event based on the dataset to be analyzed. Step S300: Extract the dataset difference sets recorded in the target conversion events corresponding to the same type of images to be converted. The same type of images to be converted refers to the images to be converted of the same type with different image similarities. Combine the datasets to be analyzed generated by the anime image conversion events corresponding to different types of images to be converted, and analyze the effective dataset for generating anime images of the corresponding type by applying AI processing technology. Step S400: Match the effective dataset corresponding to the image type of the image to be converted in real time using AI processing technology, and preferentially store the image data in the effective dataset into the dataset that needs to be constructed when the AI processing technology performs anime processing on the image. Step S400 includes the following specific steps: When the similarity R is related to the similarity index E, obtain the image type matching the image features of the real-time image to be converted, and retrieve the image with the highest similarity to the real-time image to be converted of the same image type in the historical record as the reference image. Extract the effective dataset recorded in the target conversion event where the reference image is located as the effective dataset of the real-time converted image. When the similarity R is not related to the similarity index E, obtain the image type corresponding to the real-time image to be converted, and extract the effective dataset recorded for the image type as the effective dataset of the real-time converted image.

2. The method for processing cartoon images based on big data according to claim 1, characterized in that: The step S100 includes the following specific steps: Step S110: Mark the anime image conversion events in which the same image to be converted generates m output anime images in each type of image to be converted and construct an inspection conversion event package, where m≥2. The inspection conversion event package consists of an image to be converted and m1 different output anime images, where 2≤m1≤m. Step S120: Sort the output anime images in the inspection conversion event package in the order of output time. Sequentially obtain the user operation content and the user operation duration T after each output anime image is generated and displayed. When the user operation content is download, save, or screenshot, the output feature value is A1. When the user operation content is other operation items except download, save, and screenshot, the output feature value is A2, and A2 < A1. Use the formula: P = k1*A + k2*T to calculate the user preference bias index P of each output anime image in the inspection conversion event package, where A = {A1, A2}; k1 represents the reference coefficient of the feature value, and k2 represents the reference coefficient of the user operation duration. Step S130: Obtain the user tendency deviation index P of all output animation image records in the inspection conversion event package, using the formula: L = {[∑(P-P0) 2 ] / m1} 1 / 2 , calculate the discrete index L corresponding to the inspection conversion event package; where P0 represents the average value of the user tendency deviation index recorded in the inspection conversion event package, set the discrete index threshold L0, and when L>L0, output the corresponding inspection conversion event package as the target event package, and the animation image conversion event in the target event package is the target conversion event.

3. The method for processing cartoon images based on big data according to claim 2, characterized in that: The step S200 includes the following specific steps: Step S210: Obtain the average value P0 of the user tendency deviation index recorded in the target event package, output the target conversion event corresponding to the user tendency deviation index that is greater than the average value as the target conversion event with user tendency deviation and mark it as the first conversion event, and output the target conversion event corresponding to the user tendency deviation index that is less than or equal to the average value as the target conversion event without user tendency deviation and mark it as the second conversion event; Step S220: Extract the to-be-analyzed data set C corresponding to the i-th first conversion event 1i , C 1i ={b1,b2,...,b n1 } i ,b1,b2,...,b n1 represents the 1st, 2nd, ...n1th image data of the dataset constructed in the first transformation event, and the dataset to be analyzed C corresponding to the jth second transformation event 2j , C 2j ={d1,d2,...,d n2 } j ,d1,d2,...,d n2 represents the 1st, 2nd, ...n2th image data of the constructed data set in the second transformation event, i≤w1, j≤w2; w1 represents the total number of the first transformation events, and w2 represents the total number of the second transformation events; When w1 and w2 are 1, calculate the first data set difference set Z1, Z1 = C 1i -(C 1i ∩C 2j ); When w1 or w2 is greater than 1, the transformation events with different numbers of transformation events corresponding to the data sets to be analyzed are intersected to obtain the common data set C. 01 , when there is a total data set C 01 When the data set is the data set corresponding to the first conversion event, use the formula: Z2=C 01 -(C 01 ∩C), when there is a total data set C 01 When the data set is the data set corresponding to the second conversion event, the formula is: Z2 = C-(C 01 ∩C), calculate the second data set distinction set Z2, C represents the data set to be analyzed corresponding to the number of transformation events; When w1 and w2 are greater than 1, the intersection of the corresponding data sets to be analyzed for the two types of conversion events is obtained to obtain the common data set C. 01 , C 02 , using the formula: Z3=C 01 -(C 01 ∩C 02 ), calculate the third data set distinction set Z3.

4. The method for processing cartoon images based on big data according to claim 3, characterized in that: The step S300 includes the following specific steps: Step S310: extracting the same type of to-be-converted images with different image similarities as target conversion images and forming a conversion image set, traversing each target conversion image in the conversion image set to calculate the similarity R between any two target conversion images, and sorting the corresponding two target conversion images in the form of data packets from large to small according to the similarity values ​​based on the similarity, and generating a first sequence; extracting the data set difference set in which each data packet in the first sequence records the target conversion image, and calculating the similarity index E of the data set difference set, E=e1 / e0, wherein e1 represents the number of identical image data in the data packet record of the two data set difference sets, and e0 represents the minimum value of image data in the data packet record of the two data set difference sets; Step S320: taking the similarity R and the similarity index E corresponding to each data packet as input, calculating the linear correlation coefficient between the two, if the similarity R is correlated with the similarity index E; then taking each data set difference set of the same type of image to be converted as the preliminary selected data set corresponding to the target conversion event; if the similarity R is not correlated with the similarity index E, then outputting the intersection of the difference sets of each data set in the conversion image set as the preliminary selected data set corresponding to the type of image to be converted; Step S330: Mark the data sets to be analyzed generated by the animation image conversion events corresponding to different types of images to be converted as control data sets, intersect the control data set with the preliminary selected data set to obtain a first data set, and screen out the first data set based on the preliminary selected data set to generate effective data sets for various types of animation images generated by applying AI processing technology.

5. A big data-based animation image processing system, such as using a big data-based animation image processing method according to any one of claims 1 to 4, characterized in that: It includes an animation image conversion event acquisition module, a target conversion event analysis module, a data set distinction set generation module, a valid data set analysis module and a matching storage module; The animation image conversion event acquisition module is used to acquire the animation image conversion event generated by applying AI processing technology; The target conversion event analysis module is used to analyze the target conversion event in the animation image conversion event corresponding to each type of image to be converted; The data set difference set generation module is used to analyze the data set difference set between the output animation image with user tendency deviation and the output animation image without user tendency deviation in the target conversion event based on the data set to be analyzed; The effective data set analysis module is used to analyze and output the effective data set of the corresponding type of animation image generated by applying AI processing technology; The matching storage module is used to match the effective data set corresponding to the image type of the image to be converted using the AI ​​processing technology in real time, and preferentially stores the image data in the effective data set into the data set required to be constructed when the AI ​​processing technology performs animation processing on the image.

6. The big data-based animation image processing system according to claim 5, characterized in that: The target conversion event analysis module includes a conversion event package construction unit, a user tendency deviation index calculation unit, a discrete index calculation unit and a target conversion event output unit; The inspection conversion event package construction unit is used to mark the same image to be converted in each type of image to be converted to generate an animated image conversion event for outputting an animated image and to construct an inspection conversion event package; The user tendency deviation index calculation unit is used to sort the output animation images in the inspection conversion event package according to the output time sequence; sequentially obtain the user operation content and user operation duration of each output animation image after the output animation image is generated and displayed, and calculate the user tendency deviation index of each output animation image in the inspection conversion event package; The discrete index calculation unit is used to obtain the user tendency deviation index of all output animation image records in the inspection conversion event package, and calculate the discrete index corresponding to the inspection conversion event package; The target conversion event output unit is used to set a discrete index threshold value. When the discrete index is greater than the discrete index threshold value, the corresponding conversion event package is output as a target event package, and the animation image conversion event in the target event package is the target conversion event.

7. The big data-based animation image processing system according to claim 6, characterized in that: The data set distinction set generation module includes a conversion event classification unit and a data set distinction set classification calculation unit; The conversion event classification unit is used to obtain the average value of the user tendency deviation index of the target event record, output the target conversion event corresponding to the user tendency deviation index that is greater than the average value as the target conversion event with user tendency deviation and mark it as the first conversion event, and output the target conversion event corresponding to the user tendency deviation index that is less than or equal to the average value as the target conversion event without user tendency deviation and mark it as the second conversion event; The data set difference set classification calculation unit is used to calculate the first data set difference set, the second data set difference set and the third data set difference set based on the first conversion event and the second conversion event number output.

8. The big data-based animation image processing system according to claim 7, characterized in that: The effective data set analysis module includes a conversion image set construction unit, a first sequence generation unit, a similarity index calculation unit, a linear analysis unit and an effective data set output unit; The conversion image set construction unit is used to extract the to-be-converted images of the same type but with different image similarities as target conversion images and form a conversion image set; The first sequence generating unit is used to sort the corresponding two target conversion images in the form of data packets from large to small according to the similarity values ​​based on the similarity, and generate a first sequence; The similarity index calculation unit is used to calculate the similarity index of the difference set of data sets; The linear analysis unit is used to take the similarity and similarity index corresponding to each data packet as input, calculate the linear correlation coefficient between the two to determine the relevant results, and output the preliminary selection of the data set based on the results; The effective data set output unit is used to mark the data sets to be analyzed generated by the animation image conversion events corresponding to different types of images to be converted as control data sets, intersecting the control data set with the preliminary selected data set to obtain a first data set, and filtering out the first data set on the basis of the preliminary selected data set to generate effective data sets of various types of animation images generated by applying AI processing technology.

Images