CN114912972A

CN114912972A - Block chain based network promotion marketing platform and method thereof

Info

Publication number: CN114912972A
Application number: CN202210356264.5A
Authority: CN
Inventors: 刘二松
Original assignee: Individual
Current assignee: Individual
Priority date: 2022-04-06
Filing date: 2022-04-06
Publication date: 2022-08-16

Abstract

The application relates to a block chain-based network promotion marketing platform and a method thereof, wherein the method comprises the following steps: when network popularization is carried out, a popularization trigger instruction triggered by a user is obtained, and when the popularization trigger instruction is matched with a preset standard trigger instruction, a to-be-popularized information loading column is generated; acquiring a current commodity picture to be promoted loaded by the information loading column to be promoted, inputting the current commodity picture to be promoted to a preset vectorization result output model, and outputting a vectorization result of the promoted commodity by the vectorization result output model; based on a block chain technology, performing Hash chain on the vectorization result of the promoted commodity in a Hash's certificate mode, and simultaneously generating a three-dimensional simulated commodity graph according to the vectorization result of the promoted commodity; and generating a network promotion instruction based on the three-dimensional simulation commodity diagram. The invention realizes high-efficiency network popularization and marketing.

Description

Block chain based network promotion marketing platform and method thereof

Technical Field

The application relates to the technical field of computers, in particular to a block chain-based network popularization marketing platform and a method thereof.

Background

The blockchain is a novel application mode of computer technologies such as distributed data storage, point-to-point transmission, a consensus mechanism and an encryption algorithm. The block chain is an important concept of bitcoin, is essentially a decentralized database, and is used as a bottom-layer technology of bitcoin, namely a series of data blocks which are generated by correlation by using a cryptographic method, wherein each data block contains information of a batch of bitcoin network transactions for verifying the validity of the information and generating a next block. Broadly, the blockchain technique is a completely new distributed infrastructure and computing approach that utilizes blockchain data structures to verify and store data, utilizes distributed node consensus algorithms to generate and update data, utilizes cryptography to secure data transmission and access, and utilizes intelligent contracts composed of automated script code to program and manipulate data.

With the development of blockchain, blockchain has been gradually applied to various fields, such as network promotion marketing, and in the invention patent with application number CN111582929A, a method and an apparatus for marketing label transaction based on blockchain are described, the method includes the following steps: A) the first node sends the marketing label encrypted by the encryption algorithm to the blockchain network through the wireless communication module so that the blockchain network verifies the authenticity of the marketing label, and when the marketing label is verified to be true, the blockchain network writes the marketing label into a first contract; B) the first node sends transaction information to the blockchain network through the wireless communication module, the transaction information comprises a first node address, a second node address, a digital signature of the marketing label and a shared score paid by the purchase marketing label, the blockchain network writes the transaction information into a second contract, and when the second node determines that the marketing label is true according to the first contract and the second contract, the shared score paid by the purchase marketing label is prestored into the second contract.

Although the invention has a plurality of data transmission modes, the invention can meet the requirements of users on diversified data transmission modes. However, in order to provide reliable data for network popularization and marketing with higher efficiency, the collected entity picture data needs to be vectorized, and then efficient data processing in the subsequent popularization process is realized more quickly, however, no corresponding technology is available in the prior art for realizing efficient network popularization and marketing.

Disclosure of Invention

Therefore, it is necessary to provide a block chain-based network promotion marketing platform and a method thereof, which can implement efficient network promotion and marketing.

The technical scheme of the invention is as follows:

a blockchain-based network promotion marketing method, the method comprising:

step S100: when network popularization is carried out, a popularization trigger instruction triggered by a user is obtained, and when the popularization trigger instruction is matched with a preset standard trigger instruction, a to-be-popularized information loading column is generated;

step S200: acquiring a current commodity picture to be promoted loaded by the information loading column to be promoted, inputting the current commodity picture to be promoted to a preset vectorization result output model, and outputting a vectorization result of the promoted commodity by the vectorization result output model;

step S300: based on a block chain technology, performing Hash chain linking on the vectorization result of the promoted commodity in a Hash-Messaging mode, and simultaneously generating a three-dimensional simulated commodity graph according to the vectorization result of the promoted commodity;

step S400: and generating a network promotion instruction based on the three-dimensional simulated commodity map, wherein the network promotion instruction is used for sending the three-dimensional simulated commodity map to a preset network promotion target address, and generating a promotion display interface at the network promotion target address after the three-dimensional simulated commodity map is sent to the network promotion target address, and the promotion display interface is used for displaying the three-dimensional simulated commodity map.

Specifically, step S300: acquiring a current commodity picture to be promoted loaded by the information loading column to be promoted, inputting the current commodity picture to be promoted to a preset vectorization result output model, and outputting a vectorization result of the promoted commodity by the vectorization result output model; before, still include:

step S310: before network promotion, performing picture processing on commodities to be promoted, and acquiring pictures of the commodities to be promoted, wherein one commodity to be promoted corresponds to at least one picture of the commodities to be promoted;

step S320: and performing model building training processing on the to-be-promoted commodity picture based on the to-be-promoted commodity picture, and generating the vectorization result output model after the model building training processing is completed.

Specifically, step S320: performing model building training processing on the to-be-promoted commodity picture based on the to-be-promoted commodity picture, and generating the vectorization result output model after the model building training processing is completed; the method specifically comprises the following steps:

step S321: acquiring promoted commodities in the to-be-promoted commodity picture as promoted commodity samples, manufacturing promoted commodity vector data sets by using salted films of the promoted commodity samples and corresponding region images, then carrying out normalization processing on the promoted commodity vector data sets, and finally carrying out data enhancement processing on the normalized promoted commodity vector data sets through vertical overturning, horizontal overturning and random rotating to obtain promoted commodity vector training sets;

step S322: building a promoted commodity mask generation model, wherein the promoted commodity mask generation model comprises a fusion network and a semantic segmentation network; the fusion network comprises a large-scale module, a medium-scale module and a small-scale module, the large-scale module comprises nine cascaded convolutional layers, and the input of the large-scale network is the promoted commodity vector training set; the input of the medium-scale module is a second convolution layer of the large-scale module, firstly pooling features obtained by the second convolution layer to obtain features reduced by one time, then inputting the features to the latter six convolution layers for convolution operation, and finally obtaining the features with the same size as the large-scale module through twice upsampling; the small-scale module comprises three cascaded convolution layers, the input of the small-scale module is the third convolution layer of the medium-scale module, the features obtained by the third convolution layer are firstly pooled to obtain the features reduced by one time, then the features are input to the three convolution layers for convolution operation, finally the features with the same size as the large-scale module are obtained through four-time up-sampling, and finally the three sets of the features with the same size of the large-scale module, the medium-scale module and the small-scale module are added to obtain the layered multi-scale features; inputting the obtained hierarchical multi-scale features into the semantic segmentation network, and outputting to obtain a mask of the promoted commodity; calculating the obtained promoted commodity mask by adopting a contour tracing algorithm to obtain an edge point set of the densely promoted commodity, then sampling the obtained edge point set of the densely promoted commodity according to a preset interval by adopting an equal-interval point sampling algorithm to obtain an edge point set of the equidistantly promoted commodity, normalizing the obtained coordinate values of the edge point set of the equidistantly promoted commodity in a mode of dividing the coordinate values by the length and the width of an image to obtain a vector point set of the promoted commodity, taking the output characteristics of the fusion network as input, selecting characteristic values of corresponding coordinate positions on the characteristics by using the coordinates of the obtained vector point set of the promoted commodity, and outputting the characteristic values of the promoted commodity vector;

step S323: constructing a vector point set iteration model, wherein the vector point set iteration model comprises a classification network and a regression network; the classification network is used for judging whether each point is a foreground point or a background point, reserving the foreground point if the foreground point is the foreground point, and deleting the background point if the background point is the background point, and comprises a classification full-link layer and a normalization index function activation layer which are cascaded; the regression network is used for coordinate point regression and comprises a regression full-link layer and a coordinate regression layer which are cascaded, the vector characteristics of the promoted commodities are firstly input into the regression full-link layer to obtain low-dimensional characteristics, then the low-dimensional characteristics are input into the coordinate regression layer, and the layer is a multilayer sensor to obtain a coordinate prediction deviant; judging whether the generalized commodity is a vector point on the generalized commodity according to the classification prediction result by combining the point coordinates of the generalized commodity vector point set, and carrying out coordinate migration through a coordinate prediction deviation value to obtain a generalized commodity prediction point; matching the promoted commodity prediction point and the real point by adopting a combined optimization algorithm, wherein the matching items comprise a category confidence coefficient and an Euclidean distance, the category confidence coefficient is from the classification network, the Euclidean distance is the Euclidean metric for calculating the distance between the real point and the promoted commodity prediction point, the higher the confidence coefficient is, the smaller the Euclidean distance is, the better the matching is, and the final matching degree is based on the weighted summation of the confidence coefficient and the Euclidean distance; after the predicted points and the real points of the promoted commodities are matched, cost function calculation is carried out, and for the classification network, a cross entropy cost function is adopted to evaluate the difference; for the regression network, evaluating the gap by adopting a regression cost function, and completing the iteration of the vector point set of the promoted commodity through model fitting and parameter iteration;

step S324: and training the model until the model converges, and storing the trained model, wherein the trained model is a vectorization result output model.

Specifically, in step S323: calculating the Euclidean measurement between the real point and the predicted point of the promoted commodity by adopting the following formula:

d＝sqrt((x1-x2)^+(y1-y2)^)；

wherein x1 and y1 are x-axis coordinates and y-axis coordinates of the real point, respectively, and x2 and y2 are x-axis coordinates and y-axis coordinates of the promoted merchandise prediction point, respectively.

Specifically, step S500: generating a network promotion instruction based on the three-dimensional simulated commodity map, wherein the network promotion instruction is used for sending the three-dimensional simulated commodity map to a preset network promotion target address and generating a promotion display interface at the network promotion target address after the three-dimensional simulated commodity map is sent to the network promotion target address, and the promotion display interface is used for displaying the three-dimensional simulated commodity map; the method specifically comprises the following steps:

step S510: generating a commodity information calling instruction based on the three-dimensional simulation commodity diagram;

step S520: calling current commodity information data corresponding to the three-dimensional simulation commodity map from a prestored commodity information database according to the commodity information calling instruction;

step S530: binding the current commodity information data with the three-dimensional simulation commodity drawing, displaying information of the current commodity information data on each part of the three-dimensional simulation commodity drawing, and then generating a displayable three-dimensional simulation commodity drawing;

step S540: and generating the network promotion instruction according to the displayable three-dimensional simulation commodity picture, wherein the network promotion instruction is used for sending the displayable three-dimensional simulation commodity picture to a preset network promotion target address, and generating a promotion display interface at the network promotion target address after the network promotion target address is sent, and the promotion display interface is used for displaying the displayable three-dimensional simulation commodity picture.

Specifically, a network popularization marketing platform based on block chains, the platform includes:

the network promotion module is used for acquiring a promotion trigger instruction triggered by a user during network promotion, and generating a to-be-promoted information loading column when the promotion trigger instruction is matched with a preset standard trigger instruction;

the column loading module is used for acquiring a current commodity picture to be promoted, which is loaded by the information to be promoted loading column, inputting the current commodity picture to be promoted to a preset vectorization result output model, and outputting a vectorization result of the promoted commodity by the vectorization result output model;

the object vector module is used for performing Hash chaining on the vectorization result of the promoted commodity in a Hash-Ching mode based on a block chain technology and generating a three-dimensional simulated commodity map according to the vectorization result of the promoted commodity;

and the promotion instruction module is used for generating a network promotion instruction based on the three-dimensional simulation commodity map, the network promotion instruction is used for sending the three-dimensional simulation commodity map to a preset network promotion target address, and generating a promotion display interface at the network promotion target address after the three-dimensional simulation commodity map is sent to the network promotion target address, and the promotion display interface is used for displaying the three-dimensional simulation commodity map.

Specifically, the object vector module includes:

the image processing module is used for performing image processing on the commodities to be promoted and acquiring images of the commodities to be promoted before network promotion, wherein one commodity to be promoted corresponds to at least one image of the commodities to be promoted;

and the result output module is used for carrying out model building training processing on the to-be-promoted commodity picture based on the to-be-promoted commodity picture and generating the vectorization result output model after the model building training processing is finished.

Specifically, the result output module includes:

the first model module is used for acquiring a promoted commodity in the to-be-promoted commodity picture as a promoted commodity sample, manufacturing a promoted commodity vector data set by using a promoted commodity sample salting film and a corresponding region image, then performing normalization processing on the promoted commodity vector data set, and finally performing data enhancement processing on the normalized promoted commodity vector data set through vertical overturning, horizontal overturning and random rotating to obtain a promoted commodity vector training set;

the second model module is used for building a promoted commodity mask generation model, and the promoted commodity mask generation model comprises a fusion network and a semantic segmentation network; the fusion network further comprises a large-scale module, a medium-scale module and a small-scale module, the large-scale module comprises nine cascaded convolutional layers, and the input of the large-scale network is the promoted commodity vector training set; the input of the mesoscale module is a second convolutional layer of the large-scale module, firstly pooling is carried out on the features obtained by the second convolutional layer to obtain the features reduced by one time, then the features are input to the following six convolutional layers to carry out convolution operation, and finally the features with the same size as the large-scale module are obtained by twice upsampling; the small-scale module comprises three cascaded convolution layers, the input of the small-scale module is the third convolution layer of the medium-scale module, firstly, pooling is carried out on the characteristics obtained by the third convolution layer to obtain characteristics reduced by one time, then, the characteristics are input to the three convolution layers to carry out convolution operation, finally, characteristics with the same size as that of the large-scale module are obtained through four-time up-sampling, and finally, three groups of characteristics with the same size of the large-scale module, the medium-scale module and the small-scale module are added to obtain layered multi-scale characteristics; inputting the obtained hierarchical multi-scale features into the semantic segmentation network, and outputting to obtain a mask of the promoted commodity; calculating the obtained promoted commodity mask by adopting a contour tracing algorithm to obtain an edge point set of the densely promoted commodity, then sampling the obtained edge point set of the densely promoted commodity according to a preset interval by adopting an equal-interval point sampling algorithm to obtain an edge point set of the equidistantly promoted commodity, normalizing the obtained coordinate values of the edge point set of the equidistantly promoted commodity in a mode of dividing the coordinate values by the length and the width of an image to obtain a vector point set of the promoted commodity, taking the output characteristics of the fusion network as input, selecting characteristic values of corresponding coordinate positions on the characteristics by using the coordinates of the obtained vector point set of the promoted commodity, and outputting the characteristic values of the promoted commodity vector;

the third model module is used for building a vector point set iteration model, and the vector point set iteration model comprises a classification network and a regression network; the classification network is used for judging whether each point is a foreground point or a background point, if the point is the foreground point, the point is reserved, if the point is the background point, the point is deleted, the classification network comprises a classification full-connection layer and a normalization index function activation layer which are cascaded, firstly, a fixed dimension of a promoted commodity vector feature is mapped to a class space with a space dimension of two through the classification full-connection layer to obtain a class feature, and then the class feature is input to the normalization index function activation layer to obtain a classification prediction result; the regression network is used for coordinate point regression and comprises a regression full-link layer and a coordinate regression layer which are cascaded, the vector characteristics of the promoted commodities are firstly input into the regression full-link layer to obtain low-dimensional characteristics, then the low-dimensional characteristics are input into the coordinate regression layer, and the layer is a multilayer sensor to obtain a coordinate prediction deviant; judging whether the vector points are vector points on the promoted commodities or not through the classification prediction result by combining point coordinates of the vector point set of the promoted commodities, and performing coordinate migration through a coordinate prediction deviation value to obtain a promoted commodity prediction point; matching the promoted commodity prediction point and the real point by adopting a combined optimization algorithm, wherein a matching item comprises a category confidence coefficient and an Euclidean distance, the category confidence coefficient is from the classification network, the Euclidean distance is an Euclidean metric for calculating the distance between the real point and the promoted commodity prediction point, the higher the confidence coefficient is, the smaller the Euclidean distance is, the better the matching is, and the final matching degree is based on the weighted summation of the confidence coefficient and the Euclidean distance; after matching the predicted points and the real points of the promoted commodities, calculating a cost function, and evaluating the difference of the classification network by adopting a cross entropy cost function; for the regression network, evaluating the gap by adopting a regression cost function, and completing the iteration of the vector point set of the promoted commodity through model fitting and parameter iteration;

and the fourth model module is used for training the models until the models are converged and storing the trained models, wherein the trained models are vectorized result output models.

Specifically, the computer device comprises a memory and a processor, wherein the memory stores a computer program, and the processor realizes the steps of the block chain-based network promotion marketing method when executing the computer program.

Specifically, a computer readable storage medium, on which a computer program is stored, the computer program, when being executed by a processor, implements the steps of the above block chain based network promotion marketing method.

The invention has the following technical effects:

1. the block chain-based network promotion marketing platform and the method thereof sequentially acquire a promotion trigger instruction triggered by a user during network promotion, and generate a to-be-promoted information loading column when the promotion trigger instruction is matched with a preset standard trigger instruction; acquiring a current commodity picture to be promoted loaded by the information loading column to be promoted, inputting the current commodity picture to be promoted to a preset vectorization result output model, and outputting a vectorization result of the promoted commodity by the vectorization result output model; based on a block chain technology, performing Hash chain on the vectorization result of the promoted commodity in a Hash's certificate mode, and simultaneously generating a three-dimensional simulated commodity graph according to the vectorization result of the promoted commodity; and generating a network promotion instruction based on the three-dimensional simulated commodity map, wherein the network promotion instruction is used for sending the three-dimensional simulated commodity map to a preset network promotion target address and generating a promotion display interface at the network promotion target address after the three-dimensional simulated commodity map is sent to the network promotion target address, and the promotion display interface is used for displaying the three-dimensional simulated commodity map, so that efficient network promotion and marketing can be realized.

2. The method comprises the steps of obtaining a promoted commodity mask through a promoted commodity mask generating model, then converting the promoted commodity mask into a promoted commodity vector point set by adopting a contour tracing algorithm and an equidistant point sampling algorithm, then performing point classification and point coordinate regression by using a vector point set iteration model, performing combined iteration on a classification network and a regression network, and finally outputting a high-precision promoted commodity vectorization result;

3. the method for extracting the mask of the high-precision promoted commodity obtains the mask of the promoted commodity by combining the hierarchical multi-scale feature fusion and the semantic segmentation, so that high-resolution information of an image can be retained to the maximum extent, and the method is favorable for extracting the mask of the high-precision promoted commodity;

4. according to the invention, the large-scale module is not provided with the pooling layer, so that the characteristic scale can be ensured not to be reduced, and the characteristic resolution is always kept at a higher level;

5. according to the method, the density of the promoted commodity edge point set obtained by calculating the promoted commodity mask by adopting the contour tracing algorithm is too high, so that the method cannot be directly applied to point set iteration, therefore, the promoted commodity edge point set is sampled by adopting an equidistant point sampling algorithm according to the preset interval to obtain the equidistant promoted commodity edge point set, so that the density of the promoted commodity edge point set can be greatly reduced, and the subsequent point set iteration is facilitated;

6. the regression network designed by the invention can correct the coordinates of the point set, so that the points can be regressed to the positions of the corner points of the promoted goods, and the position precision of the vector point set can be effectively improved;

7. the point matching is carried out by adopting a combined optimization algorithm, so that the point matching problem caused by the possibility of a plurality of goods promotion targets existing in a single to-be-promoted goods picture is solved;

8. the method processes the promoted commodity mask into the promoted commodity vector point set, and provides the vector point set iteration model to realize the promoted commodity vector point set iteration, so that a more regular promoted commodity edge is obtained, a vector polygon point set representing the promoted commodity can be obtained, and the applicability is greatly improved.

Drawings

FIG. 1 is a schematic flow chart of a block chain-based marketing method for network promotion in one embodiment;

FIG. 2 is a block diagram of a blockchain-based network promotion marketing platform in one embodiment;

FIG. 3 is a diagram illustrating an internal structure of a computer device according to an embodiment.

Detailed Description

In order to make the objects, technical solutions and advantages of the present application more apparent, the present application is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of and not restrictive on the broad application.

In one embodiment, as shown in fig. 1, there is provided a block chain-based network promotion marketing method, the method including:

specifically, when network promotion is performed, a user needs to perform a trigger operation first, at this time, the trigger operation is the promotion trigger instruction, and in order to prevent false touch, that is, to prevent false promotion due to the fact that the promotion operation is not triggered, the promotion trigger instruction needs to be compared with a preset standard trigger instruction.

And during comparison, when the promotion trigger instruction is matched with a preset standard trigger instruction, judging that the promotion trigger instruction is not a false touch, and generating a to-be-promoted information loading column.

Further, the generated information loading column to be promoted is mainly used for identifying what the specific commodity to be promoted is.

specifically, after the current commodity to be promoted is prepared, shooting processing can be performed on the current commodity to be promoted, for example, through a mobile phone or a professional camera device.

And after shooting, acquiring the current commodity picture to be promoted, and loading the current commodity picture to be promoted into the column through the information loading column to be promoted.

Then, the current commodity picture to be promoted is input to the preset vectorization result output model, so that the vectorization result output model outputs a vectorization result of the promoted commodity, namely, the generation of the vectorization result of the promoted commodity is realized.

particularly, the maintenance and the operation of an internal data system depend on the operation and the maintenance and the operation of platforms such as a data center and the like under a centralized network system through a block chain technology, so that the cost is reduced. In addition, the nodes of the block chain can participate by anyone, and each node can also verify the correctness of the recording results of other nodes while participating in the recording, so that the maintenance efficiency is improved, and the cost is reduced.

And then the vectorization result of the promoted commodity can be efficiently and accurately recorded.

And after data are recorded, generating a three-dimensional simulated commodity picture according to the vectorization result of the promoted commodity.

Step S400: generating a network promotion instruction based on the three-dimensional simulation commodity map, wherein the network promotion instruction is used for sending the three-dimensional simulation commodity map to a preset network promotion target address, and generating a promotion display interface at the network promotion target address after the network promotion instruction is sent to the network promotion target address, and the promotion display interface is used for displaying the three-dimensional simulation commodity map.

Specifically, the effect of displaying the three-dimensional simulation commodity map is achieved by generating a network promotion instruction and generating a promotion display interface.

The method comprises the steps that a promotion trigger instruction triggered by a user is obtained when network promotion is performed in sequence, and when the promotion trigger instruction is matched with a preset standard trigger instruction, a to-be-promoted information loading column is generated; acquiring a current commodity picture to be promoted loaded by the information loading column to be promoted, inputting the current commodity picture to be promoted to a preset vectorization result output model, and outputting a vectorization result of the promoted commodity by the vectorization result output model; based on a block chain technology, performing Hash chain on the vectorization result of the promoted commodity in a Hash's certificate mode, and simultaneously generating a three-dimensional simulated commodity graph according to the vectorization result of the promoted commodity; and generating a network promotion instruction based on the three-dimensional simulated commodity map, wherein the network promotion instruction is used for sending the three-dimensional simulated commodity map to a preset network promotion target address and generating a promotion display interface at the network promotion target address after the three-dimensional simulated commodity map is sent to the network promotion target address, and the promotion display interface is used for displaying the three-dimensional simulated commodity map, so that efficient network promotion and marketing can be realized.

In one embodiment, step S300: acquiring a current commodity picture to be promoted loaded by the information loading column to be promoted, inputting the current commodity picture to be promoted to a preset vectorization result output model, and outputting a vectorization result of the promoted commodity by the vectorization result output model; before, still include:

specifically, the popularized commodity is provided with a plurality of drawing pictures, and the database and the training data are stored by processing the pictures of the popularized commodity in advance.

Furthermore, after the pictures of the commodities to be promoted are obtained, efficient and rapid data obtaining is achieved.

Specifically, the model building training process is provided for realizing the building of the vectorized result output model.

In one embodiment, step S320: performing model building training processing on the to-be-promoted commodity picture based on the to-be-promoted commodity picture, and generating the vectorization result output model after the model building training processing is completed; the method specifically comprises the following steps:

step S322: building a promoted commodity mask generation model, wherein the promoted commodity mask generation model comprises a fusion network and a semantic segmentation network; the fusion network further comprises a large-scale module, a medium-scale module and a small-scale module, the large-scale module comprises nine cascaded convolutional layers, and the input of the large-scale network is the promoted commodity vector training set; the input of the mesoscale module is a second convolutional layer of the large-scale module, firstly pooling is carried out on the features obtained by the second convolutional layer to obtain the features reduced by one time, then the features are input to the following six convolutional layers to carry out convolution operation, and finally the features with the same size as the large-scale module are obtained by twice upsampling; the small-scale module comprises three cascaded convolution layers, the input of the small-scale module is the third convolution layer of the medium-scale module, the features obtained by the third convolution layer are firstly pooled to obtain the features reduced by one time, then the features are input to the three convolution layers for convolution operation, finally the features with the same size as the large-scale module are obtained through four-time up-sampling, and finally the three sets of the features with the same size of the large-scale module, the medium-scale module and the small-scale module are added to obtain the layered multi-scale features; inputting the obtained hierarchical multi-scale features into the semantic segmentation network, and outputting to obtain a promoted commodity mask; calculating the obtained promoted commodity mask by adopting a contour tracing algorithm to obtain an edge point set of the densely promoted commodity, then sampling the obtained edge point set of the densely promoted commodity according to a preset interval by adopting an equidistant point sampling algorithm to obtain an edge point set of the equidistantly promoted commodity, then normalizing the obtained coordinate values of the edge point set of the equidistantly promoted commodity in a mode of dividing the coordinate values by the length and the width of an image to obtain a vector point set of the promoted commodity, taking the output characteristics of the fusion network as input, selecting characteristic values of corresponding coordinate positions on the characteristics by using the coordinates of the obtained vector point set of the promoted commodity, and outputting to obtain vector characteristics of the promoted commodity;

specifically, a promoted commodity mask is obtained through a promoted commodity mask generation model, then the promoted commodity mask is converted into a promoted commodity vector point set through a contour tracing algorithm and an equidistant point sampling algorithm, then point classification and point coordinate regression are carried out through a vector point set iteration model, combined iteration is carried out on a classification network and a regression network, and finally a high-precision promoted commodity vectorization result is output.

In addition, the semantic segmentation network can adopt the existing semantic segmentation network.

And moreover, the method of combining hierarchical multi-scale feature fusion and semantic segmentation is adopted to obtain the promoted commodity mask, so that high-resolution information of the image can be retained to the maximum extent, and the high-precision promoted commodity mask can be extracted favorably.

Furthermore, the large-scale module does not design the pooling layer, so that the feature scale can be ensured not to be reduced, and the feature resolution can be kept at a higher level all the time.

Step S323: constructing a vector point set iteration model, wherein the vector point set iteration model comprises a classification network and a regression network; the classification network is used for judging whether each point is a foreground point or a background point, if the point is the foreground point, the point is reserved, if the point is the background point, the point is deleted, the classification network comprises a classification full-connection layer and a normalization index function activation layer which are cascaded, firstly, a fixed dimension of a promoted commodity vector feature is mapped to a class space with a space dimension of two through the classification full-connection layer to obtain a class feature, and then the class feature is input to the normalization index function activation layer to obtain a classification prediction result; the regression network is used for coordinate point regression and comprises a cascade regression full-link layer and a coordinate regression layer, wherein the vector characteristics of the promoted commodities are input into the regression full-link layer to obtain low-dimensional characteristics, and then the low-dimensional characteristics are input into the coordinate regression layer which is a multilayer perceptron to obtain a coordinate prediction deviant; judging whether the vector points are vector points on the promoted commodities or not through the classification prediction result by combining point coordinates of the vector point set of the promoted commodities, and performing coordinate migration through a coordinate prediction deviation value to obtain a promoted commodity prediction point; matching the promoted commodity prediction point and the real point by adopting a combined optimization algorithm, wherein the matching items comprise a category confidence coefficient and an Euclidean distance, the category confidence coefficient is from the classification network, the Euclidean distance is the Euclidean metric for calculating the distance between the real point and the promoted commodity prediction point, the higher the confidence coefficient is, the smaller the Euclidean distance is, the better the matching is, and the final matching degree is based on the weighted summation of the confidence coefficient and the Euclidean distance; after matching the predicted points and the real points of the promoted commodities, calculating a cost function, and evaluating the difference of the classification network by adopting a cross entropy cost function; for the regression network, evaluating the gap by adopting a regression cost function, and completing the iteration of the vector point set of the promoted commodity through model fitting and parameter iteration;

specifically, the combinatorial optimization algorithm may adopt a hungarian algorithm.

Further, the regression cost function may be an average absolute error cost function, a mean square error cost function, or other cost function.

Meanwhile, the density of the promoted commodity edge point set obtained by calculating the promoted commodity mask by adopting the contour tracing algorithm is too high, so that the method cannot be directly applied to point set iteration.

Specifically, the coordinates of the point set can be corrected through the designed regression network, so that the points can be regressed to the positions of corner points of the promoted goods, and the position precision of the vector point set can be effectively improved;

in addition, point matching is carried out by adopting a combined optimization algorithm, so that the problem of point matching caused by the fact that a plurality of goods promotion targets possibly exist in a single to-be-promoted goods picture is solved;

certainly, on one hand, the promoted commodity mask is processed into a promoted commodity vector point set, and meanwhile, a vector point set iteration model is provided to realize promoted commodity vector point set iteration, so that more regular promoted commodity edges are obtained, a polygonal point set representing the promoted commodity vector can be obtained, and the applicability is greatly improved.

Step S324: and training the model until the model is converged, and storing the trained model, wherein the trained model is a vectorization result output model.

In one embodiment, in step S323: calculating the Euclidean measurement between the real point and the predicted point of the promoted commodity by adopting the following formula:

d＝sqrt((x1-x2)^+(y1-y2)^)；

wherein x1 and y1 are x-axis coordinates and y-axis coordinates of the real point, respectively, and x2 and y2 are x-axis coordinates and y-axis coordinates of the predicted point of the promoted commodity, respectively.

In another embodiment, in step S323: calculating the euclidean metric between the true point and the predicted point of the promoted commodity may also be performed using the following formula:

d＝sqrt((x1-x2)^+(y1-y2)^+(z1-z2)^)；

wherein x1 and y1 are the x-axis coordinate and y-axis coordinate of the real point, respectively, x2 and y2 are the x-axis coordinate and y-axis coordinate of the promoted commodity prediction point, respectively, z1 is the ordinate of the real point, and z2 is the ordinate of the promoted commodity prediction point.

Further, the above formula can be adopted to calculate according to the two-dimensional space and the three-dimensional space where the real point and the promoted commodity prediction point are actually located, so that the Euclidean metric between the real point and the promoted commodity prediction point is accurately calculated.

In one embodiment, step S500: generating a network promotion instruction based on the three-dimensional simulated commodity map, wherein the network promotion instruction is used for sending the three-dimensional simulated commodity map to a preset network promotion target address and generating a promotion display interface at the network promotion target address after the three-dimensional simulated commodity map is sent to the network promotion target address, and the promotion display interface is used for displaying the three-dimensional simulated commodity map; the method specifically comprises the following steps:

step S510: generating a commodity information calling instruction based on the three-dimensional simulation commodity graph;

specifically, the commodity information calling instruction is generated through the three-dimensional simulation commodity map, and then the data can be conveniently called in the follow-up process.

specifically, the current commodity information data corresponding to the three-dimensional simulated commodity map is stored in the commodity information database in advance, and the three-dimensional simulated commodity map is finely analyzed through the establishment of the commodity information database, so that the commodity introduction effect is improved.

Step S530: binding the current commodity information data with the three-dimensional simulation commodity picture, displaying information of the current commodity information data on each part of the three-dimensional simulation commodity picture, and then generating a three-dimensional simulation commodity picture capable of being displayed;

specifically, the current commodity information data and the three-dimensional simulation commodity image are bound through binding processing, and information of the current commodity information data is displayed on each part of the three-dimensional simulation commodity image, so that high-efficiency display of commodities and corresponding commodity introduction is achieved, and display effect is improved.

In one embodiment, as shown in fig. 2, a blockchain-based network promotion marketing platform includes:

the network promotion module is used for acquiring a promotion trigger instruction triggered by a user when network promotion is carried out, and generating a to-be-promoted information loading column when the promotion trigger instruction is matched with a preset standard trigger instruction;

In one embodiment, the object vector module comprises:

In one embodiment, the result output module includes:

the second model module is used for building a promoted commodity mask generation model, and the promoted commodity mask generation model comprises a fusion network and a semantic segmentation network; the fusion network further comprises a large-scale module, a medium-scale module and a small-scale module, the large-scale module comprises nine cascaded convolutional layers, and the input of the large-scale network is the promoted commodity vector training set; the input of the mesoscale module is a second convolutional layer of the large-scale module, firstly pooling is carried out on the features obtained by the second convolutional layer to obtain the features reduced by one time, then the features are input to the following six convolutional layers to carry out convolution operation, and finally the features with the same size as the large-scale module are obtained by twice upsampling; the small-scale module comprises three cascaded convolution layers, the input of the small-scale module is the third convolution layer of the medium-scale module, the features obtained by the third convolution layer are firstly pooled to obtain the features reduced by one time, then the features are input to the three convolution layers for convolution operation, finally the features with the same size as the large-scale module are obtained through four-time up-sampling, and finally the three sets of the features with the same size of the large-scale module, the medium-scale module and the small-scale module are added to obtain the layered multi-scale features; inputting the obtained hierarchical multi-scale features into the semantic segmentation network, and outputting to obtain a mask of the promoted commodity; calculating the obtained promoted commodity mask by adopting a contour tracing algorithm to obtain an edge point set of the densely promoted commodity, then sampling the obtained edge point set of the densely promoted commodity according to a preset interval by adopting an equal-interval point sampling algorithm to obtain an edge point set of the equidistantly promoted commodity, normalizing the obtained coordinate values of the edge point set of the equidistantly promoted commodity in a mode of dividing the coordinate values by the length and the width of an image to obtain a vector point set of the promoted commodity, taking the output characteristics of the fusion network as input, selecting characteristic values of corresponding coordinate positions on the characteristics by using the coordinates of the obtained vector point set of the promoted commodity, and outputting the characteristic values of the promoted commodity vector;

the third model module is used for building a vector point set iteration model, and the vector point set iteration model comprises a classification network and a regression network; the classification network is used for judging whether each point is a foreground point or a background point, if the point is the foreground point, the point is reserved, if the point is the background point, the point is deleted, the classification network comprises a classification full-connection layer and a normalization index function activation layer which are cascaded, firstly, a fixed dimension of a promoted commodity vector feature is mapped to a class space with a space dimension of two through the classification full-connection layer to obtain a class feature, and then the class feature is input to the normalization index function activation layer to obtain a classification prediction result; the regression network is used for coordinate point regression and comprises a cascade regression full-link layer and a coordinate regression layer, wherein the vector characteristics of the promoted commodities are input into the regression full-link layer to obtain low-dimensional characteristics, and then the low-dimensional characteristics are input into the coordinate regression layer which is a multilayer perceptron to obtain a coordinate prediction deviant; judging whether the generalized commodity is a vector point on the generalized commodity according to the classification prediction result by combining the point coordinates of the generalized commodity vector point set, and carrying out coordinate migration through a coordinate prediction deviation value to obtain a generalized commodity prediction point; matching the promoted commodity prediction point and the real point by adopting a combined optimization algorithm, wherein the matching items comprise a category confidence coefficient and an Euclidean distance, the category confidence coefficient is from the classification network, the Euclidean distance is the Euclidean metric for calculating the distance between the real point and the promoted commodity prediction point, the higher the confidence coefficient is, the smaller the Euclidean distance is, the better the matching is, and the final matching degree is based on the weighted summation of the confidence coefficient and the Euclidean distance; after the predicted points and the real points of the promoted commodities are matched, cost function calculation is carried out, and for the classification network, a cross entropy cost function is adopted to evaluate the difference; for the regression network, evaluating the gap by adopting a regression cost function, and completing the iteration of the vector point set of the promoted commodity through model fitting and parameter iteration;

In one embodiment, the blockchain-based network promotion marketing platform further comprises:

the first commodity module is used for generating a commodity information calling instruction based on the three-dimensional simulation commodity map;

the second commodity module is used for calling current commodity information data corresponding to the three-dimensional simulation commodity map from a prestored commodity information database according to the commodity information calling instruction;

the third commodity module is used for binding the current commodity information data with the three-dimensional simulation commodity drawing, displaying the information of the current commodity information data on each part of the three-dimensional simulation commodity drawing, and then generating a displayable three-dimensional simulation commodity drawing;

and the network promotion instruction is used for sending the displayable three-dimensional simulated commodity map to a preset network promotion target address, and generating a promotion display interface at the network promotion target address after the network promotion target address is sent to, wherein the promotion display interface is used for displaying the displayable three-dimensional simulated commodity map.

In one embodiment, as shown in fig. 3, a computer device includes a memory and a processor, the memory stores a computer program, and the processor implements the steps of the block chain based network promotion marketing method when executing the computer program.

A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, implements the steps of the above block chain-based network popularization marketing method.

It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by hardware instructions of a computer program, which can be stored in a non-volatile computer-readable storage medium, and when executed, can include the processes of the embodiments of the methods described above. Any reference to memory, storage, database, or other medium used in the embodiments provided herein may include non-volatile and/or volatile memory, among others. Non-volatile memory can include read-only memory (ROM), Programmable ROM (PROM), Electrically Programmable ROM (EPROM), Electrically Erasable Programmable ROM (EEPROM), or flash memory. Volatile memory can include Random Access Memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms such as Static RAM (SRAM), Dynamic RAM (DRAM), Synchronous DRAM (SDRAM), Double Data Rate SDRAM (DDRSDRAM), Enhanced SDRAM (ESDRAM), synchronous Link (Synchlink) DRAM (SLDRAM), Rambus (Rambus) direct RAM (RDRAM), direct bused dynamic RAM (DRDRAM), and bused dynamic RAM (RDRAM).

The technical features of the above embodiments can be arbitrarily combined, and for the sake of brevity, all possible combinations of the technical features in the above embodiments are not described, but should be considered as the scope of the present specification as long as there is no contradiction between the combinations of the technical features.

The above-mentioned embodiments only express several embodiments of the present application, and the description thereof is more specific and detailed, but not construed as limiting the scope of the invention. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the concept of the present application, which falls within the scope of protection of the present application. Therefore, the protection scope of the present patent shall be subject to the appended claims.

Claims

1. A block chain-based network promotion marketing method is characterized by comprising the following steps:

step S300: based on a block chain technology, performing Hash chain on the vectorization result of the promoted commodity in a Hash's certificate mode, and simultaneously generating a three-dimensional simulated commodity graph according to the vectorization result of the promoted commodity;

2. The block chain-based network popularization marketing method of claim 1, wherein the step S300: acquiring a current commodity picture to be promoted loaded by the information loading column to be promoted, inputting the current commodity picture to be promoted to a preset vectorization result output model, and outputting a vectorization result of the promoted commodity by the vectorization result output model; before, still include:

step S310: before network popularization, picture processing is carried out on a commodity to be popularized, pictures of the commodity to be popularized are obtained, wherein one commodity to be popularized corresponds to at least one picture of the commodity to be popularized;

3. The blockchain-based network promotional marketing method according to claim 2, wherein the step S320: performing model building training processing on the to-be-promoted commodity picture based on the to-be-promoted commodity picture, and generating the vectorization result output model after the model building training processing is completed; the method specifically comprises the following steps:

step S321: acquiring promoted commodities in the pictures of the commodities to be promoted as promoted commodity samples, manufacturing salted films of the promoted commodity samples and corresponding region images into promoted commodity vector data sets, then carrying out normalization processing on the promoted commodity vector data sets, and finally carrying out data enhancement processing on the normalized promoted commodity vector data sets through vertical overturning, horizontal overturning and random rotation to obtain promoted commodity vector training sets;

step S322: building a promoted commodity mask generation model, wherein the promoted commodity mask generation model comprises a fusion network and a semantic segmentation network; the fusion network further comprises a large-scale module, a medium-scale module and a small-scale module, the large-scale module comprises nine cascaded convolutional layers, and the input of the large-scale network is the promoted commodity vector training set; the input of the mesoscale module is a second convolutional layer of the large-scale module, firstly pooling is carried out on the features obtained by the second convolutional layer to obtain the features reduced by one time, then the features are input to the following six convolutional layers to carry out convolution operation, and finally the features with the same size as the large-scale module are obtained by twice upsampling; the small-scale module comprises three cascaded convolution layers, the input of the small-scale module is the third convolution layer of the medium-scale module, the features obtained by the third convolution layer are firstly pooled to obtain the features reduced by one time, then the features are input to the three convolution layers for convolution operation, finally the features with the same size as the large-scale module are obtained through four-time up-sampling, and finally the three sets of the features with the same size of the large-scale module, the medium-scale module and the small-scale module are added to obtain the layered multi-scale features; inputting the obtained hierarchical multi-scale features into the semantic segmentation network, and outputting to obtain a mask of the promoted commodity; calculating the obtained promoted commodity mask by adopting a contour tracing algorithm to obtain an edge point set of the densely promoted commodity, then sampling the obtained edge point set of the densely promoted commodity according to a preset interval by adopting an equal-interval point sampling algorithm to obtain an edge point set of the equidistantly promoted commodity, normalizing the obtained coordinate values of the edge point set of the equidistantly promoted commodity in a mode of dividing the coordinate values by the length and the width of an image to obtain a vector point set of the promoted commodity, taking the output characteristics of the fusion network as input, selecting characteristic values of corresponding coordinate positions on the characteristics by using the coordinates of the obtained vector point set of the promoted commodity, and outputting the characteristic values of the promoted commodity vector;

step S323: constructing a vector point set iteration model, wherein the vector point set iteration model comprises a classification network and a regression network; the classification network is used for judging whether each point is a foreground point or a background point, if the point is the foreground point, the point is reserved, if the point is the background point, the point is deleted, the classification network comprises a classification full-connection layer and a normalization index function activation layer which are cascaded, firstly, a fixed dimension of a promoted commodity vector feature is mapped to a class space with a space dimension of two through the classification full-connection layer to obtain a class feature, and then the class feature is input to the normalization index function activation layer to obtain a classification prediction result; the regression network is used for coordinate point regression and comprises a cascade regression full-link layer and a coordinate regression layer, wherein the vector characteristics of the promoted commodities are input into the regression full-link layer to obtain low-dimensional characteristics, and then the low-dimensional characteristics are input into the coordinate regression layer which is a multilayer perceptron to obtain a coordinate prediction deviant; judging whether the generalized commodity is a vector point on the generalized commodity according to the classification prediction result by combining the point coordinates of the generalized commodity vector point set, and carrying out coordinate migration through a coordinate prediction deviation value to obtain a generalized commodity prediction point; matching the promoted commodity prediction point and the real point by adopting a combined optimization algorithm, wherein the matching items comprise a category confidence coefficient and an Euclidean distance, the category confidence coefficient is from the classification network, the Euclidean distance is the Euclidean metric for calculating the distance between the real point and the promoted commodity prediction point, the higher the confidence coefficient is, the smaller the Euclidean distance is, the better the matching is, and the final matching degree is based on the weighted summation of the confidence coefficient and the Euclidean distance; after matching the predicted points and the real points of the promoted commodities, calculating a cost function, and evaluating the difference of the classification network by adopting a cross entropy cost function; for the regression network, evaluating the gap by adopting a regression cost function, and completing the iteration of the vector point set of the promoted commodity through model fitting and parameter iteration;

4. The blockchain-based network promotion marketing method according to claim 3, wherein in the step S323: calculating the Euclidean measurement between the real point and the predicted point of the promoted commodity by adopting the following formula:

d＝sqrt((x1-x2)^+(y1-y2)^)；

5. The blockchain-based network promotion marketing method according to any one of claims 1 to 4, wherein the step S500: generating a network promotion instruction based on the three-dimensional simulated commodity map, wherein the network promotion instruction is used for sending the three-dimensional simulated commodity map to a preset network promotion target address and generating a promotion display interface at the network promotion target address after the network promotion instruction is sent to the network promotion target address, and the promotion display interface is used for displaying the three-dimensional simulated commodity map; the method specifically comprises the following steps:

step S540: and generating the network promotion instruction according to the displayable three-dimensional simulated commodity map, wherein the network promotion instruction is used for sending the displayable three-dimensional simulated commodity map to a preset network promotion target address, and generating a promotion display interface at the network promotion target address after the network promotion target address is sent, and the promotion display interface is used for displaying the displayable three-dimensional simulated commodity map.

6. A blockchain-based cyber-promotional marketing platform, the platform comprising:

7. The blockchain-based network promotional marketing platform of claim 6 wherein said object vector module comprises:

and the result output module is used for carrying out model building training processing on the pictures of the commodities to be promoted based on the pictures of the commodities to be promoted and generating the vectorization result output model after the model building training processing is finished.

8. The blockchain-based network promotional marketing platform of claim 6 wherein said result output module comprises:

the second model module is used for building a promoted commodity mask generation model, and the promoted commodity mask generation model comprises a fusion network and a semantic segmentation network; the fusion network further comprises a large-scale module, a medium-scale module and a small-scale module, the large-scale module comprises nine cascaded convolutional layers, and the input of the large-scale network is the promoted commodity vector training set; the input of the mesoscale module is a second convolutional layer of the large-scale module, firstly pooling is carried out on the features obtained by the second convolutional layer to obtain the features reduced by one time, then the features are input to the following six convolutional layers to carry out convolution operation, and finally the features with the same size as the large-scale module are obtained by twice upsampling; the small-scale module comprises three cascaded convolution layers, the input of the small-scale module is the third convolution layer of the medium-scale module, firstly, pooling is carried out on the characteristics obtained by the third convolution layer to obtain characteristics reduced by one time, then, the characteristics are input to the three convolution layers to carry out convolution operation, finally, characteristics with the same size as that of the large-scale module are obtained through four-time up-sampling, and finally, three groups of characteristics with the same size of the large-scale module, the medium-scale module and the small-scale module are added to obtain layered multi-scale characteristics; inputting the obtained hierarchical multi-scale features into the semantic segmentation network, and outputting to obtain a promoted commodity mask; calculating the obtained promoted commodity mask by adopting a contour tracing algorithm to obtain an edge point set of the densely promoted commodity, then sampling the obtained edge point set of the densely promoted commodity according to a preset interval by adopting an equidistant point sampling algorithm to obtain an edge point set of the equidistantly promoted commodity, then normalizing the obtained coordinate values of the edge point set of the equidistantly promoted commodity in a mode of dividing the coordinate values by the length and the width of an image to obtain a vector point set of the promoted commodity, taking the output characteristics of the fusion network as input, selecting characteristic values of corresponding coordinate positions on the characteristics by using the coordinates of the obtained vector point set of the promoted commodity, and outputting to obtain vector characteristics of the promoted commodity;

the third model module is used for building a vector point set iteration model, and the vector point set iteration model comprises a classification network and a regression network; the classification network is used for judging whether each point is a foreground point or a background point, reserving the foreground point if the foreground point is the foreground point, and deleting the background point if the background point is the background point, and comprises a classification full-link layer and a normalization index function activation layer which are cascaded; the regression network is used for coordinate point regression and comprises a cascade regression full-link layer and a coordinate regression layer, wherein the vector characteristics of the promoted commodities are input into the regression full-link layer to obtain low-dimensional characteristics, and then the low-dimensional characteristics are input into the coordinate regression layer which is a multilayer perceptron to obtain a coordinate prediction deviant; judging whether the generalized commodity is a vector point on the generalized commodity according to the classification prediction result by combining the point coordinates of the generalized commodity vector point set, and carrying out coordinate migration through a coordinate prediction deviation value to obtain a generalized commodity prediction point; matching the promoted commodity prediction point and the real point by adopting a combined optimization algorithm, wherein a matching item comprises a category confidence coefficient and an Euclidean distance, the category confidence coefficient is from the classification network, the Euclidean distance is an Euclidean metric for calculating the distance between the real point and the promoted commodity prediction point, the higher the confidence coefficient is, the smaller the Euclidean distance is, the better the matching is, and the final matching degree is based on the weighted summation of the confidence coefficient and the Euclidean distance; after matching the predicted points and the real points of the promoted commodities, calculating a cost function, and evaluating the difference of the classification network by adopting a cross entropy cost function; for the regression network, evaluating the gap by adopting a regression cost function, and completing the iteration of the vector point set of the promoted commodity through model fitting and parameter iteration;

9. A computer device comprising a memory and a processor, the memory storing a computer program, wherein the processor implements the steps of the method of any one of claims 1 to 5 when executing the computer program.

10. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the steps of the method of any one of claims 1 to 5.