CN117634864B

CN117634864B - Intelligent construction task optimization method and system based on image analysis

Info

Publication number: CN117634864B
Application number: CN202410099092.7A
Authority: CN
Inventors: 郑小鼎; 朱正伟; 陈兵; 王志太; 张员; 程强强; 马军卫; 张领雷
Original assignee: HUAREN CONSTRUCTION GROUP CO Ltd
Current assignee: HUAREN CONSTRUCTION GROUP CO Ltd
Priority date: 2024-01-24
Filing date: 2024-01-24
Publication date: 2024-04-05
Anticipated expiration: 2044-01-24
Also published as: CN117634864A

Abstract

The embodiment of the application provides an intelligent construction task optimization method and system based on image analysis, which can effectively match a design image with an actual construction video by acquiring first building design image data and first construction video data and determining building space vectors of elements or fragments covered in the first building design image data and the first construction video data by utilizing a first neural network, so that the space vector relation between building elements and the construction video fragments is parameterized and described. On the basis, a construction task optimization model can be obtained by calculating vector offset parameters and updating parameters of the first neural network, so that a target reference construction task can be generated according to target building design image data, the construction process is effectively guided, and the execution of the construction task is optimized. Therefore, the construction efficiency is improved, the construction difficulty is reduced, the construction quality and the safety are improved, and the resource consumption is saved.

Description

Intelligent construction task optimization method and system based on image analysis

Technical Field

The application relates to the technical field of intelligent construction sites, in particular to an intelligent construction site construction task optimization method and system based on image analysis.

Background

Building construction is a complex process involving comprehensive adjustment and optimization of various factors including personnel, equipment, materials, environment, etc. Traditional building construction management mainly depends on experience and visual judgment, is low in efficiency and easy to make mistakes, and cannot adapt to different building designs and construction environments. Therefore, there is a need for a more efficient, accurate and adaptive construction task generation method.

In recent years, with the development of artificial intelligence technology, related art has begun to attempt construction task generation using neural networks. However, existing methods for generating construction tasks based on neural networks still have some problems, for example, they often consider only static features of a building design, and ignore dynamic features of a construction process, which may cause the generated construction tasks to be inconsistent with the actual construction process. In addition, existing methods do not handle data differences in different architectural designs and construction environments well.

Disclosure of Invention

In order to at least overcome the defects in the prior art, the purpose of the application is to provide an intelligent construction task optimization method and system based on image analysis.

In a first aspect, the present application provides an intelligent construction task optimization method based on image analysis, applied to an intelligent construction task optimization system based on image analysis, the method comprising:

acquiring first building design image data and first construction video data, wherein the first building design image data and the construction execution characteristics of the first construction video data are matched, the first building design image data comprise a plurality of first building construction elements, and the first construction video data comprise a plurality of first construction video fragments;

determining building space vectors of all first building construction elements covered in the first building design image data according to a first neural network, wherein the building space vectors of the first building construction elements represent the first building construction elements and meet a set building vector distribution rule;

determining building space vectors of all first construction video segments covered in the first construction video data according to the first neural network, wherein the building space vectors of the first construction video segments represent the first construction video segments and meet the set building vector distribution rule;

Determining a first vector offset parameter from the building space vector of the respective first building construction element to the building space vector of the respective first construction video segment and a second vector offset parameter from the building space vector of the respective first construction video segment to the building space vector of the respective first building construction element;

and carrying out parameter updating on the first neural network according to the first vector offset parameter and the second vector offset parameter to generate a construction task optimization model, wherein the construction task optimization model is used for generating a target reference construction task according to target building design image data.

In a possible implementation manner of the first aspect, the determining, according to the first neural network, a building space vector of each first building construction element included in the first building design image data includes:

performing feature extraction on the first building design image data according to the first neural network to generate building mapping features of each first building construction element;

and carrying out space conversion on the building mapping characteristics of each first building construction element according to the first neural network to generate building space vectors of each first building construction element.

In a possible implementation manner of the first aspect, the method further includes:

according to the building space vector of each first building construction element and the building space vector of each first construction video segment, synchronously calibrating each first building construction element and each first construction video segment to generate a first segment quantity of each first building construction element corresponding to the first construction video segment;

determining second segment quantities of the first construction video segments corresponding to the first construction elements according to the construction mapping characteristics of the first construction elements;

determining segment quantity deviation parameters between a first segment quantity and a second segment quantity of the first construction video segment corresponding to each first building construction element;

the step of updating parameters of the first neural network according to the first vector offset parameter and the second vector offset parameter to generate a construction task optimization model comprises the following steps:

and carrying out parameter updating on the first neural network according to the fragment quantity deviation parameter, the first vector deviation parameter and the second vector deviation parameter to generate a construction task optimization model.

In a possible implementation manner of the first aspect, the determining, according to the first neural network, a building space vector of each first construction video segment included in the first construction video data includes:

extracting features of the first construction video data to generate construction video content vectors of the first construction video segments;

and performing space conversion on the construction video content vectors of the first construction video segments according to the first neural network to generate building space vectors of the first construction video segments.

In a possible implementation manner of the first aspect, the first construction video data is first template construction acquisition data or attention video data of the first template construction acquisition data, and the first neural network includes a first feature restoration unit; the method further comprises the steps of:

performing feature restoration on the construction video content vectors of each first construction video segment according to the first feature restoration unit to generate first restoration construction acquisition data;

determining a first comparison cost parameter between the first template construction acquisition data and the first restoration construction acquisition data;

according to the first comparison cost parameter, the first vector offset parameter and the second vector offset parameter, carrying out parameter updating on the first neural network to generate a construction task optimization model;

in a possible implementation manner of the first aspect, the first feature recovery unit includes a first loading subunit, at least two first feature extraction subunits, and a first result generation subunit, where any one of the first feature extraction subunits includes at least two filter layers with the same interval weight and different filter scales, and the filter layers of the different first feature extraction subunits correspond to different interval weights;

the feature reduction is performed on the construction video content vectors of the first construction video segments according to the first feature reduction unit, and first reduction construction collection data is generated, including:

converting the construction video content vector of each first construction video segment into a loading content vector of a first dimension according to the first loading subunit;

According to each filtering layer included in the first feature extraction subunit, performing extension filtering on the loading content vector of the first dimension to generate filtering feature data corresponding to each filtering layer, and fusing the filtering feature data corresponding to each filtering layer to generate result feature data of the first feature extraction subunit;

for any one of the first feature extraction subunits except the first feature extraction subunit, performing extension filtering on the result feature data of the previous first feature extraction subunit according to each filtering layer included in the any one of the first feature extraction subunits to generate filtering feature data corresponding to each filtering layer, and fusing the filtering feature data corresponding to each filtering layer to generate the result feature data of the any one of the first feature extraction subunits;

and converting the result characteristic data of the last first characteristic extraction subunit into the first restoration construction acquisition data according to the first result generation subunit.

In a possible implementation manner of the first aspect, the updating parameters of the first neural network according to the first vector offset parameter and the second vector offset parameter, to generate a construction task optimization model, includes:

Updating the functional function parameters of the first neural network according to the first vector offset parameters and the second vector offset parameters to generate a second neural network, wherein the second neural network comprises a feature processing unit and a second feature restoring unit;

acquiring second construction video data, wherein the second construction video data is second template construction acquisition data or attention video data of the second template construction acquisition data, and the second construction video data comprises a plurality of second construction video segments;

extracting features of the second construction video data to generate construction video content vectors of all second construction video segments;

performing feature restoration on the construction video content vectors of each second construction video segment according to the second feature restoration unit to generate second restoration construction acquisition data;

determining a second comparison cost parameter between the second template construction acquisition data and the second restoration construction acquisition data;

updating the function parameters of the second characteristic reduction unit according to the second comparison cost parameters to generate a third characteristic reduction unit;

and determining the construction task optimization model according to the feature processing unit and the third feature reduction unit.

acquiring second building design image data, wherein the second building design image data and the construction execution characteristics of the second construction video data are matched, and the second building design image data comprises a plurality of second building construction elements; determining building space vectors of all second building construction elements covered in the second building design image data according to the feature processing unit, performing space conversion on construction video content vectors of all second construction video segments, and generating building space vectors of all second construction video segments;

determining a third vector offset parameter from the building space vector of the respective second building construction element to the building space vector of the respective second construction video segment and a fourth vector offset parameter from the building space vector of the respective second construction video segment to the building space vector of the respective second building construction element;

the step of updating the function parameters of the second feature reduction unit according to the second comparison cost parameter to generate a third feature reduction unit, including:

And updating the function parameters of the second feature reduction unit according to the third vector offset parameter, the fourth vector offset parameter and the second comparison cost parameter to generate a third feature reduction unit.

In a possible implementation manner of the first aspect, the determining the construction task optimization model according to the feature processing unit and the third feature reduction unit includes:

acquiring third construction video data, wherein the third construction video data is third template construction acquisition data or attention video data of the third template construction acquisition data, and the third construction video data comprises a plurality of third construction video segments;

extracting features of the third construction video data to generate construction video content vectors of all third construction video segments;

performing feature restoration on the construction video content vectors of each third construction video segment according to the third feature restoration unit to generate third restoration construction acquisition data; performing feature reduction on the construction video content vectors of each third construction video segment according to a fourth feature reduction unit to generate fourth reduction construction acquisition data, wherein the parameter quantity of the fourth feature reduction unit is smaller than that of the third feature reduction unit;

According to a third comparison cost parameter between the third reduction construction acquisition data and the fourth reduction construction acquisition data;

updating the function parameters of the fourth feature reduction unit according to the third comparison cost parameters to generate a target feature reduction unit;

determining the construction task optimization model according to the target feature processing unit and the target feature reduction unit;

the target feature reduction unit comprises a target loading subunit, at least two target feature extraction subunits and a target result generation subunit, wherein any one target feature extraction subunit comprises at least two filter layers with the same interval weight and different filter scales, and the filter layers of different target feature extraction subunits correspond to different interval weights;

in a possible implementation manner of the first aspect, the determining the construction task optimization model according to the feature processing unit and the target feature reduction unit includes:

integrating each filter layer in the any target feature extraction subunit into an integrated filter layer aiming at any target feature extraction subunit, and generating an updated feature extraction subunit, wherein the interval weight of the integrated filter layer and the filter layer in the any target feature extraction subunit is the same, and the filter scale of the integrated filter layer is not smaller than the filter scale of each filter layer in the any target feature extraction subunit;

Integrating the target loading subunit, at least two updating feature extraction subunits and the target result generation subunit to generate a target feature reduction unit;

determining the construction task optimization model according to the feature processing unit and the target feature reduction unit;

in a possible implementation manner of the first aspect, the integrating the arbitrary target feature extraction subunit including each filter layer into an integrated filter layer, generating an updated feature extraction subunit includes:

updating functional function parameters of the first filter layer aiming at the first filter layer with the filter scale smaller than the filter scale of the integrated filter layer, so as to generate an updated first filter layer, wherein the filter scale of the updated first filter layer is the same as the filter scale of the integrated filter layer;

and determining the updated characteristic extraction subunit according to the updated functional function parameters of the first filter layer and the updated functional function parameters of the second filter layer, wherein the filter scale of the second filter layer is the same as that of the integrated filter layer.

Acquiring target building design image data, wherein the target building design image data comprises a plurality of candidate building construction elements;

determining building space vectors of all candidate building construction elements covered in the target building design image data according to a construction task optimization model, wherein the building space vectors of the candidate building construction elements represent the candidate building construction elements and meet a set building vector distribution rule;

determining building space vectors of all candidate construction video segments according to the construction task optimization model and building space vectors of all candidate construction elements, wherein the building space vectors of the candidate construction video segments represent the candidate construction video segments and meet the set building vector distribution rule;

generating a target candidate construction task according to the construction space vector of each candidate construction video segment according to the construction task optimization model, wherein candidate construction video data corresponding to the target candidate construction task is matched with construction execution characteristics of the target construction design image data, and the candidate construction video data comprises each candidate construction video segment;

in a possible implementation manner of the first aspect, the determining, according to a construction task optimization model, a building space vector of each candidate building construction element included in the target building design image data includes:

Extracting features of the target building design image data according to the construction task optimization model to generate building mapping features of each candidate building construction element;

performing space conversion on building mapping features of each candidate building construction element according to the construction task optimization model to generate building space vectors of each candidate building construction element;

determining the segment quantity of the candidate construction video segments corresponding to each candidate construction element according to the construction mapping characteristics of each candidate construction element;

the construction task optimization model is used for determining the construction space vector of each candidate construction video segment according to the construction space vector of each candidate construction element, and the construction task optimization model comprises the following steps:

according to the construction task optimization model and the segment quantity of the candidate construction video segments corresponding to the candidate construction elements, building space vectors of the candidate construction elements are expanded to generate building space vectors of the candidate construction video segments;

in a possible implementation manner of the first aspect, the generating, according to the construction task optimization model, a target candidate construction task according to the building space vector of each candidate construction video segment includes:

Performing space conversion on the building space vectors of the candidate construction video segments according to the construction task optimization model to generate construction video content vectors of the candidate construction video segments;

and carrying out feature reduction on the construction video content vectors of the candidate construction video segments according to the construction task optimization model to generate target candidate construction tasks.

In a second aspect, an embodiment of the present application further provides an image analysis-based intelligent construction task optimization system, where the image analysis-based intelligent construction task optimization system includes a processor and a machine-readable storage medium, where the machine-readable storage medium stores a computer program, where the computer program is loaded and executed in conjunction with the processor to implement the above image analysis-based intelligent construction task optimization method of the first aspect.

By combining the above aspects, by acquiring the first architectural design image data and the first construction video data and determining architectural space vectors of elements or segments covered therein by using the first neural network, the design image can be effectively matched with the actual construction video, thereby parametrically describing the space vector relationship between the architectural elements and the construction video segments. On the basis, a construction task optimization model can be obtained by calculating vector offset parameters and updating parameters of the first neural network, so that a target reference construction task can be generated according to target building design image data, the construction process is effectively guided, and the execution of the construction task is optimized. Therefore, the construction efficiency is improved, the construction difficulty is reduced, the construction quality and the safety are improved, and the resource consumption is saved.

Drawings

For a clearer description of the technical solutions of the embodiments of the present application, reference will be made to the accompanying drawings, which are needed to be activated, for the sake of simplicity, and it should be understood that the following drawings only illustrate some embodiments of the present application and should therefore not be considered as limiting the scope, and that other related drawings can be obtained according to these drawings without the need for inventive effort for a person skilled in the art.

Fig. 1 is a schematic flow chart of an intelligent construction task optimization method based on image analysis according to an embodiment of the present application;

fig. 2 is a schematic block diagram of an intelligent construction task optimization system based on image analysis for implementing the intelligent construction task optimization method based on image analysis according to an embodiment of the present application.

Detailed Description

The following description is presented to enable one of ordinary skill in the art to make and use the application and is provided in the context of a particular application and its requirements. It will be apparent to those having ordinary skill in the art that various changes can be made to the disclosed embodiments and that the general principles defined herein may be applied to other embodiments and applications without departing from the principles and scope of the present application. Thus, the present application is not limited to the embodiments described, but is to be accorded the widest scope consistent with the claims.

Step S110, first building design image data and first construction video data are acquired.

In this embodiment, the first building design image data and the first construction video data are construction execution feature-matched, the first building design image data including a plurality of first building construction elements, the first construction video data including a plurality of first construction video clips.

For example, in the present embodiment, first building design image data from a building design unit and first construction video data of a construction unit may be received and stored. By way of example, the first building design image data may refer to building design image data provided to meet the requirements of a construction unit, specifically including design images of individual construction elements of a building (e.g., walls, columns, windows, etc.). For example, a building design unit may provide building design image data containing a plurality of building construction elements so that a construction unit can perform construction based on the building design image data. The first construction video data may refer to video data collected by a construction unit during a construction process. These first construction video data include construction video clips of the individual construction elements of the building. For example, a construction unit may collect a plurality of construction video clips so that a building design unit can understand the conditions during construction. The first building construction element may refer to a specific construction element in the building design image data, such as a wall, a column or a window. For example, in one architectural design image data, a plurality of wall, column, window, and other construction elements may be included. The first construction video clip may refer to a specific video clip in the first construction video data, such as a worker-installed window video clip. For example, in one construction video data, a video clip of a plurality of construction elements such as a worker-installed wall, a pillar, and a window may be included.

The construction execution feature matching of the first building design image data and the first construction video data may refer to a process of matching features in the building design image data and the construction video data. For example, features of a wall in the architectural design image data may match features of a video clip of a worker installing the wall in the construction video data. For example, the image recognition algorithm and the motion detection algorithm may be used to process the first architectural design image data and the first construction video data, respectively, to extract features such as texture, shape, color, and the like in the first architectural design image data, and features such as motion, light and the like in the first construction video data. The extracted features are then matched to ensure that they are data for the same building.

And step S120, determining building space vectors of all first building construction elements contained in the first building design image data according to a first neural network, wherein the building space vectors of the first building construction elements represent the first building construction elements and meet a set building vector distribution rule.

For example, the first architectural design image data may be processed using a first neural network based on a deep learning algorithm to automatically determine the architectural space vector for each first architectural element. This building space vector may represent information of the position, size, shape, etc. of the first building construction element. For example, a wall's building space vector may include its length, width, height, etc.

That is, the first neural network is an artificial intelligence model that can identify and classify images by learning a large amount of data. In this scenario, the first neural network may be used to identify individual first architectural construction elements in the architectural design image data. A building space vector may refer to a vector representing the position and orientation of a first building construction element in space. For example, a building space vector for a first building construction element may include information such as position coordinates, direction angles, etc. for the construction element.

Setting the construction vector distribution rule may refer to restricting the spatial distribution of the construction elements according to a preset rule. For example, a rule may be set that requires the style of the distribution of all walls. In this way, the building space vector determined by the first neural network may represent the first building construction element and satisfy the building vector distribution rule set by the building design company. For example, the distribution position style of all walls is required, so that in determining the building space vector, it is necessary to ensure that all walls satisfy this rule.

Step S130, determining building space vectors of all first construction video segments covered in the first construction video data according to the first neural network, wherein the building space vectors of the first construction video segments represent the first construction video segments and meet the set building vector distribution rule.

Likewise, the first construction video data may be processed using a first neural network based on a deep learning algorithm to automatically determine a construction space vector for each first construction video segment. This construction space vector may represent information of the position, time, action, etc. of the first construction video segment. For example, a building space vector of a first construction video segment of a worker installing a window may include information about its start time, end time, location of the worker, action, etc.

Step S140, determining a first vector offset parameter from the building space vector of the respective first building construction element to the building space vector of the respective first construction video segment and a second vector offset parameter from the building space vector of the respective first construction video segment to the building space vector of the respective first building construction element.

For example, a mathematical algorithm may be used to calculate an offset parameter from the building space vector of the first building construction element to the building space vector of the first construction video segment, and an offset parameter from the building space vector of the first construction video segment to the building space vector of the first building construction element. These offset parameters may help to understand the differences between building design and construction, thereby optimizing construction tasks.

Illustratively, assume that the first architectural design image data contains three first architectural construction elements (wall a, wall B, and window C), and that the first construction video data contains three first construction video segments (video segment 1 of worker-installed wall a, video segment 2 of worker-installed wall B, and video segment 3 of worker-installed window C).

First, it is necessary to extract a building space vector for each building construction element, and a building space vector for each construction video clip using a neural network model. For example, the building space vector of wall a may be represented as [ x1, y1, z1], the building space vector of wall B may be represented as [ x2, y2, z2], the building space vector of window C may be represented as [ x3, y3, z3], the building space vector of video clip 1 of worker-mounted wall a may be represented as [ x4, y4, z4], the building space vector of video clip 2 of worker-mounted wall B may be represented as [ x5, y5, z5], and the building space vector of video clip 3 of worker-mounted window C may be represented as [ x6, y6, z6].

Next, it is necessary to calculate a first vector offset parameter from the building space vector of each building construction element to the building space vector of each construction video segment, and a second vector offset parameter from the building space vector of each construction video segment to the building space vector of each building construction element. For example, a first vector offset parameter from a building space vector of wall a to a building space vector of video segment 1 of worker-installed wall a may be expressed as [ dx1, dy1, dz1], a second vector offset parameter from a building space vector of video segment 1 of worker-installed wall a to a building space vector of wall a may be expressed as [ dx2, dy2, dz2], a first vector offset parameter from a building space vector of wall B to a building space vector of video segment 2 of worker-installed wall B may be expressed as [ dx3, dy3, dz3], a second vector offset parameter from a building space vector of video segment 2 of worker-installed wall B to a building space vector of wall B may be expressed as [ dx4, dy4, dz4], a first vector offset parameter from a building space vector of window C to a building space vector of video segment 3 of worker-installed window C may be expressed as [ dx5, dy5, dz5], a first vector offset parameter from a building space vector of video segment 3 of worker-installed window C to a building space vector of window C may be expressed as [ dx3, dy3, dz3] and a second vector offset from a building space vector of video segment of worker-installed window C may be expressed as [ dx6, dy 6].

And step S150, carrying out parameter updating on the first neural network according to the first vector offset parameter and the second vector offset parameter to generate a construction task optimization model, wherein the construction task optimization model is used for generating a target reference construction task according to target building design image data.

For example, the first neural network may be updated based on the calculated offset parameters to generate a construction task optimization model. This construction task optimization model may generate an optimal construction task from the target building design image data, for example, predicting problems that may occur during construction, or recommending an optimal construction solution.

For example, the first vector offset parameter and the second vector offset parameter may be used to adjust the input data of the first neural network model to better match the architectural design image data and the construction video data. In particular, the first vector offset parameter may be used to adjust the positional information of the architectural element, while the second vector offset parameter may be used to adjust the directional information of the architectural element.

After training is completed, a construction task optimization model is obtained, so that a construction task can be generated according to the building design image data. Specifically, the building design image data can be used as input data and input into the construction task optimization model, and the construction task optimization model automatically generates a construction task which can be used as a reference of a construction unit to help to finish the construction task better. In practical application, the performance of the construction task optimization model can be optimized by continuously adjusting the first vector offset parameter and the second vector offset parameter so as to better meet different design requirements. For example, if an error or inaccuracy exists in the construction task generated by the construction task optimization model, the input data of the model can be changed by adjusting the first vector offset parameter and the second vector offset parameter, so that the accuracy and reliability of the construction task optimization model are improved.

Based on the above steps, in this embodiment, by acquiring the first architectural design image data and the first construction video data and determining the architectural space vector of each element or segment covered therein by using the first neural network, the design image and the actual construction video can be effectively matched, so that the parametric description is performed on the space vector relationship between the architectural elements and the construction video segments. On the basis, a construction task optimization model can be obtained by calculating vector offset parameters and updating parameters of the first neural network, so that a target reference construction task can be generated according to target building design image data, the construction process is effectively guided, and the execution of the construction task is optimized. Therefore, the construction efficiency is improved, the construction difficulty is reduced, the construction quality and the safety are improved, and the resource consumption is saved.

In one possible implementation, step S120 may include:

step S121, performing feature extraction on the first architectural design image data according to the first neural network, and generating architectural mapping features of each first architectural element.

Step S122, performing space conversion on the building mapping features of the first building construction elements according to the first neural network, and generating building space vectors of the first building construction elements.

For example, the first architectural design image data includes a plurality of first architectural construction elements, such as walls, windows, and doors. The first architectural design image data may be feature extracted using a first neural network to generate architectural mapping features for each first architectural element. For example, for a wall, the height, width, length and other features of the wall can be extracted; for the window, the size, shape, position and other characteristics of the window can be extracted; for the door, the type, size, position, etc. characteristics thereof can be extracted.

After the building mapping features are extracted, the building mapping features of each first building construction element may be spatially transformed using the first neural network to generate a building space vector for each first building construction element. For example, for a wall, its height, width, length, etc. characteristics may be converted into coordinate values in three-dimensional space; for a window, the size, shape, position and other characteristics of the window can be converted into coordinate values in a three-dimensional space; for a door, its type, size, position, etc. characteristics can be converted into coordinate values in three-dimensional space. Thus, a building space vector for each first building construction element is obtained.

In one possible embodiment, the method further comprises:

and step A110, synchronously calibrating the first building construction elements and the first construction video fragments according to the building space vectors of the first building construction elements and the building space vectors of the first construction video fragments, and generating first fragment amounts of the first construction video fragments corresponding to the first building construction elements.

And step A120, determining a second segment quantity of each first building construction element corresponding to the first construction video segment according to the building mapping characteristics of each first building construction element.

Step A130, determining segment quantity deviation parameters between a first segment quantity and a second segment quantity of the first construction video segment corresponding to each first building construction element.

Thus, in step S150, the first neural network may be updated according to the segment deviation parameter, the first vector deviation parameter, and the second vector deviation parameter, so as to generate a construction task optimization model.

Assuming that the building space vector of each first building construction element and the building space vector of each construction video segment have been determined, these building space vectors may be used to synchronously calibrate each first building construction element and each construction video segment, generating a first segment quantity of each first building construction element corresponding to the construction video segment. For example, for a wall, the first building space vector may be compared with the building space vectors of the first construction video segments, a portion of the wall in each first construction video segment may be determined, and the length, width, height, etc. of the portion may be calculated as the first segment amount.

After determining the first segment amount of each first construction element corresponding to the construction video segment, a second segment amount of each first construction element corresponding to the construction video segment may be determined using the building mapping feature. For example, for a wall, a portion of the wall in each first construction video clip may be calculated using the features of height, width, length, etc. in the building map feature, and the features of length, width, height, etc. of the portion may be calculated as the second clip quantity.

After determining the first segment amount and the second segment amount of each first building construction element corresponding to the first construction video segment, a segment amount deviation parameter therebetween may be calculated. For example, for a wall, a deviation of characteristics such as length, width, height, etc., between a first segment amount and a second segment amount of the wall in each first construction video segment may be calculated as a segment amount deviation parameter.

And finally, parameter updating can be performed on the first neural network by using the fragment deviation parameter, the first vector deviation parameter and the second vector deviation parameter to generate a construction task optimization model. For example, these parameters may be used to adjust weights and offsets of the neural network so that the neural network is better able to predict segment quantity deviation parameters for each first architectural element corresponding to a constructed video segment.

In one possible implementation, step S130 may include:

and step S131, extracting the characteristics of the first construction video data to generate construction video content vectors of the first construction video segments.

And step S132, performing space conversion on the construction video content vectors of the first construction video segments according to the first neural network to generate building space vectors of the first construction video segments.

For example, a plurality of first construction video pieces, such as a worker installing a wall, a window, a door, and the like, are included in the first construction video data. The construction video data may be feature extracted using a first neural network to generate a construction video content vector for each first construction video segment. For example, for a first construction video clip of a worker installing a wall, features of the worker's actions, tools used, speed of installation, etc. may be extracted and combined into one construction video content vector.

After the construction video content vector is extracted, the construction video content vector for each first construction video segment may be spatially transformed using the first neural network to generate a construction space vector for each first construction video segment. For example, for a first construction video clip whose construction video content vector is to be converted into a coordinate value in three-dimensional space as a construction space vector, a worker installs a wall body. Thus, a construction space vector for each first construction video segment is obtained.

In one possible implementation, the first construction video data is first template construction acquisition data or attention video data of the first template construction acquisition data, and the first neural network includes a first feature reduction unit. The method further comprises the steps of:

and step B110, performing feature restoration on the construction video content vectors of the first construction video segments according to the first feature restoration unit to generate first restoration construction acquisition data.

And step B120, determining a first comparison cost parameter between the first template construction acquisition data and the first restoration construction acquisition data.

Thus, in step S150, the first neural network may be updated according to the first comparison cost parameter, the first vector offset parameter, and the second vector offset parameter, so as to generate a construction task optimization model.

For example, the first template construction collection data includes a plurality of first construction video clips, such as workers installing walls, windows, doors, and the like. And performing feature restoration on the construction video content vector of each first construction video segment by using a first feature restoration unit to generate first restoration construction acquisition data. For example, for a construction video clip of a worker installing a wall, its construction video content vector may be restored to the original video data as the first restored construction acquisition data. Thus, each first reduction construction acquisition data is obtained.

After the first restoration construction acquisition data is generated, a first comparison cost parameter may be used to determine a difference between the first template construction acquisition data and the first restoration construction acquisition data. For example, the first comparison cost parameter may be calculated using an index such as mean square error, average absolute error, or the like.

And finally, parameter updating can be performed on the first neural network by using the first comparison cost parameter, the first vector offset parameter and the second vector offset parameter, so as to generate a construction task optimization model. For example, the weights and biases of the first neural network may be updated using a gradient descent algorithm so that the first neural network is better able to predict the first comparison cost parameter.

In a possible implementation manner, the first feature recovery unit includes a first loading subunit, at least two first feature extraction subunits, and a first result generation subunit, where any one of the first feature extraction subunits includes at least two filter layers with the same interval weights and different filter scales, and the filter layers of the different first feature extraction subunits correspond to different interval weights.

Step B110 may include:

and step B111, converting the construction video content vectors of the first construction video segments into loading content vectors with a first dimension according to the first loading subunit.

And step B112, performing extension filtering on the loading content vector of the first dimension according to each filtering layer included in the first feature extraction subunit, generating filtering feature data corresponding to each filtering layer, and fusing the filtering feature data corresponding to each filtering layer to generate result feature data of the first feature extraction subunit.

And step B113, for any one of the first feature extraction subunits except the first feature extraction subunit, performing extended filtering on the result feature data of the previous first feature extraction subunit according to each filtering layer included in the any one of the first feature extraction subunits to generate filtering feature data corresponding to each filtering layer, and fusing the filtering feature data corresponding to each filtering layer to generate the result feature data of the any one of the first feature extraction subunits.

And step B114, converting the result characteristic data of the last first characteristic extraction subunit into the first restoration construction acquisition data according to the first result generation subunit.

Assume that there is a construction video content vector of a first construction video clip, in which a plurality of features are included, such as actions of a worker, tools used, speed of installation, etc. The first loading subunit may be used to convert the construction video content vector into a loaded content vector of a first dimension. For example, the construction video content vector may be converted into a 1024-dimensional loading content vector.

After the loaded content vector of the first dimension is generated, the loaded content vector can be subjected to extended filtering by using each filtering layer included in the first feature extraction subunit, so as to generate filtering feature data corresponding to each filtering layer. For example, 3 filter layers may be used to filter the loaded content vector separately, generating 3 filter characteristic data. The 3 filtered feature data may then be fused to generate resultant feature data for the first feature extraction subunit.

After the result feature data of the first feature extraction subunit is generated, any one of the first feature extraction subunits other than the first feature extraction subunit may be used to perform extension filtering on the result feature data of the last first feature extraction subunit, so as to generate filter feature data corresponding to each filter layer. For example, 3 filter layers may be used to filter the resulting feature data of the last first feature extraction subunit, respectively, to generate 3 filtered feature data. Then, the 3 filter feature data may be fused to generate the result feature data of any one of the first feature extraction subunits.

After the result feature data of any one of the first feature extraction subunits is generated, the result feature data of the last first feature extraction subunit may be converted into first restoration construction acquisition data using the first result generation subunit. For example, the resulting feature data of the last first feature extraction subunit may be converted to first restoration construction acquisition data using a neural network.

In one possible implementation, step S150 may include:

step S151, updating the functional function parameters of the first neural network according to the first vector offset parameter and the second vector offset parameter, to generate a second neural network, where the second neural network includes a feature processing unit and a second feature restoring unit.

For example, the first neural network has learned some knowledge about the construction task. Now, it is desirable to use this knowledge to optimize construction tasks. In this embodiment, the first vector offset parameter and the second vector offset parameter may be used to update the functional function parameters of the first neural network to generate the second neural network. In the second neural network, a feature processing unit and a second feature restoration unit are added to help better process the construction video data.

Step S152, obtaining second construction video data, where the second construction video data is second template construction collection data or attention video data of the second template construction collection data, and the second construction video data includes a plurality of second construction video segments.

For example, assuming that further optimization of the construction task is desired, some second construction video data may be acquired. The second construction video data may be second template construction acquisition data or attention video data of the second template construction acquisition data. The second construction video data may include a plurality of second construction video segments, each of which contains some image or video data pertaining to a construction task.

And step S153, extracting the characteristics of the second construction video data to generate construction video content vectors of all the second construction video segments.

For example, to better understand the second construction video data, it may be feature extracted to generate construction video content vectors for respective second construction video segments. These vectors may represent key features in each segment, such as the shape, size, color, etc. of the building.

And step S154, performing feature restoration on the construction video content vectors of the second construction video segments according to the second feature restoration unit to generate second restoration construction acquisition data.

For example, the second feature reduction unit may be used to perform feature reduction on the construction video content vector of the second construction video segment to generate second reduction construction acquisition data. These data may represent key features in each segment, such as the shape, size, color, etc. of the building.

Step S155, determining a second comparison cost parameter between the second template construction acquisition data and the second restoration construction acquisition data.

For example, a second comparison cost parameter between the second template construction collection data and the second restoration construction collection data may be used to determine the similarity therebetween. These parameters may represent key features in each segment, such as the shape, size, color, etc. of the building.

Step S156, updating the function parameters of the second feature reduction unit according to the second comparison cost parameter, and generating a third feature reduction unit.

For example, the second comparison cost parameter may be used to update the functional function parameters of the second feature reduction unit, generating the third feature reduction unit. These parameters may represent key features in each segment, such as the shape, size, color, etc. of the building.

And step S157, determining the construction task optimization model according to the characteristic processing unit and the third characteristic reduction unit.

For example, a feature processing unit and a third feature reduction unit may be used to determine a construction task optimization model so that key features in each segment, e.g., shape, size, color, etc., of a building may be more accurately represented.

In one possible embodiment, the method further comprises:

and step C110, acquiring second building design image data, wherein the second building design image data and the construction execution characteristics of the second construction video data are matched, and the second building design image data comprises a plurality of second building construction elements. And determining building space vectors of all second building construction elements covered in the second building design image data according to the feature processing unit, and performing space conversion on the construction video content vectors of all second construction video segments to generate building space vectors of all second construction video segments.

Step C120, determining a third vector offset parameter from the building space vector of the respective second building construction element to the building space vector of the respective second construction video segment and a fourth vector offset parameter from the building space vector of the respective second construction video segment to the building space vector of the respective second building construction element.

Thus, in step S156, the functional function parameters of the second feature recovery unit may be updated according to the third vector offset parameter, the fourth vector offset parameter, and the second comparison cost parameter, so as to generate a third feature recovery unit.

For example, assuming that further optimization of the construction task is desired, some second architectural design image data may be acquired. These second building design image data may be matched with construction performance characteristics in the second construction video data, such as the shape, size, color, etc. of the building. These second building design image data may include a plurality of second building construction elements, such as walls, doors, windows, and the like.

A feature processing unit may be used to determine a building space vector for each second building construction element encompassed in the second building design image data. For example, image recognition techniques may be used to determine the location, size, shape, etc. of each second architectural element encompassed in the second architectural design image data. Then, the construction video content vectors of the second construction video segments may be spatially converted to generate building space vectors for each of the second construction video segments. For example, the construction video content vector of the second construction video segment may be converted into the building space using coordinate conversion techniques.

A coordinate conversion technique may be used to determine a third vector offset parameter from the building space vector of each second building construction element to the building space vector of each second construction video segment and a fourth vector offset parameter from the building space vector of each second construction video segment to the building space vector of each second building construction element. For example, a euclidean distance formula may be used to calculate a third vector offset parameter from the building space vector of each second building construction element to the building space vector of each second construction video segment and a fourth vector offset parameter from the building space vector of each second construction video segment to the building space vector of each second building construction element.

The third feature recovery unit may be generated by updating the functional function parameters of the second feature recovery unit using the third vector offset parameter, the fourth vector offset parameter, and the second comparison cost parameter. For example, a gradient descent algorithm may be used to update the functional function parameters of the second feature recovery unit, generating a third feature recovery unit.

Based on the above steps, the construction task can be optimized using the second building design image data, a third feature reduction unit is generated, and a construction task optimization model is determined using the feature processing unit and the third feature reduction unit, so that key features in each segment, such as the shape, size, color, and the like of a building, can be more accurately represented.

In one possible implementation, step S157 may include:

step S1571, third construction video data is obtained, wherein the third construction video data is third template construction acquisition data or attention video data of the third template construction acquisition data, and the third construction video data comprises a plurality of third construction video segments.

For example, some third construction video data may be acquired, assuming that further optimization of the construction task is desired. The third construction video data may be third template construction acquisition data or attention video data of the third template construction acquisition data. The third construction video data may include a plurality of third construction video segments, each third construction video segment containing some image or video data pertaining to a construction task.

And step S1572, extracting features of the third construction video data to generate construction video content vectors of all third construction video segments.

For example, to better understand the third construction video data, it may be feature extracted to generate construction video content vectors for respective third construction video segments. These vectors may represent key features in each segment, such as the shape, size, color, etc. of the building.

And step S1573, performing feature restoration on the construction video content vectors of the third construction video segments according to the third feature restoration unit to generate third restoration construction acquisition data. And carrying out feature reduction on the construction video content vectors of each third construction video segment according to a fourth feature reduction unit to generate fourth reduction construction acquisition data, wherein the parameter quantity of the fourth feature reduction unit is smaller than that of the third feature reduction unit.

For example, the third feature reduction unit may be used to perform feature reduction on the construction video content vector of the third construction video segment to generate third reduction construction acquisition data. These data may represent key features in each segment, such as the shape, size, color, etc. of the building.

Step S1574, according to a third comparison cost parameter between the third reduction construction collection data and the fourth reduction construction collection data.

For example, the construction video content vector of the third construction video segment may be subjected to feature restoration by using a fourth feature restoration unit to generate fourth restoration construction acquisition data. These data may represent key features in each segment, such as the shape, size, color, etc. of the building. The parameter quantity of the fourth characteristic reduction unit is smaller than that of the third characteristic reduction unit, so that the calculated quantity can be reduced, and the calculation efficiency is improved.

Step S1575, updating the function parameters of the fourth feature reduction unit according to the third comparison cost parameter, and generating a target feature reduction unit.

For example, a third comparison cost parameter between the third and fourth restoration construction acquisition data may be used. These parameters may represent differences between the third and fourth reduction construction acquisition data, e.g., shape, size, color, etc., of the building.

Step S1576, determining the construction task optimization model according to the feature processing unit and the target feature reduction unit.

For example, the third comparison cost parameter may be used to update the functional function parameters of the fourth feature reduction unit to generate the target feature reduction unit. These parameters may represent updated functional function parameters of the fourth feature reduction unit, such as the shape, size, color, etc. of the building.

In step S1577, the target feature restoration unit includes a target loading subunit, at least two target feature extraction subunits, and a target result generation subunit, where any one target feature extraction subunit includes at least two filter layers with the same interval weights and different filter scales, and the filter layers of different target feature extraction subunits correspond to different interval weights.

For example, a construction task optimization model may be determined using a feature processing unit and a target feature reduction unit, so that key features in each segment, such as the shape, size, color, etc., of a building may be more accurately represented.

In summary, the third construction video data may be used to optimize the construction task, generate the target feature reduction unit, and determine the construction task optimization model using the feature processing unit and the target feature reduction unit, so that the key features in each segment, such as the shape, size, color, etc., of the building may be more accurately represented.

In a possible implementation manner, in step S1576, for the arbitrary target feature extraction subunit, integrating each filter layer included in the arbitrary target feature extraction subunit into an integrated filter layer, and generating an updated feature extraction subunit, where the interval weights of the integrated filter layer and the filter layers included in the arbitrary target feature extraction subunit are the same, and the filter scale of the integrated filter layer is not smaller than the filter scale of each filter layer included in the arbitrary target feature extraction subunit. And then integrating the target loading subunit, at least two updating feature extraction subunits and the target result generation subunit to generate a target feature reduction unit. And finally, determining the construction task optimization model according to the target feature reduction unit and the target feature reduction unit.

For example, assume that there is one target feature extraction subunit including three filter layers, filter layer a, filter layer B, and filter layer C, respectively. The three filter layers may be integrated into an integrated filter layer D, generating an updated feature extraction subunit. The interval weight of the integrated filter layer D is the same as the interval weights of the filter layer A, the filter layer B and the filter layer C, and the filter scale of the integrated filter layer D is not smaller than the filter scale of the filter layer A, the filter layer B and the filter layer C.

It is assumed that there are two update feature extraction subunits, update feature extraction subunit 1 and update feature extraction subunit 2, respectively. The target load subunit, the update feature extraction subunit 1, the update feature extraction subunit 2, and the target result generation subunit may be integrated to generate a target feature recovery unit. Thus, the construction task optimization model can be determined using the feature processing unit and the target feature reduction unit, so that key features in each segment, such as the shape, size, color, etc., of a building can be more accurately represented.

In summary, the target feature extraction subunit and the target result generation subunit may be used to generate the target feature reduction unit, and the feature processing unit and the target feature reduction unit may be used to determine the construction task optimization model, so that the key features in each segment, such as the shape, size, color, and the like, of the building may be more accurately represented.

In a possible implementation manner, the integrating the any one target feature extraction subunit into an integrated filter layer including each filter layer, and generating an updated feature extraction subunit includes: and updating the functional function parameters of the first filter layer aiming at the first filter layer with the filter scale smaller than the filter scale of the integrated filter layer, so as to generate an updated first filter layer, wherein the filter scale of the updated first filter layer is the same as the filter scale of the integrated filter layer. And then, determining the updated feature extraction subunit according to the updated functional function parameters of the first filter layer and the updated functional function parameters of the second filter layer, wherein the filtering scale of the second filter layer is the same as that of the integrated filter layer.

For example, assume that the target feature extraction subunit includes three filter layers, filter layer a, filter layer B, and filter layer C, respectively. Wherein, the filtering scale of the filtering layer A is smaller than the filtering scale of the integrated filtering layer. The function parameters of the filter layer a can be updated to generate an updated filter layer a ', so that the filter scale of the updated filter layer a' is the same as the filter scale of the integrated filter layer.

After generating the updated filter layer a ', the updated feature extraction subunit may be determined according to the updated functional function parameters of the filter layer a' and the functional function parameters of the filter layer B. The filtering scale of the filtering layer B is the same as that of the integrated filtering layer.

Thus, the updated filter layer a' and filter layer B may be used to generate an updated feature extraction subunit, and the updated feature extraction subunit and the target result generation subunit may be used to generate a target feature reduction unit, so that key features in each segment, such as the shape, size, color, etc., of a building may be more accurately represented.

In a possible implementation manner, at a stage of applying the embodiment, the method further includes:

step S160, obtaining target building design image data, the target building design image data including a plurality of candidate building construction elements.

And S170, determining building space vectors of all candidate building construction elements contained in the target building design image data according to a construction task optimization model, wherein the building space vectors of the candidate building construction elements represent the candidate building construction elements and meet a set building vector distribution rule.

And step S180, determining building space vectors of all candidate construction video segments according to the construction task optimization model and the building space vectors of all candidate construction elements, wherein the building space vectors of the candidate construction video segments represent the candidate construction video segments and meet the set building vector distribution rule.

Step S190, generating a target candidate construction task according to the construction space vector of each candidate construction video segment according to the construction task optimization model, wherein candidate construction video data corresponding to the target candidate construction task is matched with construction execution characteristics of the target building design image data, and the candidate construction video data comprises each candidate construction video segment.

For example, the target building design image data may include a plurality of candidate building construction elements, such as windows, doors, walls, and the like.

On this basis, a construction task optimization model may be used to determine building space vectors for each candidate building construction element contained in the target building design image data. These building space vectors may characterize candidate building construction elements and are required to meet set building vector distribution rules.

Next, a construction task optimization model may be used to determine building space vectors for each candidate construction video segment. These building space vectors may characterize the candidate construction video segments and are required to meet set building vector distribution rules.

Then, a construction task optimization model may be used to generate target candidate construction tasks from the building space vectors of each candidate construction video segment. The candidate construction video data corresponding to these target candidate construction tasks matches the construction execution characteristics of the target building design image data, for example, the window should be located on a wall, the door should be located on a wall, and so on.

Thus, it is possible to generate a target candidate construction task from the building space vectors of the respective candidate building construction elements using the construction task optimization model, and to match the construction execution characteristics of the target building design image data using the candidate construction video data corresponding to the target candidate construction task. These target candidate construction tasks may be used to generate final construction tasks, e.g., a window should be located on a wall, a door should be located on a wall, etc.

In one possible implementation, step S170 may include:

And extracting features of the target building design image data according to the construction task optimization model, and generating building mapping features of each candidate building construction element.

And carrying out space conversion on the building mapping characteristics of each candidate building construction element according to the construction task optimization model, and generating building space vectors of each candidate building construction element.

In one possible embodiment, the method further comprises: and determining the segment quantity of the candidate construction video segments corresponding to each candidate construction element according to the construction mapping characteristics of each candidate construction element.

For example, target building design image data including a plurality of candidate building construction elements, such as windows, doors, walls, and the like. The construction task optimization model may be used to perform feature extraction on the target building design image data to generate building mapping features for each candidate building construction element. These building mapping features may represent the shape, size, location, etc. of the candidate building construction elements.

After generating the building mapping features of each candidate building construction element, the construction task optimization model may be used to spatially transform the building mapping features of each candidate building construction element to generate a building space vector for each candidate building construction element. These building space vectors may represent the location and orientation of candidate building construction elements in three-dimensional space.

The construction mapping characteristics of each candidate construction element may be used to determine the segment amounts of the candidate construction video segments corresponding to each candidate construction element. For example, if one candidate building construction element is a window, it may be determined that the segment count of the candidate construction video segment corresponding to this window is 2, because the window has two portions, one being the left portion of the window and the other being the right portion of the window.

Step S180 may include:

and expanding the building space vector of each candidate building construction element according to the construction task optimization model and the segment quantity of the candidate construction video segment corresponding to each candidate building construction element, and generating the building space vector of each candidate construction video segment.

For example, the construction task optimization model may be used to expand the construction space vector of each candidate construction element according to the segment quantity of the candidate construction video segment corresponding to each candidate construction element, and generate the construction space vector of each candidate construction video segment. For example, if one candidate building construction element is a window, the construction task optimization model may be used to expand the building space vector of the window according to the segment number of the candidate construction video segment corresponding to the window, to generate the building space vector of the window.

Step S190 may include: and carrying out space conversion on the building space vectors of the candidate construction video segments according to the construction task optimization model to generate construction video content vectors of the candidate construction video segments, and carrying out feature reduction on the construction video content vectors of the candidate construction video segments according to the construction task optimization model to generate target candidate construction tasks.

For example, assume that a plurality of candidate building construction elements, e.g., windows, doors, walls, etc., are included in the target building design image data. The construction task optimization model may be used to perform feature extraction on the target building design image data to generate building mapping features for each candidate building construction element. Then, the construction task optimization model may be used to spatially transform the building mapping features of each candidate building construction element to generate a building space vector for each candidate building construction element. And then, using a construction task optimization model to expand the construction space vectors of the candidate construction elements according to the segment quantities of the candidate construction video segments corresponding to the candidate construction elements, and generating the construction space vectors of the candidate construction video segments. Then, the construction task optimization model may be used to spatially transform the building space vectors of each candidate construction video segment, generating construction video content vectors of each candidate construction video segment. And finally, performing feature reduction on the construction video content vectors of each candidate construction video segment by using the construction task optimization model to generate a target candidate construction task.

For example, assume a window whose building space vector is (x, y, z), where x represents the width of the window, y represents the height of the window, and z represents the depth of the window. And a construction task optimization model can be used for generating a construction video content vector of the window according to the building space vector of the window. And then, performing feature restoration on the construction video content vector of the window by using the construction task optimization model to generate a target candidate construction task. This target candidate construction task may be expressed as: "in the case where the window has a width x, a height y, and a depth z, the window should be installed on a wall, and the window should be oriented south. This target candidate construction task may help determine the installation location and orientation of the window, thereby guiding the actual construction process.

Fig. 2 provides an intelligent construction task optimization system 100 based on image analysis in the embodiment of the application, which comprises a processor 1001, a memory 1003 and sensor node codes stored on the memory 1003, wherein the processor 1001 executes the sensor node codes to implement the steps of a sensing data processing method applied to intelligent photovoltaics.

The intelligent work site construction task optimization system 100 based on image analysis shown in fig. 2 includes: a processor 1001 and a memory 1003. The processor 1001 is coupled to the memory 1003, such as via a bus 1002. Optionally, the intelligent worksite construction task optimization system 100 based on image analysis may further include a transceiver 1004, where the transceiver 1004 may be used for data interaction between the server and other servers, such as transmission of data and/or reception of data, etc. It should be noted that, the transceiver 1004 in actual scheduling is not limited to one, and the structure of the intelligent construction task optimization system 100 based on image analysis is not limited to the embodiment of the present application.

The processor 1001 may be a CPU (Central Processing Unit ), general purpose processor, DSP (Digital Signal Processor, data signal processor), ASIC (Application SpecificIntegrated Circuit ), FPGA (Field Programmable Gate Array, field programmable gate array) or other programmable logic device, transistor logic device, hardware components, or any combination thereof. Which may implement or perform the various exemplary logic blocks, modules, and circuits described in connection with this disclosure. The processor 1001 may also be a combination that implements computing functionality, such as a combination comprising one or more microprocessors, a combination of a DSP and a microprocessor, or the like.

Bus 1002 may include a path to transfer information between the components. Bus 1002 may be a PCI (Peripheral Component Interconnect, peripheral component interconnect standard) bus, or EISA (ExtendedIndustry Standard Architecture ) bus, among others. The bus 1002 may be divided into an address bus, a data bus, a control bus, and the like. For ease of illustration, only one thick line is shown in fig. 2, but not only one bus or one type of bus.

The Memory 1003 may be, but is not limited to, ROM (Read Only Memory) or other type of static storage device that can store static information and instructions, RAM (Random Access Memory ) or other type of dynamic storage device that can store information and instructions, EEPROM (Electrically ErasableProgrammable Read Only Memory ), CD-ROM (Compact DiscRead Only Memory, compact disc Read Only Memory) or other optical disk storage, optical disk storage (including compact disc, laser disc, optical disc, digital versatile disc, blu-ray disc, etc.), magnetic disk storage media, other magnetic storage devices, or any other medium that can be used to carry or store sensor node code and that can be Read by a computer.

The memory 1003 is used for storing sensor node codes for executing the embodiments of the present application, and is controlled to be executed by the processor 1001. The processor 1001 is configured to execute the sensor node code stored in the memory 1003 to implement the steps shown in the foregoing method embodiment.

Wherein the server includes, but is not limited to: mobile terminals such as mobile phones, notebook computers, PADs, etc., and stationary terminals such as digital TVs, desktop computers, etc.

The embodiment of the application provides a computer readable storage medium, and the computer readable storage medium stores sensor node codes, which can realize the steps and corresponding contents of the embodiment of the method when being executed by a processor.

It should be understood that, although the flowcharts of the embodiments of the present application indicate the respective operation steps by arrows, the order of implementation of these steps is not limited to the order indicated by the arrows. In some implementations of embodiments of the present application, the implementation steps in the flowcharts may be performed in other orders as desired, unless explicitly stated herein. Furthermore, some or all of the steps in the flowcharts may include a plurality of sub-steps or a plurality of stages, depending on the actual implementation scenario. Some or all of these sub-steps or phases may be performed at the same time, or each of these sub-steps or phases may be performed at different times, respectively. In the scenario that the execution time is different, the execution sequence of the sub-steps or stages can be flexibly configured based on requirements, which is not limited by the embodiment of the present application.

The foregoing is merely an optional implementation manner of the implementation scenario of the application, and it should be noted that, for those skilled in the art, other similar implementation manners according to the technical ideas of the application are adopted without departing from the technical ideas of the application, and also belong to the protection scope of the embodiments of the application.

Claims

1. An intelligent construction task optimization method based on image analysis is characterized by comprising the following steps:

2. The intelligent construction task optimization method based on image analysis according to claim 1, wherein the determining, according to a first neural network, building space vectors of respective first building construction elements included in the first building design image data includes:

3. The intelligent worksite construction task optimization method based on image analysis according to claim 2, characterized in that the method further comprises:

4. The intelligent construction task optimization method based on image analysis according to claim 1, wherein the determining, according to the first neural network, building space vectors of respective first construction video segments included in the first construction video data includes:

5. The intelligent construction task optimization method based on image analysis according to claim 4, wherein the first construction video data is first template construction acquisition data or attention video data of the first template construction acquisition data, and the first neural network comprises a first feature reduction unit; the method further comprises the steps of:

the first feature recovery unit comprises a first loading subunit, at least two first feature extraction subunits and a first result generation subunit, wherein any one of the first feature extraction subunits comprises at least two filter layers with the same interval weight and different filter scales, and the filter layers of different first feature extraction subunits correspond to different interval weights;

6. The intelligent construction task optimization method based on image analysis according to any one of claims 1 to 5, wherein the performing parameter update on the first neural network according to the first vector offset parameter and the second vector offset parameter to generate a construction task optimization model includes:

7. The intelligent worksite construction task optimization method based on image analysis according to claim 6, further comprising:

8. The intelligent construction task optimization method based on image analysis according to claim 6, wherein the determining the construction task optimization model according to the feature processing unit and the third feature reduction unit includes:

determining the construction task optimization model according to the target feature reduction unit and the target feature reduction unit;

the determining the construction task optimization model according to the feature processing unit and the target feature reduction unit comprises the following steps:

the step of integrating the filter layers into an integrated filter layer to generate an updated feature extraction subunit, where the step of integrating the filter layers into an integrated filter layer includes:

9. The intelligent worksite construction task optimization method based on image analysis according to claim 1, further comprising:

the determining, according to a construction task optimization model, a building space vector of each candidate building construction element included in the target building design image data includes:

the method further comprises the steps of:

generating a target candidate construction task according to the construction task optimization model and the building space vector of each candidate construction video segment, wherein the target candidate construction task comprises the following steps:

10. An image analysis-based intelligent worksite construction task optimization system, comprising a processor and a machine-readable storage medium having stored therein machine-executable instructions loaded and executed by the processor to implement the image analysis-based intelligent worksite construction task optimization method of any one of claims 1-9.