CN116543322A

CN116543322A - Intelligent property routing inspection method based on community potential safety hazards

Info

Publication number: CN116543322A
Application number: CN202310560244.4A
Authority: CN
Inventors: 李玥辰
Original assignee: Shenzhen Being Perfect Community Service Technology Co ltd
Current assignee: Shenzhen Being Perfect Community Service Technology Co ltd
Priority date: 2023-05-17
Filing date: 2023-05-17
Publication date: 2023-08-04

Abstract

The invention relates to the technical field of intelligent patrol, and discloses a property intelligent patrol method based on community potential safety hazards, which comprises the following steps: generating a coverage detection path of the community initial map according to the unmanned aerial vehicle acquisition parameter set, and controlling the unmanned aerial vehicle to acquire a community point cloud sequence along the coverage detection path; object detection, semantic recognition, point cloud splicing and three-dimensional reconstruction are carried out on community point clouds in the community point cloud sequence, and a community inspection model is obtained; path planning is carried out on the community inspection model according to the inspection period to obtain an inspection path set; controlling the unmanned aerial vehicle to shoot the point positions along the patrol path set to obtain a patrol chart set; and carrying out texture enhancement, multistage feature extraction, safety analysis and safety labeling on each inspection picture in the inspection picture set to obtain a community inspection result. The invention further provides a property intelligent patrol device based on the community potential safety hazard. The invention can improve the efficiency of community safety inspection.

Description

Intelligent property routing inspection method based on community potential safety hazards

Technical Field

The invention relates to the technical field of intelligent patrol, in particular to a property intelligent patrol method based on community safety hidden danger.

Background

The community is a living area where a community is located by a community group living together, and the community property is an organization and a mechanism for managing, maintaining and serving property buildings in the community.

The existing community safety inspection technology is mainly based on manual safety inspection technology, community workers are arranged to conduct safety inspection on public facilities such as building walls, street lamps and air water pipelines in communities in a fixed time period, inspection results are recorded, in practical application, the manual safety inspection technology is used for carrying out inspection work based on long-term repeatability of the workers, high requirements are placed on physical strength, eyesight and judgment of the workers, and the efficiency of community safety inspection is low possibly.

Disclosure of Invention

The invention provides a property intelligent patrol method and device based on community safety hidden danger, and mainly aims to solve the problem of low efficiency in community safety patrol.

In order to achieve the above purpose, the invention provides a property intelligent patrol method based on community potential safety hazards, which comprises the following steps:

Acquiring a community initial map and an unmanned aerial vehicle acquisition parameter set, generating a coverage detection path of the community initial map according to the unmanned aerial vehicle acquisition parameter set, and controlling the unmanned aerial vehicle to perform point cloud coverage acquisition along the coverage detection path to obtain a community point cloud sequence;

selecting community point clouds in the community point cloud sequence one by one as target community point clouds, sequentially carrying out probability filtering, object detection and semantic identification on the target community point clouds to obtain a patrol object semantic set, marking the target community point clouds as target patrol point clouds by using the patrol object semantic set, splicing all the target patrol point clouds as community patrol point clouds, and three-dimensionally reconstructing the community patrol point clouds as a community patrol model;

splitting the patrol object in the community patrol model into a plurality of patrol period object groups according to the patrol period, performing visual field analysis on each patrol period object group to obtain a period patrol point group set, and performing path planning on the period patrol point group set according to the community patrol model to obtain a patrol path set, wherein the visual field analysis is performed on each patrol period object group to obtain the period patrol point group set, and the method comprises the following steps: selecting the object groups in the inspection period one by one as target inspection object groups, and selecting the inspection objects in the target inspection object groups one by one as target inspection objects; generating a shooting space of the target patrol object according to the focal length interval of the unmanned aerial vehicle, and selecting shooting points in the shooting space one by one as target shooting points; calculating an offset shielding coefficient of the target shooting point position by using the following shielding view algorithm:

Wherein O is _i The offset shielding coefficient of the target shooting point position i is shown, M is the total pixel number of the total shooting of the unmanned aerial vehicle, delta is shielding weight, j is the shooting pixel serial number of the unmanned aerial vehicle, v _i,j Means the unit vector of the j-th pixel ray in all pixel rays from the target shooting point i to the surface of the target inspection object, wherein the unit vector is a vector dot product symbol, and p _i,j The method comprises the steps that shielding coefficients of a j-th pixel ray in all pixel rays from a target shooting point i to the surface of a target inspection object are referred to, wherein I is a vector modulo symbol, and w is a normal vector of the inner surface of the target inspection object; selecting the target shooting point with the smallest offset shielding coefficient as the inspection point of the target inspection object, collecting all the inspection points corresponding to the target inspection object group into a period inspection point group, and collecting all the period inspection point groupsIntegrating a period inspection point group set;

selecting the patrol paths in the patrol path set one by one as target patrol paths according to the patrol periods, controlling the unmanned aerial vehicle to carry out point location shooting along the target patrol paths to obtain period patrol chart groups, and collecting all period patrol chart groups into a patrol chart set;

Selecting the inspection pictures of the inspection atlas one by one as target inspection pictures, sequentially carrying out texture enhancement and multistage feature extraction operation on the target inspection pictures to obtain an enhancement feature set, carrying out safety analysis on the target inspection pictures according to the enhancement feature set to obtain object safety semantics, and carrying out safety labeling on the community inspection model by utilizing all the object safety semantics to obtain a community inspection result.

Optionally, the generating the coverage detection path of the community initial map according to the unmanned aerial vehicle collection parameter set includes:

calculating the acquisition length and the acquisition width according to the unmanned aerial vehicle acquisition parameter set by using the following coverage projection algorithm:

wherein R is ₁ Refers to the acquisition width, R ₂ The unmanned aerial vehicle acquisition parameter set comprises an acquisition length, a lifting symbol, a longitudinal point cloud acquisition angle, a rolling angle, a transverse point cloud acquisition angle and a flying height, wherein the acquisition length is the acquisition length, the lifting symbol is an upward rounding symbol, θ is a longitudinal point cloud acquisition angle in the unmanned aerial vehicle acquisition parameter set, α is a rolling angle in the unmanned aerial vehicle acquisition parameter set, β is a transverse point cloud acquisition angle in the unmanned aerial vehicle acquisition parameter set, and d is a flying height in the unmanned aerial vehicle acquisition parameter set;

generating an acquisition block according to the acquisition length and the acquisition width;

And performing coverage scanning on the initial map of the community according to a preset overlapping step length and the acquisition block to obtain a coverage detection path.

Optionally, the sequentially performing probability filtering, object detection and semantic identification operations on the target community point cloud to obtain a patrol object semantic set, including:

probability filtering is carried out on the target community point cloud by using a preset Gaussian density algorithm, and a target filtering point cloud is obtained;

sequentially performing object detection and object segmentation operation on the target filtering point cloud to obtain an object point cloud set;

and carrying out semantic recognition operation on object point clouds in the object point clouds one by one to obtain a patrol object semantic set.

Optionally, the probability filtering of the target community point cloud by using a preset gaussian density algorithm to obtain a target filtered point cloud includes:

performing median filtering on the target community point cloud to obtain a target denoising point cloud;

extracting geometric features of the target denoising point cloud to obtain a point cloud geometric feature set;

calculating the point cloud probability of each point cloud in the target denoising point cloud according to the point cloud geometric feature set by using a preset Gaussian density algorithm:

wherein N (x) refers to a point cloud probability of a point cloud x in the target denoising point cloud, pi is a perimeter rate, N is a feature dimension of each point cloud geometric feature in the point cloud geometric feature set, det () is a determinant symbol, Σ _x Is Gao Sixie variance matrix of the point cloud geometric features of the point cloud geometric feature collection point cloud x, exp is an exponential function symbol, mu _x The Gaussian mean vector of the point cloud geometrical characteristics concentrated point cloud x is referred to, and T is a transposed symbol;

and carrying out probability threshold filtering on the target denoising point cloud according to the point cloud probability to obtain a target filtering point cloud.

Optionally, the three-dimensionally reconstructing the community inspection point cloud into a community inspection model includes:

triangulating the community inspection point cloud into a triangular community model;

extracting a triangular vertex set and a triangular surface set from the triangular community model, and carrying out matching recombination on the triangular vertex set and the triangular surface set to obtain a triangular information set;

calculating a reconstruction triangle vertex of each triangular surface in the triangle community model according to the triangle information set;

performing triangular reconstruction on the triangular community model according to the reconstructed triangular vertex to obtain a reconstructed community model;

and carrying out surface smoothing operation on the reconstructed community model to obtain a community inspection model.

Optionally, the performing path planning on the period inspection point set according to the community inspection model to obtain an inspection path set includes:

Selecting a time period inspection point group in the time period inspection point group set one by one as a target inspection point group, and generating a start inspection point and an end inspection point of the target inspection point group according to the inventory coordinates of the unmanned aerial vehicle;

calculating the inventory distance between each inspection point in the target inspection point group and the inventory coordinate;

sequencing all the inspection points in the target inspection point group according to the order from small to large of the inventory distance to obtain an initial inspection point sequence;

inserting the start patrol point into the head of the initial patrol point, and inserting the end patrol point into the tail of the initial patrol point sequence to obtain a target patrol point sequence;

performing path planning on the target inspection point sequence according to the community inspection model to obtain an initial inspection path;

and carrying out path optimization on the initial inspection path to obtain a target inspection path, and converging all the target inspection paths into an inspection path set.

Optionally, the performing texture enhancement and multi-level feature extraction operations on the target inspection picture in sequence to obtain an enhanced feature set includes:

carrying out Gaussian filtering on the target inspection picture to obtain a target filtering picture;

Performing texture erosion operation on the target filter picture to obtain a target inspection texture, and performing image segmentation operation on the target filter picture by using the target inspection texture to obtain a target texture picture;

generating a gray level histogram of the target texture picture, and carrying out gray level enhancement on the target texture picture by using the gray level histogram to obtain a target enhanced picture;

and respectively extracting texture features, color features and shape features of the target enhanced picture, and collecting the texture features, the color features and the shape features into an enhanced feature group.

Optionally, the performing security analysis on the target inspection picture according to the enhanced feature set to obtain object security semantics includes:

fusing the enhanced feature group into a target patrol feature by using a preset attention fusion algorithm;

taking the patrol object semantics corresponding to the target patrol picture as target object semantics, and selecting an object security model corresponding to the target object semantics from a preset security recognition model library as a target object security model;

and normalizing the target inspection characteristics by using the target object security model to obtain a target security code, and taking the interval semantics of the target security code as the object security semantics of the target inspection picture.

Optionally, the fusing the enhancement feature set into the target patrol feature by using a preset attention fusion algorithm includes:

performing global dimension reduction operation on the texture features in the enhancement feature group to obtain standard texture features;

performing global dimension reduction operation on the color features in the enhancement feature group to obtain standard color features;

performing global dimension reduction operation on the shape features in the enhancement feature group to obtain standard shape features;

fusing the standard texture features, the standard color features and the standard shape features into target inspection features by using the following attention fusion algorithm:

wherein A refers to the target inspection characteristic, softmax is a normalization function,omega, sigma are the fusion weight matrix of the attention fusion algorithm, Q is the standard texture feature, K is the standard color feature, T is a transposed symbol, f is a feature dimension, and the feature dimensions of the standard texture feature, the standard color feature, and the standard shape feature are equal, V is the standard shape feature.

In order to solve the problems, the invention also provides a property intelligent patrol device based on community safety hazards, which comprises:

The system comprises a point cloud acquisition module, a control module and a control module, wherein the point cloud acquisition module is used for acquiring a community initial map and an unmanned aerial vehicle acquisition parameter set, generating a coverage detection path of the community initial map according to the unmanned aerial vehicle acquisition parameter set, and controlling the unmanned aerial vehicle to perform point cloud coverage acquisition along the coverage detection path to obtain a community point cloud sequence;

the three-dimensional reconstruction module is used for selecting community point clouds in the community point cloud sequence one by one as target community point clouds, sequentially carrying out probability filtering, object detection and semantic recognition operation on the target community point clouds to obtain a patrol object semantic set, marking the target community point clouds as target patrol point clouds by using the patrol object semantic set, splicing all the target patrol point clouds into community patrol point clouds, and three-dimensionally reconstructing the community patrol point clouds into a community patrol model;

the path planning module is used for splitting the patrol objects in the community patrol model into a plurality of patrol period object groups according to the patrol period, performing visual field analysis on each patrol period object group to obtain a period patrol point group set, and performing path planning on the period patrol point group set according to the community patrol model to obtain a patrol path set, wherein the visual field analysis is performed on each patrol period object group to obtain the period patrol point group set, and the path planning module comprises the following steps: selecting the object groups in the inspection period one by one as target inspection object groups, and selecting the inspection objects in the target inspection object groups one by one as target inspection objects; generating a shooting space of the target patrol object according to the focal length interval of the unmanned aerial vehicle, and selecting shooting points in the shooting space one by one as target shooting points; calculating an offset shielding coefficient of the target shooting point position by using the following shielding view algorithm:

Wherein O is _i The offset shielding coefficient of the target shooting point position i is shown, M is the total pixel number of the total shooting of the unmanned aerial vehicle, delta is shielding weight, j is the shooting pixel serial number of the unmanned aerial vehicle, v _i,j Means the unit vector of the j-th pixel ray in all pixel rays from the target shooting point i to the surface of the target inspection object, wherein the unit vector is a vector dot product symbol, and p _i,j The method comprises the steps that shielding coefficients of a j-th pixel ray in all pixel rays from a target shooting point i to the surface of a target inspection object are referred to, wherein I is a vector modulo symbol, and w is a normal vector of the inner surface of the target inspection object; selecting the target shooting point with the smallest offset shielding coefficient as a patrol point of the target patrol object, collecting all patrol points corresponding to the target patrol object group into a time period patrol point group, and collecting all time period patrol point groups into a time period patrol point group set;

the inspection shooting module is used for selecting inspection paths in the inspection path set one by one as target inspection paths according to the inspection time period, controlling the unmanned aerial vehicle to carry out point location shooting along the target inspection paths to obtain time period inspection chart sets, and collecting all the time period inspection chart sets into an inspection chart set;

The semantic analysis module is used for selecting the inspection pictures of the inspection atlas one by one as target inspection pictures, sequentially carrying out texture enhancement and multi-level feature extraction operation on the target inspection pictures to obtain an enhancement feature set, carrying out safety analysis on the target inspection pictures according to the enhancement feature set to obtain object safety semantics, and carrying out safety labeling on the community inspection model by utilizing all the object safety semantics to obtain a community inspection result.

The embodiment of the invention generates the coverage detection path of the community initial map according to the acquisition parameter set of the unmanned aerial vehicle by acquiring the community initial map and the acquisition parameter set of the unmanned aerial vehicle, controls the unmanned aerial vehicle to carry out point cloud coverage acquisition along the coverage detection path to obtain a community point cloud sequence, can acquire the global point cloud sequence of the community by utilizing an unmanned aerial vehicle-mounted point cloud scanner, thereby facilitating subsequent path planning and intelligent inspection, sequentially carries out probability filtering, object detection and semantic recognition operation on the target community point clouds by selecting the community point clouds in the community point cloud sequence one by one as target community point clouds, obtains the object semantic set for inspection, can enhance the details of the point clouds, marks the object needing security inspection in the target community point clouds, marks the target community point clouds into target inspection point clouds by utilizing the object semantic set for inspection, all target inspection point clouds are spliced into community inspection point clouds, the community inspection point clouds are three-dimensionally reconstructed into a community inspection model, a three-dimensional model of a community to be inspected safely can be reconstructed, avoidance and path planning of obstacles are convenient to follow for each object needing to be inspected safely, inspection objects in the community inspection model are split into a plurality of inspection period object groups according to inspection periods, visual field analysis is conducted on each inspection period object group to obtain period inspection point group sets, most suitable shooting points can be selected for each inspection object to obtain a period inspection chart set with better quality, path planning is conducted on the period inspection point group sets according to the community inspection model to obtain an inspection path set, and shooting paths which pass through all the shooting points and can be automatically restored can be generated, therefore, the inspection object in the inspection period is photographed at one time, and the inspection speed is improved.

The inspection path in the inspection path set is selected one by one according to the inspection time period to serve as a target inspection path, the unmanned aerial vehicle is controlled to conduct point location shooting along the target inspection path to obtain time period inspection image groups, all the time period inspection image groups are assembled into an inspection image set, the inspection image sets can be shot and sampled in the time period when each object to be inspected is easiest to observe, accordingly the accuracy of intelligent inspection is improved, the inspection images of the inspection image sets are selected one by one to serve as target inspection images, the target inspection images are sequentially subjected to texture enhancement and multistage feature extraction operation to obtain enhancement feature groups, the target inspection images are subjected to safety analysis according to the enhancement feature groups to obtain object safety semantics, the community inspection model is subjected to safety labeling by utilizing all the object safety semantics to obtain community inspection results, safety analysis can be achieved according to multiple features of each inspection object, the safety analysis accuracy is improved, and the community safety inspection efficiency is improved. Therefore, the intelligent property inspection method and device based on the community safety hidden danger can solve the problem of low efficiency in community safety inspection.

Drawings

FIG. 1 is a schematic flow chart of a real estate intelligent patrol method based on community safety hazards according to an embodiment of the present invention;

FIG. 2 is a schematic flow chart of a method for reconstructing a community inspection model according to an embodiment of the present invention;

FIG. 3 is a flowchart illustrating a method for generating a set of inspection paths according to an embodiment of the present invention;

FIG. 4 is a functional block diagram of a real estate intelligent patrol device based on community safety hazards according to an embodiment of the present invention;

the achievement of the objects, functional features and advantages of the present invention will be further described with reference to the accompanying drawings, in conjunction with the embodiments.

Detailed Description

It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the invention.

The embodiment of the application provides a property intelligent patrol method based on community potential safety hazards. The execution main body of the intelligent property inspection method based on the community safety hidden danger comprises at least one of electronic equipment which can be configured to execute the method provided by the embodiment of the application, such as a server side and a terminal. In other words, the intelligent property patrol method based on the community safety hidden trouble can be executed by software or hardware installed in the terminal equipment or the server equipment, and the software can be a blockchain platform. The service end includes but is not limited to: a single server, a server cluster, a cloud server or a cloud server cluster, and the like. The server may be an independent server, or may be a cloud server that provides cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (Content Delivery Network, CDN), and basic cloud computing services such as big data and artificial intelligence platforms.

Referring to fig. 1, a flow chart of a real estate intelligent patrol method based on community safety hazards according to an embodiment of the present invention is shown. In this embodiment, the intelligent property inspection method based on community safety hazards includes:

s1, acquiring a community initial map and an unmanned aerial vehicle acquisition parameter set, generating a coverage detection path of the community initial map according to the unmanned aerial vehicle acquisition parameter set, and controlling the unmanned aerial vehicle to perform point cloud coverage acquisition along the coverage detection path to obtain a community point cloud sequence.

In the embodiment of the invention, the community initial map refers to a regional map of a community to be inspected, the community initial map can be a two-dimensional plan view or a three-dimensional model view with a scale, the unmanned aerial vehicle acquisition parameter set refers to various parameters during unmanned aerial vehicle flight and monitoring, such as flight height, a point cloud acquisition angle and a rolling angle of the unmanned aerial vehicle when the unmanned aerial vehicle performs point cloud acquisition, wherein the point cloud acquisition angle comprises a transverse point cloud acquisition angle and a longitudinal point cloud acquisition angle, the transverse point cloud acquisition angle refers to an included angle between left and right views of a point cloud scanner of the unmanned aerial vehicle, the longitudinal point cloud acquisition angle refers to an included angle between upper and lower views of the point cloud scanner of the unmanned aerial vehicle, and the rolling angle refers to an angle of the unmanned aerial vehicle rotating around a front axis and a rear axis of the aircraft.

In the embodiment of the present invention, the generating the coverage detection path of the community initial map according to the unmanned aerial vehicle acquisition parameter set includes:

According to the embodiment of the invention, the coverage projection algorithm is utilized to calculate the acquisition length and the acquisition width according to the unmanned aerial vehicle acquisition parameter set, so that the single coverage range of the unmanned aerial vehicle in the process of acquiring the community point cloud can be determined, and the coverage detection path can be conveniently planned.

In detail, the generating the collection block according to the collection length and the collection width refers to generating a rectangular area with the collection length as the length and the collection width as the width, and taking the rectangular area as the collection area.

Specifically, performing coverage scanning on the initial map of the community according to a preset overlapping step length and the acquisition block to obtain a coverage detection path refers to generating an acquisition bit length according to the overlapping step length and the acquisition block, performing coverage scanning on the initial map of the community according to the acquisition bit length by using a scanning line mode to obtain an initial detection path, and lifting Gao Dudi of the initial detection path to the flying height to obtain the coverage detection path.

In detail, the controlling the unmanned aerial vehicle to perform point cloud coverage acquisition along the coverage detection path to obtain a community point cloud sequence refers to controlling the unmanned aerial vehicle to fly along the coverage detection path, and performing point cloud scanning on a community by using a point cloud scanner on the unmanned aerial vehicle when the unmanned aerial vehicle is positioned at the central point of each acquisition area in the coverage detection path to obtain the community point cloud sequence.

In the embodiment of the invention, the community initial map and the unmanned aerial vehicle acquisition parameter set are acquired, the coverage detection path of the community initial map is generated according to the unmanned aerial vehicle acquisition parameter set, the unmanned aerial vehicle is controlled to carry out point cloud coverage acquisition along the coverage detection path to obtain the community point cloud sequence, and the unmanned aerial vehicle-mounted point cloud scanner can be utilized to acquire the global point cloud sequence of the community, so that the follow-up path planning and intelligent inspection are convenient.

S2, selecting community point clouds in the community point cloud sequence one by one as target community point clouds, sequentially carrying out probability filtering, object detection and semantic recognition operation on the target community point clouds to obtain a patrol object semantic set, marking the target community point clouds as target patrol point clouds by using the patrol object semantic set, splicing all the target patrol point clouds into community patrol point clouds, and three-dimensionally reconstructing the community patrol point clouds into a community patrol model.

In the embodiment of the invention, the inspection object semantic set is a set formed by a plurality of inspection object semantics, and the inspection object semantic set comprises inspection object semantics of all object point clouds in the target community point cloud, wherein the inspection object semantics are important objects needing to be subjected to potential safety hazard inspection, such as street lamps, building walls, fire-fighting facilities, water meters and the like.

In the embodiment of the invention, probability filtering, object detection and semantic identification operations are sequentially performed on the target community point cloud to obtain a patrol object semantic set, which comprises the following steps:

In the embodiment of the present invention, the sequentially performing object detection and object segmentation operations on the target filtering point clouds to obtain an object point cloud set refers to splitting each point cloud into different object point clouds by using algorithms such as a support vector machine (Support Vector Machine, abbreviated as SVM) and a random forest according to the point cloud geometric feature of each point cloud in the target filtering point clouds, and collecting all object point clouds into the object point cloud set.

In detail, the probability filtering is performed on the target community point cloud by using a preset gaussian density algorithm to obtain a target filtered point cloud, which comprises the following steps:

In the embodiment of the present invention, the extracting geometric features of the target denoising point cloud to obtain a point cloud geometric feature set refers to extracting features such as a point cloud density, a point cloud coordinate, a point cloud normal vector, a point cloud curvature, a point cloud color, etc. of each point cloud in the target denoising point cloud, and integrating the features such as the point cloud density, the point cloud coordinate, the point cloud normal vector, the point cloud curvature, the point cloud color, etc. into a point cloud geometric feature, and integrating the point cloud geometric features of all the point clouds in the target denoising point cloud into a point cloud geometric feature set.

In the embodiment of the invention, the point cloud probability of each point cloud in the target denoising point cloud is calculated according to the point cloud geometric feature set by utilizing the Gaussian density algorithm, and the point cloud probability of each point cloud can be calculated by combining the Gaussian distribution probability of the point cloud geometric feature of each point cloud in the target denoising point cloud, so that the detection of outlier point clouds and abnormal point clouds is realized.

Specifically, the performing probability threshold filtering on the target denoising point cloud according to the point cloud probability to obtain a target filtering point cloud refers to screening the point cloud with the point cloud probability smaller than a preset probability threshold as an outlier point cloud from the target denoising point cloud to obtain the target filtering point cloud.

In detail, a deep learning algorithm such as a support vector machine (Support Vector Machine, abbreviated as SVM) and a convolutional neural network (Convolutional Neural Network, abbreviated as CNN) can be utilized to perform semantic recognition operation on object point clouds in the object point clouds one by one, so as to obtain a patrol object semantic set, wherein the patrol object semantic set contains semantics of objects needing to be subjected to safe patrol, such as street lamps, well covers, water meters, building walls, traffic lights and the like.

In detail, marking the target community point cloud as a target patrol point cloud by using the patrol object semantic set refers to marking the corresponding object point cloud in the target community point cloud by using each patrol object semantic in the patrol object semantic set to obtain the target patrol point cloud; the step of splicing all the target patrol point clouds into community patrol point clouds refers to splicing all the target patrol point clouds according to the point cloud sequence of the community point cloud sequence, the structure of the point clouds and the overlapping step length, so as to obtain the community patrol point clouds.

Specifically, referring to fig. 2, the three-dimensionally reconstructing the community inspection point cloud into a community inspection model includes:

s21, triangulating the community inspection point cloud into a triangular community model;

s22, extracting a triangular vertex set and a triangular surface set from the triangular community model, and carrying out matching recombination on the triangular vertex set and the triangular surface set to obtain a triangular information set;

s23, calculating a reconstruction triangle vertex of each triangle surface in the triangle community model according to the triangle information set;

s24, performing triangular reconstruction on the triangular community model according to the reconstructed triangular vertex to obtain a reconstructed community model;

S25, performing surface smoothing operation on the reconstructed community model to obtain a community inspection model.

In detail, the community inspection point cloud can be triangulated into a triangular community model by using a triangulating algorithm such as Delaunay triangulating algorithm, voronoi triangulating algorithm or Alpha-shape triangulating algorithm, wherein the triangular vertex set refers to a set formed by vertexes of each triangular surface in the triangular community model, and the triangular surface set refers to a set formed by each triangular surface in the triangular community model; and matching and reorganizing the triangular vertex set and the triangular surface set to obtain a triangular information set, namely matching and reorganizing the triangular vertices of each triangular surface to form triangular information, and collecting all the triangular information to form the triangular information set.

Specifically, a reconstruction algorithm such as Poisson reconstruction, marching cube reconstruction, ball pivot reconstruction and the like can be utilized to calculate a reconstruction triangle vertex of each triangle surface in the triangle community model according to the triangle information set, and the triangle community model is subjected to triangle reconstruction according to the reconstruction triangle vertex to obtain a reconstruction community model; and carrying out surface smoothing operation on the reconstructed community model by using Laplacian smoothing and mean value curvature flow smoothing to obtain a community inspection model.

According to the embodiment of the invention, the community point clouds in the community point cloud sequence are selected one by one to serve as target community point clouds, probability filtering, object detection and semantic recognition operations are sequentially carried out on the target community point clouds to obtain the patrol object semantic set, the details of the point clouds can be enhanced, objects needing to be subjected to safety patrol in the target community point clouds are marked, the target community point clouds are marked as target patrol point clouds by utilizing the patrol object semantic set, all the target patrol point clouds are spliced to form community patrol point clouds, the community patrol point clouds are three-dimensionally reconstructed to form a community patrol model, a three-dimensional model of a community to be subjected to safety patrol can be reconstructed, and obstacle avoidance and path planning can be conveniently carried out on each object needing to be subjected to safety patrol in the follow-up process.

S3, splitting the patrol objects in the community patrol model into a plurality of patrol period object groups according to the patrol period, performing visual field analysis on each patrol period object group to obtain a period patrol point group set, and performing path planning on the period patrol point group set according to the community patrol model to obtain a patrol path set.

In the embodiment of the invention, the inspection period is a preset period of time convenient for property inspection, for example, a period of time from 11 pm to 12 pm and a period of time from 18 pm to 19 pm, and the inspected object is an object model of a part of the point cloud, which is marked with the semantics of the inspected object, in the target inspected point cloud.

In the embodiment of the invention, the patrol objects in the community patrol model are divided into a plurality of patrol period object groups according to patrol period, for example, objects such as building wall surfaces, fire-fighting facilities, water meters and the like are watched in a period from 11 pm to 12 pm, and objects such as street lamps, traffic lamps and the like are watched in a period from 18 pm to 19 pm

In the embodiment of the present invention, the visual field analysis is performed on each inspection time period object group to obtain a time period inspection point group set, including:

selecting the object groups in the inspection period one by one as target inspection object groups, and selecting the inspection objects in the target inspection object groups one by one as target inspection objects;

generating a shooting space of the target patrol object according to the focal length interval of the unmanned aerial vehicle, and selecting shooting points in the shooting space one by one as target shooting points;

calculating an offset shielding coefficient of the target shooting point position by using the following shielding view algorithm:

wherein O is _i The offset shielding coefficient of the target shooting point position i is shown, M is the total pixel number of the total shooting of the unmanned aerial vehicle, delta is shielding weight, j is the shooting pixel serial number of the unmanned aerial vehicle, v _i,j Means the unit vector of the j-th pixel ray in all pixel rays from the target shooting point i to the surface of the target inspection object, wherein the unit vector is a vector dot product symbol, and p _i,j The method refers to the shielding coefficient of the j-th pixel ray in all pixel rays from the target shooting point i to the surface of the target patrol object, wherein the I is a vector modulo symbol, and w refers to the target patrol objectA normal vector of the inner surface;

and selecting the target shooting point with the smallest offset shielding coefficient as the inspection point of the target inspection object, collecting all the inspection points corresponding to the target inspection object group into a time period inspection point group, and collecting all the time period inspection point groups into a time period inspection point group set.

In detail, the focal length section refers to a shooting distance section in which the unmanned aerial vehicle can focus, the shooting space refers to a spherical wall space which is formed by taking the center of the outer surface of the target inspection object as an origin and taking the focal length section as a radius and is not overlapped with other objects, and the shooting point position refers to a point position space in which the unmanned aerial vehicle shoots and occupies a position in the shooting space.

In the embodiment of the invention, the ratio of the shielding object in each shooting point position picture and the offset degree of the shooting point position from the center line of the inspection object can be calculated by calculating the offset shielding coefficient of the target shooting point position by using the shielding view algorithm, so that the advantages and disadvantages of the shooting pictures are measured.

In detail, referring to fig. 3, the performing path planning on the period inspection point set according to the community inspection model to obtain an inspection path set includes:

s31, selecting a time period inspection point group in the time period inspection point group set one by one as a target inspection point group, and generating a start inspection point and an end inspection point of the target inspection point group according to the inventory coordinates of the unmanned aerial vehicle;

s32, calculating the inventory distance between each inspection point in the target inspection point group and the inventory coordinate;

s33, sequencing all the inspection points in the target inspection point group according to the order from small to large of the inventory distance to obtain an initial inspection point sequence;

s34, inserting the start patrol point into the head of the initial patrol point, and inserting the end patrol point into the tail of the initial patrol point sequence to obtain a target patrol point sequence;

s35, carrying out path planning on the target inspection point sequence according to the community inspection model to obtain an initial inspection path;

s36, carrying out path optimization on the initial inspection path to obtain a target inspection path, and collecting all the target inspection paths into an inspection path set.

In detail, the inventory coordinates refer to position coordinates of the unmanned aerial vehicle during the flight stopping and safety inspection, and the generating of the start inspection point and the end inspection point of the target inspection point group according to the inventory coordinates of the unmanned aerial vehicle refers to taking the point where the inventory coordinates are located as the start inspection point and the end inspection point of the target inspection point group.

Specifically, the inventory distance between each inspection point in the target inspection point group and the inventory coordinates can be calculated by using a euclidean distance algorithm or a manhattan distance algorithm, an algorithm such as an algorithm A, a Dijkstra algorithm or an ant colony algorithm can be used for avoiding obstacles in the community inspection model, an initial inspection path sequentially passing through the target inspection point sequence is generated, and the path of the initial inspection path can be optimized by using a genetic algorithm or a simulated annealing algorithm to obtain the target inspection path.

In the embodiment of the invention, the patrol objects in the community patrol model are split into a plurality of patrol period object groups according to the patrol period, the visual field analysis is carried out on each patrol period object group to obtain a period patrol point group set, the most suitable shooting point can be selected for each patrol object to obtain a period patrol chart set with better quality, and the path planning is carried out on the period patrol point group set according to the community patrol model to obtain a patrol path set, so that shooting paths which pass through all the shooting points and can be automatically reset can be generated, the patrol objects in the patrol period can be shot at one time, and the patrol speed is improved.

S4, selecting the patrol paths in the patrol path set one by one as target patrol paths according to the patrol periods, controlling the unmanned aerial vehicle to carry out point location shooting along the target patrol paths to obtain period patrol image groups, and collecting all the period patrol image groups into a patrol image set.

In the embodiment of the present invention, the step of selecting the routing inspection paths in the routing inspection path set one by one as the target routing inspection path according to the routing inspection period refers to the step of selecting the routing inspection paths in the routing inspection path set one by one as the target routing inspection path according to the time sequence of the routing inspection period.

Specifically, the controlling the unmanned aerial vehicle to perform point location shooting along the target inspection path to obtain a time period inspection image group refers to controlling the unmanned aerial vehicle to fly along the target inspection path in an inspection time period corresponding to the target inspection path, shooting an inspection object at the position of each inspection point to obtain time period inspection images, and collecting all the time period inspection images in the target inspection path into the time period inspection image group.

In the embodiment of the invention, the patrol paths in the patrol path set are selected one by one as the target patrol paths according to the patrol periods, the unmanned aerial vehicle is controlled to carry out point location shooting along the target patrol paths to obtain period patrol image groups, and all the period patrol image groups are collected into a patrol image set.

According to the embodiment of the invention, the patrol paths in the patrol path set are selected one by one as the target patrol paths according to the patrol periods, the unmanned aerial vehicle is controlled to carry out point location shooting along the target patrol paths to obtain the period patrol chart group, all the period patrol chart groups are assembled into the patrol chart set, shooting and sampling can be carried out in the period in which each object to be patrol is most easily observed, and therefore the intelligent patrol accuracy is improved.

S5, selecting the inspection pictures of the inspection atlas one by one as target inspection pictures, sequentially carrying out texture enhancement and multistage feature extraction operation on the target inspection pictures to obtain an enhancement feature set, carrying out safety analysis on the target inspection pictures according to the enhancement feature set to obtain object safety semantics, and carrying out safety labeling on the community inspection model by utilizing all the object safety semantics to obtain a community inspection result.

In the embodiment of the present invention, the performing texture enhancement and multi-level feature extraction operations on the target inspection image in sequence to obtain an enhanced feature set includes:

Specifically, texture erosion operation can be performed on the target filtered picture by using a sobel operator or a canny operator to obtain a target inspection texture, image segmentation operation is performed on the target filtered picture by using the target inspection texture to obtain a target texture picture, namely, a texture mask is generated according to the target inspection texture, and image segmentation operation is performed on the target filtered picture according to the texture mask to obtain the target texture picture.

In detail, the texture features of the target enhanced picture can be extracted by using a gray level co-occurrence matrix method or a Gabor filtering algorithm; obtaining color characteristics by generating a color histogram of the target enhanced picture; and convoluting the target texture picture or extracting a skeleton of the target texture picture by utilizing a multi-convolution residual error network to obtain the shape characteristic.

Specifically, the performing security analysis on the target inspection picture according to the enhanced feature set to obtain object security semantics includes:

In detail, the safety recognition model library is a model library formed by a plurality of pre-trained object safety models, each object safety model is a safety analysis model obtained by training a plurality of inspection features of an inspected object, which are marked with safety semantics, and the safety analysis model can be a neural network model such as a support vector machine (Support Vector Machine, SVM for short), a convolutional neural network (Convolutional Neural Network, CNN for short) or a transducer model.

Specifically, the normalizing the target inspection feature by using the target object security model to obtain a target security code refers to normalizing the target inspection feature by using a softmax layer or a linear normalization layer of the target object security model to obtain the target security code.

In detail, the interval semantics refers to the coding semantics interval of the target object security model, for example, an interval of 0 to 0.60 is security, an interval of 0.60 to 0.80 is existence of small potential safety hazard, and an interval of 0.80 to 1 is existence of large potential safety hazard.

In detail, the fusing the enhancement feature set into the target inspection feature by using a preset attention fusion algorithm includes:

Specifically, performing global dimension reduction operation on the texture features in the enhancement feature group to obtain standard texture features; performing global dimension reduction operation on the color features in the enhancement feature group to obtain standard color features; and performing global dimension reduction operation on the shape features in the enhancement feature group to obtain standard shape features, wherein the step of reducing dimensions of Chinese texture features, color features and shape features of the enhancement feature group to the same feature dimension and feature structure by using a global pooling layer.

In detail, the standard texture features, the standard color features and the standard shape features are fused into the target inspection features by using the following attention fusion algorithm, so that attention fusion among a plurality of features can be realized, the inherent structural relation of each feature is reserved, and the calculation dimension is reduced.

According to the embodiment of the invention, the patrol pictures of the patrol atlas are selected one by one to serve as target patrol pictures, the target patrol pictures are sequentially subjected to texture enhancement and multistage feature extraction operation to obtain the enhancement feature group, the target patrol pictures are subjected to safety analysis according to the enhancement feature group to obtain object safety semantics, the community patrol model is subjected to safety marking by utilizing all the object safety semantics to obtain community patrol results, safety analysis can be realized according to a plurality of features of each patrol object, the accuracy of the safety analysis is improved, and the community safety patrol efficiency is further improved.

The inspection path in the inspection path set is selected one by one according to the inspection time period to serve as a target inspection path, the unmanned aerial vehicle is controlled to conduct point location shooting along the target inspection path to obtain time period inspection image groups, all the time period inspection image groups are assembled into an inspection image set, the inspection image sets can be shot and sampled in the time period when each object to be inspected is easiest to observe, accordingly the accuracy of intelligent inspection is improved, the inspection images of the inspection image sets are selected one by one to serve as target inspection images, the target inspection images are sequentially subjected to texture enhancement and multistage feature extraction operation to obtain enhancement feature groups, the target inspection images are subjected to safety analysis according to the enhancement feature groups to obtain object safety semantics, the community inspection model is subjected to safety labeling by utilizing all the object safety semantics to obtain community inspection results, safety analysis can be achieved according to multiple features of each inspection object, the safety analysis accuracy is improved, and the community safety inspection efficiency is improved. Therefore, the intelligent property inspection method based on the community safety hidden danger can solve the problem of low efficiency in community safety inspection.

Fig. 4 is a functional block diagram of a real estate intelligent patrol device based on community safety hazards according to an embodiment of the present invention.

The intelligent property inspection device 100 based on community safety hazards can be installed in electronic equipment. According to the functions implemented, the intelligent property inspection device 100 based on the community safety hidden trouble may include a point cloud acquisition module 101, a three-dimensional reconstruction module 102, a path planning module 103, an inspection shooting module 104 and a semantic analysis module 105. The module of the invention, which may also be referred to as a unit, refers to a series of computer program segments, which are stored in the memory of the electronic device, capable of being executed by the processor of the electronic device and of performing a fixed function.

In the present embodiment, the functions concerning the respective modules/units are as follows:

the point cloud acquisition module 101 is configured to acquire a community initial map and an unmanned aerial vehicle acquisition parameter set, generate a coverage detection path of the community initial map according to the unmanned aerial vehicle acquisition parameter set, and control the unmanned aerial vehicle to perform point cloud coverage acquisition along the coverage detection path to obtain a community point cloud sequence;

the three-dimensional reconstruction module 102 is configured to select community point clouds in the community point cloud sequence one by one as target community point clouds, sequentially perform probability filtering, object detection and semantic recognition operations on the target community point clouds to obtain a patrol object semantic set, label the target community point clouds as target patrol point clouds by using the patrol object semantic set, splice all the target patrol point clouds as community patrol point clouds, and reconstruct the community patrol point clouds into a community patrol model in a three-dimensional manner;

The path planning module 103 is configured to split an inspection object in the community inspection model into a plurality of inspection period object groups according to an inspection period, perform visual field analysis on each inspection period object group to obtain a period inspection point group set, perform path planning on the period inspection point group set according to the community inspection model to obtain an inspection path set, where the performing visual field analysis on each inspection period object group to obtain a period inspection point group set includes: selecting the object groups in the inspection period one by one as target inspection object groups, and selecting the inspection objects in the target inspection object groups one by one as target inspection objects; generating a shooting space of the target patrol object according to the focal length interval of the unmanned aerial vehicle, and selecting shooting points in the shooting space one by one as target shooting points; calculating an offset shielding coefficient of the target shooting point position by using the following shielding view algorithm:

wherein O is _i The offset shielding coefficient of the target shooting point position i is shown, M is the total pixel number of the total shooting of the unmanned aerial vehicle, delta is shielding weight, j is the shooting pixel serial number of the unmanned aerial vehicle, v _i,j Means the unit vector of the j-th pixel ray in all pixel rays from the target shooting point i to the surface of the target inspection object, wherein the unit vector is a vector dot product symbol, and p _i,j Is taken by the objectThe point position i starts to the shielding coefficient of the j-th pixel ray in all the pixel rays of the surface of the target patrol object, the I is a vector modulo symbol, and w is a normal vector of the inner surface of the target patrol object; selecting the target shooting point with the smallest offset shielding coefficient as a patrol point of the target patrol object, collecting all patrol points corresponding to the target patrol object group into a time period patrol point group, and collecting all time period patrol point groups into a time period patrol point group set;

the inspection shooting module 104 is configured to select inspection paths in the inspection path set one by one as target inspection paths according to the inspection period, control the unmanned aerial vehicle to perform point location shooting along the target inspection paths, obtain period inspection chart sets, and collect all period inspection chart sets into an inspection chart set;

the semantic analysis module 105 is configured to select the inspection pictures of the inspection atlas one by one as target inspection pictures, sequentially perform texture enhancement and multi-level feature extraction operations on the target inspection pictures to obtain an enhancement feature set, perform security analysis on the target inspection pictures according to the enhancement feature set to obtain object security semantics, and perform security labeling on the community inspection model by using all the object security semantics to obtain a community inspection result.

In detail, each module in the intelligent property inspection device 100 based on the community safety hazard in the embodiment of the present invention adopts the same technical means as the intelligent property inspection method based on the community safety hazard described in fig. 1 to 3, and can generate the same technical effects, which is not described herein.

In the several embodiments provided in the present invention, it should be understood that the disclosed apparatus, device and method may be implemented in other manners. For example, the above-described apparatus embodiments are merely illustrative, and for example, the division of the modules is merely a logical function division, and there may be other manners of division when actually implemented.

The modules described as separate components may or may not be physically separate, and components shown as modules may or may not be physical units, may be located in one place, or may be distributed over multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

In addition, each functional module in the embodiments of the present invention may be integrated in one processing unit, or each unit may exist alone physically, or two or more units may be integrated in one unit. The integrated units can be realized in a form of hardware or a form of hardware and a form of software functional modules.

It will be evident to those skilled in the art that the invention is not limited to the details of the foregoing illustrative embodiments, and that the present invention may be embodied in other specific forms without departing from the spirit or essential characteristics thereof.

The present embodiments are, therefore, to be considered in all respects as illustrative and not restrictive, the scope of the invention being indicated by the appended claims rather than by the foregoing description, and all changes which come within the meaning and range of equivalency of the claims are therefore intended to be embraced therein. Any reference signs in the claims shall not be construed as limiting the claim concerned.

The embodiment of the application can acquire and process the related data based on the artificial intelligence technology. Among these, artificial intelligence (Artificial Intelligence, AI) is the theory, method, technique and application system that uses a digital computer or a digital computer-controlled machine to simulate, extend and extend human intelligence, sense the environment, acquire knowledge and use knowledge to obtain optimal results.

Furthermore, it is evident that the word "comprising" does not exclude other elements or steps, and that the singular does not exclude a plurality. Multiple units or means as set forth in the system embodiments may also be implemented by one unit or means in software or hardware. The terms first, second, etc. are used to denote a name, but not any particular order.

Finally, it should be noted that the above-mentioned embodiments are merely for illustrating the technical solution of the present invention and not for limiting the same, and although the present invention has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that modifications and equivalents may be made to the technical solution of the present invention without departing from the spirit and scope of the technical solution of the present invention.

Claims

1. The utility intelligent patrol method based on the potential safety hazard of the community is characterized by comprising the following steps:

s1: acquiring a community initial map and an unmanned aerial vehicle acquisition parameter set, generating a coverage detection path of the community initial map according to the unmanned aerial vehicle acquisition parameter set, and controlling the unmanned aerial vehicle to perform point cloud coverage acquisition along the coverage detection path to obtain a community point cloud sequence;

s2: selecting community point clouds in the community point cloud sequence one by one as target community point clouds, sequentially carrying out probability filtering, object detection and semantic identification on the target community point clouds to obtain a patrol object semantic set, marking the target community point clouds as target patrol point clouds by using the patrol object semantic set, splicing all the target patrol point clouds as community patrol point clouds, and three-dimensionally reconstructing the community patrol point clouds as a community patrol model;

S3: splitting the patrol object in the community patrol model into a plurality of patrol period object groups according to the patrol period, performing visual field analysis on each patrol period object group to obtain a period patrol point group set, and performing path planning on the period patrol point group set according to the community patrol model to obtain a patrol path set, wherein the visual field analysis is performed on each patrol period object group to obtain the period patrol point group set, and the method comprises the following steps:

s31: selecting the object groups in the inspection period one by one as target inspection object groups, and selecting the inspection objects in the target inspection object groups one by one as target inspection objects;

s32: generating a shooting space of the target patrol object according to the focal length interval of the unmanned aerial vehicle, and selecting shooting points in the shooting space one by one as target shooting points;

s33: calculating an offset shielding coefficient of the target shooting point position by using the following shielding view algorithm:

wherein O is _i The offset shielding coefficient of the target shooting point position i is shown, M is the total pixel number of the total shooting of the unmanned aerial vehicle, delta is shielding weight, j is the shooting pixel serial number of the unmanned aerial vehicle, v _i,j Means the unit vector of the j-th pixel ray in all pixel rays from the target shooting point i to the surface of the target inspection object, wherein the unit vector is a vector dot product symbol, and p _i, The method comprises the steps that shielding coefficients of a j-th pixel ray in all pixel rays from a target shooting point i to the surface of a target inspection object are referred to, wherein I is a vector modulo symbol, and w is a normal vector of the inner surface of the target inspection object;

s34: selecting the target shooting point with the smallest offset shielding coefficient as a patrol point of the target patrol object, collecting all patrol points corresponding to the target patrol object group into a time period patrol point group, and collecting all time period patrol point groups into a time period patrol point group set;

s4: selecting the patrol paths in the patrol path set one by one as target patrol paths according to the patrol periods, controlling the unmanned aerial vehicle to carry out point location shooting along the target patrol paths to obtain period patrol chart groups, and collecting all period patrol chart groups into a patrol chart set;

s5: selecting the inspection pictures of the inspection atlas one by one as target inspection pictures, sequentially carrying out texture enhancement and multistage feature extraction operation on the target inspection pictures to obtain an enhancement feature set, carrying out safety analysis on the target inspection pictures according to the enhancement feature set to obtain object safety semantics, and carrying out safety labeling on the community inspection model by utilizing all the object safety semantics to obtain a community inspection result.

2. The intelligent community security risk-based property inspection method of claim 1, wherein the generating the coverage detection path of the initial community map according to the unmanned aerial vehicle collection parameter set comprises:

3. The intelligent property inspection method based on community safety hazards of claim 1, wherein the sequentially performing probability filtering, object detection and semantic recognition operations on the target community point cloud to obtain an inspected object semantic set comprises:

4. The intelligent property inspection method based on community safety hazards of claim 3, wherein the probability filtering the target community point cloud by using a preset gaussian density algorithm to obtain a target filtering point cloud comprises the following steps:

5. The intelligent community security hidden danger-based property inspection method of claim 1, wherein the three-dimensionally reconstructing the community inspection point cloud into a community inspection model comprises:

6. The intelligent property inspection method based on community safety hazards of claim 1, wherein the performing path planning on the period inspection point group set according to the community inspection model to obtain an inspection path set comprises:

7. The intelligent property inspection method based on community safety hazards of claim 1, wherein the sequentially performing texture enhancement and multi-level feature extraction operations on the target inspection picture to obtain an enhanced feature set comprises:

8. The intelligent property inspection method based on community safety hazards of claim 1, wherein the performing safety analysis on the target inspection picture according to the enhanced feature set to obtain object safety semantics comprises:

9. The intelligent community safety hazard-based property inspection method according to claim 8, wherein the fusing the enhanced feature set into a target inspection feature by using a preset attention fusion algorithm comprises:

Wherein a refers to the target inspection feature, softmax is a normalization function, θ, ω, σ are fusion weight matrices of the attention fusion algorithm, Q is the standard texture feature, K is the standard color feature, T is a transposed symbol, f is a feature dimension, and feature dimensions of the standard texture feature, the standard color feature, and the standard shape feature are equal, and V refers to the standard shape feature.

10. Property intelligence inspection device based on community's potential safety hazard, its characterized in that, the device includes:

wherein O is _i The offset shielding coefficient of the target shooting point position i is shown, M is the total pixel number of the total shooting of the unmanned aerial vehicle, delta is shielding weight, j is the shooting pixel serial number of the unmanned aerial vehicle, v _i,j Means the unit vector of the j-th pixel ray in all pixel rays from the target shooting point i to the surface of the target inspection object, wherein the unit vector is a vector dot product symbol, and p _i, The method comprises the steps that shielding coefficients of a j-th pixel ray in all pixel rays from a target shooting point i to the surface of a target inspection object are referred to, wherein I is a vector modulo symbol, and w is a normal vector of the inner surface of the target inspection object; selecting the target shooting point with the smallest offset shielding coefficient as a patrol point of the target patrol object, and taking the target patrol object as a patrol point of the target patrol objectAll the inspection points corresponding to the groups are collected into time period inspection point groups, and all the time period inspection point groups are collected into time period inspection point group sets;