CN107256017A - route planning method and system - Google Patents
route planning method and system Download PDFInfo
- Publication number
- CN107256017A CN107256017A CN201710296362.3A CN201710296362A CN107256017A CN 107256017 A CN107256017 A CN 107256017A CN 201710296362 A CN201710296362 A CN 201710296362A CN 107256017 A CN107256017 A CN 107256017A
- Authority
- CN
- China
- Prior art keywords
- point
- flex point
- flex
- region
- area
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 83
- 239000000284 extract Substances 0.000 claims abstract description 52
- 238000000605 extraction Methods 0.000 claims abstract description 25
- 238000013079 data visualisation Methods 0.000 claims abstract description 14
- 238000004422 calculation algorithm Methods 0.000 claims description 91
- 238000012549 training Methods 0.000 claims description 23
- 230000004888 barrier function Effects 0.000 claims description 17
- 238000002790 cross-validation Methods 0.000 claims description 10
- 230000006978 adaptation Effects 0.000 claims description 9
- 238000013459 approach Methods 0.000 claims description 8
- 238000006243 chemical reaction Methods 0.000 claims description 5
- 238000000638 solvent extraction Methods 0.000 claims description 3
- 238000010801 machine learning Methods 0.000 description 12
- 230000000694 effects Effects 0.000 description 10
- 238000010586 diagram Methods 0.000 description 6
- 238000012545 processing Methods 0.000 description 6
- 238000012360 testing method Methods 0.000 description 6
- 238000005516 engineering process Methods 0.000 description 5
- 238000007500 overflow downdraw method Methods 0.000 description 5
- 230000004927 fusion Effects 0.000 description 4
- 230000001788 irregular Effects 0.000 description 4
- 239000002245 particle Substances 0.000 description 4
- 230000006870 function Effects 0.000 description 3
- 241000251468 Actinopterygii Species 0.000 description 2
- 238000007635 classification algorithm Methods 0.000 description 2
- 239000012141 concentrate Substances 0.000 description 2
- 238000010219 correlation analysis Methods 0.000 description 2
- 238000011156 evaluation Methods 0.000 description 2
- 238000013210 evaluation model Methods 0.000 description 2
- 238000009313 farming Methods 0.000 description 2
- 238000002156 mixing Methods 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 238000012706 support-vector machine Methods 0.000 description 2
- 238000012795 verification Methods 0.000 description 2
- 230000000007 visual effect Effects 0.000 description 2
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 2
- 206010000125 Abnormal dreams Diseases 0.000 description 1
- 241001269238 Data Species 0.000 description 1
- 238000012952 Resampling Methods 0.000 description 1
- HUTDUHSNJYTCAR-UHFFFAOYSA-N ancymidol Chemical compound C1=CC(OC)=CC=C1C(O)(C=1C=NC=NC=1)C1CC1 HUTDUHSNJYTCAR-UHFFFAOYSA-N 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 239000007795 chemical reaction product Substances 0.000 description 1
- 238000010835 comparative analysis Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 238000003066 decision tree Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000002844 melting Methods 0.000 description 1
- 230000008018 melting Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000007935 neutral effect Effects 0.000 description 1
- 238000007637 random forest analysis Methods 0.000 description 1
- 239000011435 rock Substances 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 241000894007 species Species 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 239000011800 void material Substances 0.000 description 1
Classifications
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05D—SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
- G05D1/00—Control of position, course, altitude or attitude of land, water, air or space vehicles, e.g. using automatic pilots
- G05D1/02—Control of position or course in two dimensions
- G05D1/021—Control of position or course in two dimensions specially adapted to land vehicles
- G05D1/0231—Control of position or course in two dimensions specially adapted to land vehicles using optical position detecting means
- G05D1/0246—Control of position or course in two dimensions specially adapted to land vehicles using optical position detecting means using a video camera in combination with image processing means
- G05D1/0251—Control of position or course in two dimensions specially adapted to land vehicles using optical position detecting means using a video camera in combination with image processing means extracting 3D information from a plurality of images taken from different locations, e.g. stereo vision
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05D—SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
- G05D1/00—Control of position, course, altitude or attitude of land, water, air or space vehicles, e.g. using automatic pilots
- G05D1/02—Control of position or course in two dimensions
- G05D1/021—Control of position or course in two dimensions specially adapted to land vehicles
- G05D1/0276—Control of position or course in two dimensions specially adapted to land vehicles using signals provided by a source external to the vehicle
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05D—SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
- G05D1/00—Control of position, course, altitude or attitude of land, water, air or space vehicles, e.g. using automatic pilots
- G05D1/02—Control of position or course in two dimensions
- G05D1/021—Control of position or course in two dimensions specially adapted to land vehicles
- G05D1/0276—Control of position or course in two dimensions specially adapted to land vehicles using signals provided by a source external to the vehicle
- G05D1/028—Control of position or course in two dimensions specially adapted to land vehicles using signals provided by a source external to the vehicle using a RF signal
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Automation & Control Theory (AREA)
- Remote Sensing (AREA)
- Radar, Positioning & Navigation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Aviation & Aerospace Engineering (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Multimedia (AREA)
- Electromagnetism (AREA)
- Image Analysis (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The embodiment of the present invention provides a kind of route planning method and system.Methods described includes：Data set to be predicted is obtained, the data set is the first extraction point of model in scene, and described first extracts the barycenter that point is the model；The data set is subjected to data visualization, multiple regions are obtained, each region includes multiple first and extracts point；For each region, judge to know that described first extracts point as flex point, and the region is flex point area, then obtains the second of the flex point area and extract point as flex point, described second extracts the barycenter that point is the flex point area；Each flex point is attached with default direction, conduct route is obtained.The method of the embodiment of the present invention is by determining flex point area, and flex point area's barycenter is as flex point and connects flex point for extraction, route planning can be realized automatically, so as to greatly improve operating efficiency.
Description
Technical field
The present embodiments relate to a kind of areas of information technology, particularly a kind of route planning method and system.
Background technology
Path planning refers in the environment with barrier, according to certain principle, finds from initial state to target
The collisionless path of state.Path planning is had a wide range of applications, and one of which is applied in three-dimensional scenic roaming, based on void
Intend reality technology can accurately, three-dimensional scenic is dynamically shown, so that experiencer can be roamed in three-dimensional scenic.
Wherein, it is to carry out path planning to make the most basic ring of the three-dimensional scenic that can be roamed, and guide is provided for experiencer.
The process of path planning is as follows in the prior art：For each scene, started by means of corresponding plug-in unit (such as unity plug-in units) and transported
OK, the flex point in roam procedure is set manually respectively, correlative code is write, to determine the conduct route in roam procedure.
Wherein, Unity has been widely used in 3d gaming exploitation and three-dimensional scenic is taken as the development of games engine of main flow
Build.
It is understood that because the three-dimensional scenic of roaming is numerous and it is abundant in content to show, if each scene is carried out
The operation of Roam Path is set manually, and the need of work repeated repeats to do hundreds of times, and needs every time by means of plug-in unit, pin
Originally the operation such as height of adjustment flex point and flex point, is repeated, the disaster of bad dream is brought to programmer.
At present, prior art solves the problem of path planning efficiency is low also without corresponding method.
The content of the invention
For the defect of prior art, the embodiment of the present invention provides a kind of route planning method and system.
On the one hand, the embodiment of the present invention provides a kind of route planning method, including：
Data set to be predicted is obtained, the data set is the first extraction point of model in scene, and described first extracts point
It is the barycenter of the model；
The data set is subjected to data visualization, multiple regions are obtained, each region includes multiple first and extracts point；
For each region, judge to know that described first extracts point as flex point, and the region is flex point area, then obtains institute
State the second of flex point area and extract point as flex point, described second extracts the barycenter that point is the flex point area；
Each flex point is attached with default direction, conduct route is obtained.
On the other hand, the embodiment of the present invention also provides a kind of route planning system, including：
Acquisition module, the data set to be predicted for obtaining, the data set is the first extraction point of model in scene, institute
State the barycenter that the first extraction point is the model；
Region division module, for the data set to be carried out into data visualization, obtains multiple regions, each region includes
Multiple first extract point；
Flex point determining module, for for each region, judging to know that described first extracts point as flex point, and the region
It is flex point area, then obtains the second of the flex point area and extract point as flex point, described second extracts the barycenter that point is the region；
Conduct route acquiring module, for each flex point to be attached with default direction, obtains conduct route.
As shown from the above technical solution, route planning method provided in an embodiment of the present invention and system, methods described pass through
Flex point area is determined, flex point area's barycenter is as flex point and connects flex point for extraction, route planning can be realized automatically, so as to greatly improve work
Make efficiency.
Brief description of the drawings
Fig. 1 is a kind of schematic flow sheet of route planning method provided in an embodiment of the present invention；
A kind of learning curve schematic diagram for route planning method that Fig. 2 provides for further embodiment of this invention；
A kind of flex point connection diagram for route planning method that Fig. 3 provides for further embodiment of this invention；
A kind of schematic flow sheet for route planning method that Fig. 4 provides for further embodiment of this invention；
A kind of schematic flow sheet of the step 406 for route planning method that Fig. 5 provides for further embodiment of this invention；
A kind of partial schematic diagram of two sorting algorithms of route planning method that Fig. 6 provides for further embodiment of this invention；
A kind of structural representation for route planning system that Fig. 7 provides for further embodiment of this invention.
Embodiment
To make the purpose, technical scheme and advantage of the embodiment of the present invention clearer, below in conjunction with the embodiment of the present invention
In accompanying drawing, the technical scheme in the embodiment of the present invention is explicitly described, it is clear that described embodiment be the present invention
A part of embodiment of embodiment, rather than whole embodiments.
The embodiment of the present invention can be planned using multiple technologies with realizing route, in order to better illustrate the present embodiment, to answer
With being illustrated exemplified by machine learning techniques.
Machine learning, as its name suggests, allows computer to carry out learning behavior as the mankind.
Specifically, collection has n sample (sample) data set, and each sample has known attribute, can be predicted unknown
The attribute of data.
Machine learning has been widely used in the fields such as medical science, education, agricultural, Digital Display, it is therefore an objective to by machine
Strength help people's processing or solve mass data in real world, and offered convenience for people, allow the life of people to fill
Intellectuality is expired.
Generally, machine learning can be divided into supervised learning, unsupervised learning and the class of intensified learning three.
Have some related algorithms for the study of each species, current most common mode of learning be supervised learning with
Unsupervised learning.
Illustrate the difference of two kinds of modes of learning by taking classification (classification) as an example, classification makes most simple also most universal
A class machine learning.For classification, the training data of input has feature (feature), there is label (label).It is so-called to learn
Practise, its essence is exactly to find the relation (mapping) between feature and label.
During unknown data input without label, unknown data label can be obtained by existing relation when there is a feature.
In above-mentioned assorting process, if all training datas have label, for supervised learning (supervised
learning).It is exactly unsupervised learning (unsupervised learning) if data do not have label, namely cluster
(clustering)。
Wherein, supervised learning related algorithm such as KNN (k-NearestNeighbor, K arest neighbors), is usually used in kaggle
The related multiple fields of contest, game AI, big data.
Unsupervised learning related algorithm such as K-MEANS (K classes algorithm) is usually used in data analysis association area.
In the present embodiment, route planning, such as three-dimensional range and road guide, in order to more clear can be carried out for a variety of demands
Description the present embodiment of Chu, is illustrated by taking the three-dimensional range being related in Farming Culture project as an example.
Specifically, three-dimensional range refers to switch in the three dimensions by picture construction, reaches and browses each difference
The purpose of scene.
Wherein, the roaming that video camera can be provided for experiencer in three-dimensional scenic is guided, or video camera is according to experiencer's
Operation is shown three-dimensional scenic.Three-dimensional scenic roaming can be brought outside strong visual impact to experiencer in Digital Display,
Experiencer can also be allowed to understand relevant knowledge.
Specifically, built during three dimensions by three-dimensional scenic, that is to say, that constitute three-dimensional space by multiple three-dimensional scenics
Between.
Fig. 1 shows a kind of schematic flow sheet of route planning method provided in an embodiment of the present invention.
Reference picture 1, method provided in an embodiment of the present invention specifically includes following steps：
Step 11, acquisition data set to be predicted, the data set are the first extraction point of model in scene, described first
Extract the barycenter that point is the model.
Alternatively, from described one point of model extraction, i.e., first extracts point.
In this step, it is its barycenter by each Unified Model breviary.
If model is made up of n particle, the barycenter of model is the mean place of model quality distribution, for representing in model
Mass concentration is in this image point.
Wherein, barycenter, which is extracted, can take the prior art manner of graph and image processing to be extracted, for example, at binaryzation
Reason.
If it is understood that the model irregular shape, it is impossible to extract effective center, using extraction barycenter
Mode, can rationally, accurately by the region breviary.
Step 12, by the data set carry out data visualization, obtain multiple regions, each region is carried including multiple first
Take a little.
Step 13, for each region, judging to know described first, to extract point be flex point, and the region is flex point area,
Then obtain the second of the flex point area and extract point as flex point, described second extracts the barycenter that point is the flex point area.
Alternatively, a point is extracted from the flex point area, i.e., second extracts point.
In this step, each flex point area is unified into breviary for its barycenter.
If flex point area is made up of n particle, the barycenter in flex point area is the mean place of flex point area Mass Distribution, for representing
Mass concentration is in this image point in flex point area.
Wherein, barycenter, which is extracted, can take the prior art manner of graph and image processing to be extracted, for example, at binaryzation
Reason.
If it is understood that flex point area irregular shape, it is impossible to extract effective center, using extraction barycenter
Mode, can rationally, accurately by the region breviary, further, regard the barycenter of extraction as flex point.
Step 14, each flex point is attached with default direction, obtains conduct route.
, can be from closely to remote or from as far as closely connecting each flex point after the flex point in each flex point area is obtained.
Certainly, the mode of connection can be adjusted according to the displaying content of the scene, be not limited.
The route planning method that the present embodiment is provided, at least with following technique effect：
By determining flex point area, flex point area's barycenter is as flex point and connects flex point for extraction, and route planning can be realized automatically, from
And greatly improve operating efficiency.
On the basis of above-described embodiment, the route planning method that further embodiment of this invention is provided.Methods described step
12, there can be a variety of partitioned modes, the present embodiment illustrates one of which.
Methods described includes：
Step 12 ', by the data set on coordinate system use default Optimal Clustering, carry out data visualization,
The data set is divided into multiple regions.
In this step, the model is object present in scene.
For example, the model is farm implements, crops, fish, personage, field, water etc., certainly, can be according to actual conditions
Configured, be not limited.
Alternatively, from described one point of person extraction, i.e., first extracts point.
In this step, it is its barycenter by personage's Unified Model breviary.
Alternatively, the default Optimal Clustering can be following one：DBScan(Density-Based
Spatial Clustering of Applications with Noise), the clustering algorithm based on space density, KMeans
(K classes algorithm), MiniBatchKmeans (small lot K classes a surname hair), Birth algorithms (big data algorithm).
Before the default Optimal Clustering of the use, methods described also includes：
It is multiple sample areas by sample scene partitioning for each clustering algorithm；
Judgement knows that the sample areas that clustering algorithm is obtained includes predetermined flex point, then makees the sample areas
For flex point area, otherwise, the sample areas is regard as non-flex point area.
It should be noted that determining that Optimal Clustering can have various ways, the present embodiment provides a kind of embodiment.
The predetermined flex point number that is included according to predetermined flex point number, sample areas and sample scene it is total
Number, obtains the corresponding adaptation value of the clustering algorithm；It regard the corresponding clustering algorithm of highest adaptation value as Optimal Clustering.
First, sample scene can be obtained, machine learning is carried out.
Secondly, it is multiple sample areas by sample scene partitioning for each clustering algorithm.
By machine learning scikit-learn storehouses call successively the most frequently used cluster algorithms DBScan, KMeans,
MiniBatchKmeans and Birth algorithms are clustered.
Further, the model in each sample areas can be obtained, and using the coordinate of the barycenter of each model as data source,
To judge whether the sample areas can be as flex point area, that is to say, that whether there is predetermined turn according to sample areas
Put to judge whether that flex point area can be done.
Alternatively, predetermined flex point is the flex point manually planned.
Again, statistical magnitude：Predetermined flex point number and sample that predetermined flex point number, sample areas include
The sum of this scene.
Then, the predetermined flex point number and sample included according to predetermined flex point number, sample areas
The sum of scape, obtains the corresponding adaptation value of the clustering algorithm.
Wherein, the adaptation value is used to represent resonable degree of the clustering algorithm to the flex point planning of sample scene.
Specifically, adaptation value can be calculated according to below equation：
In formula, n is that the number of scene, g are predetermined flex point number, using turning that clustering algorithm A is obtained in scene i
Point area, and have gAIndividual flex point is distributed in the flex point area gathered by clustering algorithm A, and MAR (Mean Adaption Rate, it is average
Fitting percentage) it is adaptation value.
Finally, it regard the corresponding clustering algorithm of highest adaptation value as Optimal Clustering.
The route planning method that the present embodiment is provided, at least with following technique effect：
By the way that the data set is used into default Optimal Clustering on coordinate system, data visualization is carried out, by institute
State data set and be divided into multiple regions, so as to intelligentized programme path.
On the basis of above-described embodiment, the route planning method that further embodiment of this invention is provided.Methods described step
The mode of 13 judgement can have a variety of, and the present embodiment illustrates one of which.
The step 13 is specially：
Step 13 ', for each region, using default optimal two sorting algorithm, judgement knows that described first extracts point
For flex point.
In this step, default optimal two sorting algorithm can be following one：KNN,SVM(Support Vector
Machine, SVMs), NBC (Naive Bayesian Model, model-naive Bayesian), MLP (Multi-layer
Perceptron, also referred to as multilayer perceptron, artificial neural network), (Recursive Flow Classification, are passed RFC
Return flow point class), SGD (Stochastic Gradient Descent, stochastic gradient descent).
It should be noted that determining that optimal two sorting algorithm can have various ways, the present embodiment provides a kind of embodiment.
Concretely following steps A1-A4：
Step A1, the institute for obtaining the flex point area a little, are used as flex point sample.
Step A2, the institute for obtaining the non-flex point area a little, are used as non-flex point sample.
Step A3, it is used as checking to collect the flex point sample and the non-flex point sample.
Step A4, for every one or two sorting algorithm, based on the checking collection, pass through cross-validation method and combine two points of correspondence
The grader that class algorithm is learnt is trained to checking collection, and training result is matched with the known results that checking is concentrated,
Corresponding two sorting algorithm during by matching degree highest, is used as optimal two sorting algorithm.
Alternatively, the study of two sorting algorithms of correspondence can be judged according to training set curve and checking collection curve matching degree
Ability and generalization ability.
Fig. 2 shows a kind of learning curve schematic diagram for route planning method that the present embodiment is provided.
As shown in Fig. 2 abscissa represents number of training, ordinate represents scoring.
For example, in 200 training samples, the degree of accuracy that flex point is determined in training set is 0.92, and checking is concentrated true
The degree of accuracy for determining flex point is approximately 0.4, training set curve (Training score) and checking collection curve (Cross-
Validation score) differ greatly, but increasing with number of training, training set curve and checking collection curve gradually become
In gentle, the preferable of two curve matchings is represented.
The route planning method that the present embodiment is provided, at least with following technique effect：
Judge that described first extracts whether point is flex point by using default optimal two sorting algorithm, and the region
Whether it is flex point area so that judge more accurate, reasonable, so as to intelligentized programme path.
On the basis of above-described embodiment, the route planning method that further embodiment of this invention is provided.Methods described step
After 13, possible flex point area quantity is excessive, so that flex point is excessive, causes camera to rock frequently, route planning is not
Rationally, the present embodiment provides a kind of scheme for solving this problem.
After methods described step 13, in addition to：
According to flex point occupation rate IOR judge the region whether flex point area.
The flex point occupation rate IOR accounts for the ratio of the quantity of all first extractions points in affiliated area for the quantity of the flex point
Value, if ratio is more than preset value, regard the region as flex point area；
Obtain the second of the flex point area and extract point as final flex point
Alternatively, flex point occupation rate can be obtained according to below equation：
Flex point occupation rate
In formula, IOR (inflexion occupancy rate) represents flex point occupation rate, and II represents to pass through in a certain flex point area
Two classification predictions can as flex point quantity, CCP represents total model sum in same flex point area.
It is understood that one barycenter of a model extraction, i.e. model sum are equal to barycenter sum.
Wherein, preset value (Threhold) can be 60%, can be configured, be not limited according to actual conditions certainly.
Further, IOR is worked as>When 60%, it can determine whether that the barycenter in the flex point area can be as flex point.
For example, prediction flex point number is 30, and barycenter sum is 50% for 60, IOR, then the flex point area is not suitable as
Flex point relevant range, so as to reduce the quantity in flex point area.
Correspondingly, the barycenter in flex point area is not extracted as flex point, can reduce the quantity of flex point.
Alternatively, the barycenter for extracting the flex point area includes the two-dimensional coordinate of barycenter, wherein, the two-dimensional coordinate is basis
The three-dimensional coordinate of the barycenter of collection carries out what Coordinate Conversion was obtained.
Specifically, the three-dimensional coordinate of each flex point area barycenter is gathered, the three-dimensional coordinate is subjected to Coordinate Conversion, two are obtained
Dimension coordinate.
For example, y-axis upward in three-dimensional coordinate (x, y, z) can be uniformly set to 0, obtains two-dimensional coordinate (x, z).
Correspondingly, the barycenter for extracting the model includes the two-dimensional coordinate of barycenter, wherein, the two-dimensional coordinate is that basis is adopted
The three-dimensional coordinate of the barycenter of collection carries out what Coordinate Conversion was obtained.
It is understood that three-dimensional coordinate is converted to two coordinates of plane by the method mapped using coordinate, it can subtract significantly
The complexity of small follow-up data processing.
Correspondingly, the step 14 is specially：Each final flex point is attached with default direction, conduct road is obtained
Line.
The route planning method that the present embodiment is provided, at least with following technique effect：
The ratio of the quantity of all first extractions points in affiliated area is accounted for more than default by obtaining the quantity of the flex point
The region of value, screening again is carried out to the region can simplify the quantity in flex point area, guide camera smooth.
Fig. 3 shows a kind of flex point connection diagram for route planning method that further embodiment of this invention is provided.
Reference picture 3, on the basis of above-described embodiment, the route planning method that the present embodiment is provided.Methods described step
14, obtaining the mode of conduct route can have a variety of, and the present embodiment illustrates one of which.
Methods described step 14 ', specifically include：
The flex point and additional inflection points are attached with default direction；
Wherein, the additional inflection points are obtained according to barrier point and closest approach, and the barrier point is positioned at two flex points
Point in air line distance, the closest approach is the point nearest apart from barrier point.
In this step, may also there is a point in the air line distance of two flex points, then be determined as that video camera should be kept away
The model allowed.
Further, if the route of video camera avoidance is too short, the effect hidden can not be played, if the route avoided
It is oversize, then roaming route is added, causes route planning unreasonable.
Alternatively, the instruction point that a suitable additional inflection points turn round as video camera is increased in the course of travel.
For example, when flex point 1<x1, y1>With flex point 2<x2, y2>Air line distance on exist<x′1, y '1>, by the point
It is used as barrier point.
The distance of two flex points is obtained using equation below.
In formula,<x1, y1>For the coordinate of flex point 1,<x2, y2>For the coordinate of flex point 2, Distance represents distance.
Further, the point nearest apart from barrier point, i.e. closest approach are obtained<x′1, y '1>。
Alternatively, additional inflection points are obtained using equation below：
That is, choosing barrier point<x′1, y '1>With closest approach<x″1, y "1>Midpoint be additional inflection points<x″‘1, y " '1
>。
Next, flex point 1, additional inflection points and flex point 2 are connected with default direction.
Alternatively, flex point 1, additional inflection points and flex point 2 can be connected by straight line.
Wherein, can be from closely to remote or from as far as closely connecting each flex point.
Certainly, the mode of connection can be adjusted according to the displaying content of the scene, be not limited.
The route planning method that the present embodiment is provided, at least with following technique effect：
By increasing additional inflection points so that make up the deficiency of previously determined flex point in time, can avoid the straight lines of two flex points away from
From upper barrier point, so that perfect conduct route.
Fig. 4 shows a kind of schematic flow sheet for route planning method that further embodiment of this invention is provided.
Reference picture 4, method provided in an embodiment of the present invention can be applied to a variety of three-dimensional scenics, in order to better illustrate this
Embodiment, on the basis of above-described embodiment, is said by taking the three-dimensional scenic being related in being converged applied to Farming Culture project as an example
It is bright.In for other VR game or virtual experience system, only related training data need to be changed.
It the described method comprises the following steps：
Step 401, beginning, obtain the three-dimensional scenic built.
Step 402, data acquisition, each scene include multiple models, for each scene, gather each model barycenter
Three-dimensional coordinate.
Step 403, the data collected are pre-processed, specifically, the three-dimensional coordinate matched with scene,
The three-dimensional coordinate of scene numbering belonging to carrying is obtained, three-dimensional coordinate will be also stated and carry out Coordinate Conversion, and obtain two-dimensional coordinate, described two
Dimension coordinate carries affiliated scene numbering.
Thus the two-dimensional coordinate with the barycenter of scene matching can be obtained, is adjusted for following clustering algorithm and two sorting algorithms
With.
Step 404, selection clustering algorithm, are specifically judged by the way of cross validation.
In this step, using the barycenter two-dimensional coordinate of each model in each scene as data source, by machine learning
The most frequently used cluster algorithms are called successively in scikit-learn storehouses：DBScan, KMeans, MiniBatchKmeans and
Birth algorithms are clustered.
The reasonability of step 405, each clustering algorithm of checking.
The two-dimensional coordinate point of model in three-dimensional scenic is carried out after unsupervised learning cluster, can also be carried out reasonability and be tested
Card.
Its core procedure includes：Attributions selection, sample data are collected, clustering algorithm selection, packet and for handing over
Fork experiment carries out the selected of clustering algorithm.
First, the quality of above-mentioned clustering algorithm is assessed according in the following manner：
Average fitting percentage MAR of each clustering algorithm in up to 100 roaming scences is calculated respectively, and where is comparative analysis
The MAR values of individual clustering algorithm are higher.
MAR can be calculated by below equation：
In formula, n is that the number of scene, g are predetermined flex point number, using turning that clustering algorithm A is obtained in scene i
Point area, and have gAIndividual flex point is distributed in the flex point area gathered by clustering algorithm A.
Secondly, take MAR values to come the corresponding clustering algorithm of top n and carry out soundness verification again.
Alternatively, soundness verification is carried out according to difference parameter in class inherited parameter and class.
Specifically, difference parameter is minimized and the maximized clustering algorithm of class inherited parameter in selection class.
Wherein, it is special according to class centre coordinate, coordinate points, number of clusters, unmarked sample, the feature of unmarked sample
The dimension levied, obtains the distance between class center, is used as class inherited parameter.According to each point coordinates, class centre coordinate obtains class
Interior every bit is used as difference parameter in class to the quadratic sum of the distance at class center it described.
Assuming that in some scene, the flex point number (including initial point position, be two-dimensional coordinate) manually planned is g,
And c_A class cluster is obtained using A clustering algorithms in this scene, can calculate poor between the class of i classes and j classes by below equation
Different parameter：
In formula, k represents k cluster, and i and j represent i-th and j cluster, x respectivelyjFor j-th of unmarked sample, xiRepresent
I-th of unmarked sample, For CjClass center, niIt is the points in ith cluster.
AndN represents the dimension of feature, xiuRepresent i-th it is unmarked
U-th of feature of sample, xjuRepresent u-th of feature of j-th of unmarked sample.
Difference parameter in the class of Ci classes can be calculated by below equation：
In formula, k represents k cluster, and Ci represents ith cluster, xiRepresent i-th of unmarked sample.
It should be noted that for each clustering algorithm, can all utilize cross validation, compare reasonable with other algorithms
Property.
If the clustering algorithm rationally, continues executing with step 406, if unreasonable, return to step 404 is using another
Clustering algorithm.
Step 406, the selection of two sorting algorithms are trained with data.
Fig. 5 shows a kind of schematic flow sheet of the step 406 of route planning method provided in an embodiment of the present invention.
In a step 406, suitable two sorting algorithm is selected to be trained to reach higher study to above-mentioned data set
Rate.
As shown in figure 5, the step 406 also includes step 91- steps 99：
Step 91, category label.By being labeled as a little just in flex point cluster, the point in non-flex point class cluster be it is anti-, will just
Example point and counter-example point, are used as the object of two classification.
Specifically, the result clustered using above-mentioned clustering algorithm carries out two classification based trainings.In class cluster comprising artificial flex point
Be designated generally as positive example a little, the point that other class clusters are included uniformly is labeled as counter-example, with above-mentioned cluster result as two
The training data of classification carries out two sorting algorithm selections.
Step 92, data set update, addition new data row, and each data is label information in new data row.
Specifically, concentrate addition one to arrange the data row of entitled " Result " in initial data, concentrate every for flag data
One record "Yes" or "no" flex point.
Step 93, data visualization.The point of each data row carries feature, obtains the feature being had a significant effect to flex point,
And be not feature that flex point has a significant effect, that is, obtain the weight of feature.
Specifically, histogram, statistical chart for each attribute and result queue etc. are drawn.
It can select according to Visual Chart and result data listed with aobvious influential feature as candidate feature.
Step 94, feature selecting, feature each to each two carry out correlation analysis, and each feature has multiple phase relations
Number, summation is averaged the corresponding feature of rear maximum, is used as corner feature.
Specifically, the candidate feature selected with reference to step 93, and correlation analysis is carried out to various features.
Wherein, coefficient correlation is just and value is bigger, then this attribute is more possible to selection and is characterized attribute.
If occurring over-fitting or poor fitting in step 97, just do a little feature engineering work, reduce or
The feature or assemblage characteristic of some new outputs are added into model.
Step 95, algorithms selection.An algorithm in KNN, SVM, NBC, MLP, RFC, SGD is chosen every time, into step
96。
Step 96, cross-validation method choose training set and test set.
Alternatively, roll over cross-validation method to evaluate the performance indications of each two sorting algorithms institute training pattern using K.
When being chosen due to K it is excessive or it is too small can all cause unnecessary trouble, therefore intersected using 10 the most frequently used foldings and test
Demonstration.
Specifically, data set is divided into ten parts, in turn will wherein 9 parts do training 1 part verify, the average of the result of 10 times
As the estimation to arithmetic accuracy, typically also need to the multiple 10 folding cross validation of progress and average, for example：10 10 foldings, which intersect, to be tested
Card, to obtain more accurate the result.
After step 97, selected good suitable cross validation method, the grader gone out respectively using two classification algorithm trainings
Learning curve is drawn, with the learning ability and generalization ability of judgment models.
If training set score curve and cross validation collection score curve are fitted in the learning curve that some grader is drawn
Preferably, show as shown in Figure 2, then choose this two sorting algorithm as optimal two sorting algorithm, i.e. learning ability and generalization ability by
Gradually tend towards stability.
In addition, the error rate and precision, precision ratio and recall ratio, cost-sensitive that should also consider each model are wrong
The factor such as rate carrys out the performance of evaluation model by mistake.
Alternatively, error rate can be calculated according to below equation：
Error rate
In formula, m represents sample number, f presentation class functions, i.e. fitting function, f (xi) represent predicted value ("Yes" or " no
It is " flex point), yiRepresent actual value, Π (f (xi)≠yi) indicator function is represented, it is very to distinguish value 1,0 with fictitious time in bracket.
Alternatively, can be according to below equation computational accuracy：
Precision acc (f；D)=1-E (f；D)
Wherein, error rate is lower, correspondingly, and precision is higher, represents that two sorting algorithms are better.
Alternatively, precision ratio can be calculated according to below equation：
Alternatively, recall ratio can be calculated according to below equation：
In formula, TP is real example, and FP is false positive example ("no" flex point is predicted as into "Yes" flex point), and FN is false counter-example,
TN is true counter-example.
Wherein, precision ratio and recall ratio are higher, represent that two sorting algorithms are better.
Alternatively, can be according to the sensitive error rate of below equation calculation cost：
In formula, E (f；D；Cost) it is cost-sensitive error rate, costijRepresent that by the i-th class sample predictions be jth class sample
Cost, D+Represent positive example subset ("Yes" flex point), D-Represent counter-example subset ("no" flex point).
Step 94, over-fitting or poor fitting judge.
If occurring serious over-fitting or poor fitting, return to step 94 for learning curve in step 97.That is,
If fitting is bad, then to greatest extent from extracting data feature so that algorithm and model is used.
Step 99, Model Fusion and algorithms selection.
Fig. 6 shows a kind of partial schematic diagram of two sorting algorithms of route planning method that the present embodiment is provided.
Refering to Fig. 6, for training set, based on each Model Fusion method, evaluated using Model Fusion, obtain melting for minimum
The corresponding Model Fusion method of hop algorithm, is used as best Model Fusion method.
Through after step 97, training obtained multiple graders (such as logistic on same part data set
Regression, SVM, KNN, random forest, neutral net), grader is done respectively and judged, then result is thrown
Ticket is counted, and several most results of getting tickets are end product.
Which kind of, on Model Fusion method selected, can successively it be called by means of Scikit-learning machine learning storehouse
AdaBoostClassifier (Iterative classification algorithm), BaggingClassifier (resampling classification),
VotingClassifier (ballot sorting algorithm), RandemTreesEnbedding (random number blending algorithm),
The methods such as GradientBoostingClassifier (iteration Decision Tree Algorithm) go to test which kind of Model Fusion method more
It is adapted to.
Alternatively, it can be merged according to below equation evaluation model：
Err (x)=(E [f (x) ']-f (x))2+(E[f(x)′]-f(x)′)2=Bias2+Varience2
In formula, Err (x) represents general magnificent error, and Bias represents deviation, and Varience represents variance, represents that general magnificent error can
Deviation and variance sum are decomposed into, E [f (x) '] represents to expect prediction, and f (x) represents real marking, and f (x) ' is that training set D goes to school
Obtain prediction output of the model on x.
Wherein, selection causes variance and deviation all small Model Fusion algorithms to be as far as possible used as optimal models blending algorithm.
Step 407, after the mode of data test is selected, obtain data set to be predicted, the data set is scene
The first of middle model extracts point, and described first extracts the barycenter that point is the model.
Step 404, using Optimal Clustering, obtain multiple class clusters.
Specifically, the data set is subjected to data visualization, obtains multiple regions, each region is carried including multiple first
Take a little, model two-dimensional coordinate point and its affiliated scene numbering are inputted as clustering algorithm.It is exportable to carry using clustering algorithm
The flex point cluster of mark, non-flex point cluster.
Wherein, the flex point cluster is the distributed areas for the two-dimensional coordinate point for including artificial flex point, and the non-flex point cluster is not
The distributed areas of two-dimensional coordinate point including artificial flex point.
Step 409, two classification flex point predictions.
For each region, judge to know that described first extracts point as flex point, and the region is flex point area, then obtains institute
State the second of flex point area and extract point as flex point, described second extracts the barycenter that point is the flex point area.
Specifically, using class cluster information as two sorting algorithms input.
Wherein, class cluster information includes：The affiliated class cluster numbering of scene numbering, some two-dimensional coordinate, two-dimensional coordinate, place class cluster
Whether default object is had, and whether at least one keyword, place class cluster includes keyword.
According to class cluster information and the flex point manually planned, the corresponding corner feature of class cluster information is obtained,
When being roamed due to artificial setting three-dimensional scenic, the movement of video camera is mismatched with following word, this implementation
Class cluster carries keyword in example, and keyword is shown in the lump when final three-dimensional range is shown.Keyword and flex point can be achieved
One-to-one corresponding, while the convenient relevant speed for setting adjustment captions to occur and precision.
It should be noted that the class cluster includes flex point cluster and non-flex point cluster.
For example, the affiliated class cluster numbering of scene numbering, some model coordinate, model, keyword 1, keyword 2, key
Word 3, keyword 4, keyword 5, place class cluster whether have farm implements, place class cluster whether have crops, place class cluster whether have fish,
Whether place class cluster has whether personage, place class cluster have field, place class cluster whether to have whether water, place class cluster include keyword
1st, whether whether class cluster in place include whether keyword 3, place class cluster include keyword 4, institute comprising keyword 2, place class cluster
Whether keyword 5 is included in class cluster.
Using the exportable model of optimal two sorting algorithm class cluster whether may be configured as flex point, if so, then choosing
Otherwise the center-of-mass coordinate of such cluster is determined as "no" flex point as flex point coordinate.
Further, the result clustered according to step 404 carries out two classification using optimal two sorting algorithm and judged.
Meanwhile, customized flex point occupation rate IOR and threshold value Threhold is introduced to differentiate whether whole class cluster can conduct
Flex point.
If so, then taking the barycenter point coordinates of class cluster as flex point coordinate.
IOR can be calculated according to below equation：
In formula, IOR be flex point occupation rate, II represent in a certain class cluster through two classification prediction can as flex point quantity, CCP
Represent model sum total in same class cluster.
Threshold value Threhold may be configured as 60%, work as IOR>When 60%, can determine whether in such cluster this point can as flex point,
The barycenter point coordinates of class cluster is then taken as flex point coordinate.
Step 410, path connection.
Between flex point and flex point at accessible, the mode that is linked with straight line will be taken.
And video camera is being led into another flex point as an intelligent body from a flex point, it is assumed that there is obstacle in centre
During point, then need so that still smoothly reaching next flex point while video camera avoids barrier point with beeline.
As shown in figure 3, when flex point 1<x1, y1>With flex point 2<x2, y2>There is barrier point in centre<x′1, y '1>When, first should
This fetches the two-dimensional coordinate point of barrier point 1, and is to seek distance between 2 points in plane using equation below,
So as to find out point nearest from barrier point 1 in same class cluster, closest approach is referred to as<x″1, y "1〉。
Then barrier point is chosen<x′1, y '1>With closest approach<x″1, y "1>MidpointTurned as additional
Point<x″‘1, y " '1>。
Finally flex point 1, additional inflection points and flex point 2 are connected by straight line.
Step 411, automatically generate.The route of all flex points is automatically generated according to scene display content.
Step 412, end, can set camera all the time just towards the side of straight line according to the route connected in step 411
To realizing that video camera follows the route set to be roamed with certain speed.
The route planning method that the present embodiment is provided, at least with following technique effect：
1st, without setting Roam Path manually, by determining flex point area, extract flex point area's barycenter and turned as flex point and connection
Point, route planning can be realized automatically, so as to greatly improve operating efficiency.
2nd, route planning is carried out by machine learning, objective study and test can be carried out, the subjectivity manually planned is reduced
Interference, and improve the precision of route planning.
3rd, keyword is carried in all kinds of clusters, after the barycenter for taking class cluster is flex point, keyword and flex point can be achieved
Correspond, the convenient relevant speed for setting adjustment captions to occur and precision.
It should be noted that roaming of the prior art is unreasonable due to the point of setting, or theme is not met so that body
The person's of testing impression is not good, and keyword displaying is coordinated in the embodiment of the present invention so that experiencer goes in scene of game displaying as far as possible
Roaming so that experiencer experiences more rich content.
4th, three-dimensional coordinate is converted to two coordinates of plane by the method mapped using coordinate, can reduce operand.
5th, be its barycenter by each Unified Model breviary, and be used in combination machine learning algorithm select it is more practical optimal
Clustering algorithm and two sorting algorithms, and the accuracy rate of route planning can integrally be improved using Model Fusion mode.
6th, two-dimensional coordinate point of the standard after unitized is clustered using the clustering algorithm of unsupervised learning, and with it is artificial
The data classified contrast (flex point i.e. related to artificial path planning is compared), in original clustering algorithm Performance Evaluation
On the basis of, with reference to average fitting percentage MAR methods simultaneously for clustering algorithm Performance Evaluation, calculated to select suitable cluster
Method.
7th, when whether the class cluster after judging cluster can be as flex point, also with flex point occupation rate IOR methods and threshold value
Threhold, is easy to the selection of optimal flex point in follow-up class cluster, guides camera smooth.
8th, when the line between flex point has barrier point so that video camera is while intelligent avoiding obstacles and makes
Obtain distance most short.
9th, set camera just towards the direction of straight line all the time, realize that video camera follows the route automatic roaming set.
Fig. 7 shows a kind of structural representation for route planning system that further embodiment of this invention is provided.
Reference picture 7, route planning system provided in an embodiment of the present invention, the system includes acquisition module 111, region and drawn
Sub-module 112, flex point determining module 113 and conduct route acquiring module 114, wherein：
Acquisition module 111 obtains data set to be predicted, and the data set is the first extraction point of model in scene, described
First extracts the barycenter that point is the model；Region division module 112 is used to the data set carrying out data visualization, obtains
Multiple regions, each region includes multiple first and extracts point；Flex point determining module 113 is used to be directed to each region, judges to know
Described first extracts point for flex point, and the region is flex point area, then obtains the second of the flex point area and extract point as flex point,
Described second extracts the barycenter that point is the flex point area；Conduct route acquiring module 114 is used for each flex point with default direction
It is attached, obtains conduct route.
The embodiment of the present invention can be planned using multiple technologies with realizing route, in order to better illustrate the present embodiment, to answer
With being illustrated exemplified by machine learning techniques.
Alternatively, acquisition module 111 extracts point from described one point of model extraction, i.e., first.
In this step, it is its barycenter by each Unified Model breviary.
If model is made up of n particle, the barycenter of model is the mean place of model quality distribution, for representing in model
Mass concentration is in this image point.
Wherein, barycenter, which is extracted, can take the prior art manner of graph and image processing to be extracted, for example, at binaryzation
Reason.
If it is understood that the model irregular shape, it is impossible to extract effective center, using extraction barycenter
Mode, can rationally, accurately by the region breviary.
The data set is carried out data visualization by region division module 112, obtains multiple regions, and each region includes many
Individual first extracts point.
Wherein, the flex point area represents that the region is the region related to flex point.
Flex point determining module 113 extracts a point from the flex point area, i.e., second extracts point.
In this step, each flex point area is unified into breviary for its barycenter.
If flex point area is made up of n particle, the barycenter in flex point area is the mean place of flex point area Mass Distribution, for representing
Mass concentration is in this image point in flex point area.
Wherein, barycenter, which is extracted, can take the prior art manner of graph and image processing to be extracted, for example, at binaryzation
Reason.
If it is understood that flex point area irregular shape, it is impossible to extract effective center, using extraction barycenter
Mode, can rationally, accurately by the region breviary, further, regard the barycenter of extraction as flex point.
Conduct route acquiring module 114, can be from closely to remote or from as far as closely respectively turning after the flex point in each flex point area is obtained
Point is connected.
Certainly, the mode of connection can be adjusted according to the displaying content of the scene, be not limited.
The route planning system that the present embodiment is provided, available for the method for performing above method embodiment, this implementation is no longer
Repeat.
The route planning system that the present embodiment is provided, at least with following technique effect：
Flex point area is determined by region division module, flex point determining module extracts flex point area barycenter as flex point, and by going
Footpath route obtains mould connection flex point, route planning can be realized automatically, so as to greatly improve operating efficiency.
Finally it should be noted that：Above example is only limited to the technical scheme for illustrating the embodiment of the present invention, rather than to it
System；Although the embodiment of the present invention is described in detail with reference to the foregoing embodiments, one of ordinary skill in the art should
Understand：It can still modify to the technical scheme described in foregoing embodiments, or to which part technical characteristic
Carry out equivalent substitution；And these modifications or replacement, the essence of appropriate technical solution is departed from each reality of the embodiment of the present invention
Apply the spirit and scope of a technical scheme.
Claims (10)
1. a kind of route planning method, it is characterised in that including：
Data set to be predicted is obtained, the data set is the first extraction point of model in scene, and the first extraction point is institute
State the barycenter of model；
The data set is subjected to data visualization, multiple regions are obtained, each region includes multiple first and extracts point；
For each region, judge to know that described first extracts point as flex point, and the region is flex point area, then is turned described in acquisition
The second of point area extracts point as flex point, and described second extracts the barycenter that point is the flex point area；
Each flex point is attached with default direction, conduct route is obtained.
2. according to the method described in claim 1, it is characterised in that：It is described that the data set is subjected to data visualization, obtain
Multiple regions, be specially；
The data set is used into default Optimal Clustering on coordinate system, data visualization is carried out, by the data set
It is divided into multiple regions.
3. method according to claim 2, it is characterised in that：It is described before the default Optimal Clustering of the use
Method also includes：
It is multiple sample areas by sample scene partitioning for each clustering algorithm；
Judgement knows that the sample areas that clustering algorithm is obtained includes predetermined flex point, then using the sample areas as turning
Point area, otherwise, regard the sample areas as non-flex point area.
4. method according to claim 3, it is characterised in that：Methods described also includes：
The predetermined flex point number and the sum of sample scene included according to predetermined flex point number, sample areas,
Obtain the corresponding adaptation value of the clustering algorithm；
It regard the corresponding clustering algorithm of highest adaptation value as Optimal Clustering.
5. method according to claim 4, it is characterised in that：It is described to judge to know that described first extracts point as flex point, tool
Body is：
For each region, using default optimal two sorting algorithm, judge to know that described first extracts point as flex point.
6. method according to claim 5, it is characterised in that：Before default optimal two sorting algorithm of the use, institute
Stating method also includes：
The institute for obtaining the flex point area a little, is used as flex point sample；
The institute for obtaining the non-flex point area a little, is used as non-flex point sample；
The flex point sample and non-flex point sample are collected as checking；
For every one or two sorting algorithm, based on the checking collection, pass through cross-validation method and combination two sorting algorithms of correspondence
The grader obtained is trained to checking collection, training result is matched with the known results that checking is concentrated, by matching degree most
Corresponding two sorting algorithm when high, is used as optimal two sorting algorithm.
7. according to the method described in claim 1, it is characterised in that：Described to be directed to each region, judgement knows that described first carries
It is a little flex point to take, and the region is flex point area, is specially：
According to flex point occupation rate IOR judge the region whether flex point area；
The flex point occupation rate IOR accounts for the ratio of the quantity of all first extractions points in affiliated area for the quantity of the flex point,
If ratio is more than preset value, the region is regard as flex point area；
Obtain the second of the flex point area and extract point as final flex point；
Correspondingly, it is described to be attached each flex point with default direction, conduct route is obtained, is specially：
Each final flex point is attached with default direction, conduct route is obtained.
8. according to the method described in claim 1, it is characterised in that：Described first extracts two that point includes the barycenter of the model
Dimension coordinate, or the second extraction point include a two-dimensional coordinate for the barycenter in the flex point area；Wherein, the two-dimensional coordinate is root
Carry out what Coordinate Conversion was obtained according to the three-dimensional coordinate of the barycenter of collection.
9. according to the method described in claim 1, it is characterised in that：It is described to be attached each flex point with default direction, tool
Body is,
The flex point and additional inflection points are attached with default direction；
Wherein, the additional inflection points are obtained according to barrier point and closest approach, and the barrier point is the straight line positioned at two flex points
Apart from upper point, the closest approach is the point nearest apart from barrier point.
10. a kind of route planning system, it is characterised in that including：
Acquisition module, the data set to be predicted for obtaining, the data set is first of model in scene to extract point, described the
One extracts the barycenter that point is the model；
Region division module, for the data set to be carried out into data visualization, obtains multiple regions, each region includes multiple
First extracts point；
Flex point determining module, for for each region, judging to know that described first extracts point as flex point, and the region is to turn
Point area, then obtain the second of the flex point area and extract point as flex point, and described second extracts the barycenter that point is the region；
Conduct route acquiring module, for each flex point to be attached with default direction, obtains conduct route.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710296362.3A CN107256017B (en) | 2017-04-28 | 2017-04-28 | Route planning method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710296362.3A CN107256017B (en) | 2017-04-28 | 2017-04-28 | Route planning method and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107256017A true CN107256017A (en) | 2017-10-17 |
| CN107256017B CN107256017B (en) | 2020-08-04 |
Family
ID=60027591
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710296362.3A Expired - Fee Related CN107256017B (en) | 2017-04-28 | 2017-04-28 | Route planning method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107256017B (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109260709A (en) * | 2018-09-11 | 2019-01-25 | 深圳市创梦天地科技有限公司 | A kind of generation method and device of dynamic route |
| CN110705619A (en) * | 2019-09-25 | 2020-01-17 | 南方电网科学研究院有限责任公司 | Fog concentration grade judging method and device |
| CN110766770A (en) * | 2019-10-16 | 2020-02-07 | 腾讯科技（深圳）有限公司 | Thermodynamic diagram generation method and device, readable storage medium and computer equipment |
| CN111035928A (en) * | 2019-11-19 | 2020-04-21 | 腾讯科技（深圳）有限公司 | Method and device for displaying travel route, storage medium and electronic device |
| CN111881939A (en) * | 2020-06-24 | 2020-11-03 | 东南大学 | Shared single-vehicle parking area layout method based on clustering algorithm |
| CN112729327A (en) * | 2020-12-24 | 2021-04-30 | 浙江商汤科技开发有限公司 | Navigation method, navigation device, computer equipment and storage medium |
| CN113075933A (en) * | 2021-03-30 | 2021-07-06 | 北京布科思科技有限公司 | Robot passage control method, device and equipment |
| CN117929173A (en) * | 2024-03-18 | 2024-04-26 | 中国汽车技术研究中心有限公司 | Method and device for testing and calibrating mechanical properties of rib components of automobile collision dummy |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20060155394A1 (en) * | 2004-12-16 | 2006-07-13 | International Business Machines Corporation | Method and apparatus for order-preserving clustering of multi-dimensional data |
| CN101271465A (en) * | 2007-12-12 | 2008-09-24 | 北京航空航天大学 | Lens clustering method based on information bottleneck theory |
| CN102063707A (en) * | 2011-01-05 | 2011-05-18 | 西安电子科技大学 | Mean shift based grey relation infrared imaging target segmentation method |
| US8452531B2 (en) * | 2009-02-04 | 2013-05-28 | Mitac International Corp. | Route guiding system and method thereof |
| CN103268729A (en) * | 2013-05-22 | 2013-08-28 | 北京工业大学 | Mobile robot cascading type map creating method based on mixed characteristics |
| CN106060052A (en) * | 2016-06-02 | 2016-10-26 | 深圳市豹风网络股份有限公司 | Three-dimensional navigation method for network game of mobile terminal |
| CN106097424A (en) * | 2016-06-12 | 2016-11-09 | 中国农业大学 | The method for visualizing of a kind of Agro-ecology scene and device |
| CN106295639A (en) * | 2016-08-01 | 2017-01-04 | 乐视控股（北京）有限公司 | A kind of virtual reality terminal and the extracting method of target image and device |
-
2017
- 2017-04-28 CN CN201710296362.3A patent/CN107256017B/en not_active Expired - Fee Related
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20060155394A1 (en) * | 2004-12-16 | 2006-07-13 | International Business Machines Corporation | Method and apparatus for order-preserving clustering of multi-dimensional data |
| CN101271465A (en) * | 2007-12-12 | 2008-09-24 | 北京航空航天大学 | Lens clustering method based on information bottleneck theory |
| US8452531B2 (en) * | 2009-02-04 | 2013-05-28 | Mitac International Corp. | Route guiding system and method thereof |
| CN102063707A (en) * | 2011-01-05 | 2011-05-18 | 西安电子科技大学 | Mean shift based grey relation infrared imaging target segmentation method |
| CN103268729A (en) * | 2013-05-22 | 2013-08-28 | 北京工业大学 | Mobile robot cascading type map creating method based on mixed characteristics |
| CN106060052A (en) * | 2016-06-02 | 2016-10-26 | 深圳市豹风网络股份有限公司 | Three-dimensional navigation method for network game of mobile terminal |
| CN106097424A (en) * | 2016-06-12 | 2016-11-09 | 中国农业大学 | The method for visualizing of a kind of Agro-ecology scene and device |
| CN106295639A (en) * | 2016-08-01 | 2017-01-04 | 乐视控股（北京）有限公司 | A kind of virtual reality terminal and the extracting method of target image and device |
Non-Patent Citations (6)
| Title |
|---|
| CHUNMING CHENG 等: "Rapid Construction of Virtual Scenes for Large-Scale Crops Distribution", 《IFAC PAPERSONLINE》 * |
| ROBERT LORWAY: "Reassembling epidemiology: Mapping, monitoring and making-up people in the context of HIV prevention in Indiapeople in the context of HIV prevention in India", 《SOCIAL SCIENCE & MEDICINE》 * |
| 修春华 等: "三维场景中虚拟漫游路径的优化设计方法", 《金属矿山》 * |
| 许爱军: "虚拟校园三维场景构建与漫游导航的实现", 《计算机技术与发展》 * |
| 贾晨 等: "基于宽度学习方法的多模态信息融合", 《智能系统学报》 * |
| 陈勇 等: "一种三维虚拟场景自动漫游的快速路径规划算法", 《系统仿真学报》 * |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109260709A (en) * | 2018-09-11 | 2019-01-25 | 深圳市创梦天地科技有限公司 | A kind of generation method and device of dynamic route |
| CN109260709B (en) * | 2018-09-11 | 2022-02-22 | 深圳市创梦天地科技有限公司 | Dynamic route generation method and device |
| CN110705619A (en) * | 2019-09-25 | 2020-01-17 | 南方电网科学研究院有限责任公司 | Fog concentration grade judging method and device |
| CN110766770A (en) * | 2019-10-16 | 2020-02-07 | 腾讯科技（深圳）有限公司 | Thermodynamic diagram generation method and device, readable storage medium and computer equipment |
| CN110766770B (en) * | 2019-10-16 | 2021-07-20 | 腾讯科技（深圳）有限公司 | Thermodynamic diagram generation method and device, readable storage medium and computer equipment |
| CN111035928A (en) * | 2019-11-19 | 2020-04-21 | 腾讯科技（深圳）有限公司 | Method and device for displaying travel route, storage medium and electronic device |
| CN111881939A (en) * | 2020-06-24 | 2020-11-03 | 东南大学 | Shared single-vehicle parking area layout method based on clustering algorithm |
| CN112729327A (en) * | 2020-12-24 | 2021-04-30 | 浙江商汤科技开发有限公司 | Navigation method, navigation device, computer equipment and storage medium |
| CN112729327B (en) * | 2020-12-24 | 2024-06-07 | 浙江商汤科技开发有限公司 | Navigation method, navigation device, computer equipment and storage medium |
| CN113075933A (en) * | 2021-03-30 | 2021-07-06 | 北京布科思科技有限公司 | Robot passage control method, device and equipment |
| CN113075933B (en) * | 2021-03-30 | 2023-08-29 | 北京布科思科技有限公司 | Robot passing control method, device and equipment |
| CN117929173A (en) * | 2024-03-18 | 2024-04-26 | 中国汽车技术研究中心有限公司 | Method and device for testing and calibrating mechanical properties of rib components of automobile collision dummy |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107256017B (en) | 2020-08-04 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107256017A (en) | route planning method and system | |
| CN106897738B (en) | A kind of pedestrian detection method based on semi-supervised learning | |
| CN111861975A (en) | Method, device and computer program for assisting disease analysis, and method, device and program for training computer algorithm | |
| CN103489005B (en) | A kind of Classification of High Resolution Satellite Images method based on multiple Classifiers Combination | |
| CN107818302A (en) | Non-rigid multi-scale object detection method based on convolutional neural network | |
| CN108491776A (en) | Assembly part identification method and device based on pixel classification and monitoring system | |
| CN108537102A (en) | High Resolution SAR image classification method based on sparse features and condition random field | |
| CN110689093B (en) | Image target fine classification method under complex scene | |
| CN110281949B (en) | Unified hierarchical decision-making method for automatic driving | |
| CN110383291A (en) | Understand the method for the machine learning decision based on camera data | |
| CN110322099A (en) | The training activity that assessment user carries out in interactive computer simulation | |
| CN108197636A (en) | A kind of paddy detection and sorting technique based on depth multiple views feature | |
| CN110059656A (en) | The leucocyte classification method and system for generating neural network are fought based on convolution | |
| CN111709468B (en) | Training method and device for directional artificial intelligence and storage medium | |
| CN110619059A (en) | Building marking method based on transfer learning | |
| CN113537496A (en) | Deep learning model visual construction system and application and design method thereof | |
| CN109464808A (en) | Detect game plug-in method, apparatus and terminal | |
| García Balboa et al. | Generalization-oriented road line classification by means of an artificial neural network | |
| CN106529470A (en) | Gesture recognition method based on multistage depth convolution neural network | |
| CN110175519A (en) | Method and device for identifying separation and combination identification instrument of transformer substation and storage medium | |
| CN104463207B (en) | Knowledge autoencoder network and its polarization SAR image terrain classification method | |
| CN103793054B (en) | A kind of action identification method simulating declarative memory process | |
| CN108229505A (en) | Image classification method based on FISHER multistage dictionary learnings | |
| CN110288041A (en) | Chinese herbal medicine classification model construction method and system based on deep learning | |
| CN109948708A (en) | Multispectral image feature level information fusion method when more based on the implicit canonical of iteration |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20200804 Termination date: 20210428 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |