CN110146110A - A kind of error hiding determination method of indoor environment robot line feature ICNN data correlation - Google Patents
A kind of error hiding determination method of indoor environment robot line feature ICNN data correlation Download PDFInfo
- Publication number
- CN110146110A CN110146110A CN201910418114.0A CN201910418114A CN110146110A CN 110146110 A CN110146110 A CN 110146110A CN 201910418114 A CN201910418114 A CN 201910418114A CN 110146110 A CN110146110 A CN 110146110A
- Authority
- CN
- China
- Prior art keywords
- feature
- line
- line segment
- coordinate system
- robot
- 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
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01C—MEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY
- G01C25/00—Manufacturing, calibrating, cleaning, or repairing instruments or devices referred to in the other groups of this subclass
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01C—MEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY
- G01C9/00—Measuring inclination, e.g. by clinometers, by levels
Abstract
The invention belongs to data correlation fields, and in particular to a kind of error hiding determination method of indoor environment robot line feature ICNN data correlation.The present invention provides the improvement ICNN data correlation method of a kind of error hiding based on segment positions relationship and the decision rule for preventing error hiding.It is primarily based on laser sensor data extraction environment line feature, again line characteristic parameter is chosen, and the segment positions relationship on same two-dimensional surface is summarized, aiming at the problem that be easy to causeing error hiding when line segment is parallel or conllinear, error hiding scoring model is provided respectively, proposes a kind of improvement ICNN algorithm on this basis.Modified hydrothermal process has obtained higher association accuracy with the calculating time similar in canonical algorithm.
Description
Technical field
The invention belongs to data correlation fields, and in particular to a kind of indoor environment robot line feature ICNN data correlation
Error hiding determination method.
Background technique
Mobile robot simultaneous localization and mapping SLAM refers to that mobile robot is taken in circumstances not known by itself
The sensor of band detects ambient enviroment, and the map of incremental creation environment is simultaneously self-positioning.Wherein data correlation refers to when will be current
It carves the information that sensor obtains to be matched with the feature being added in map, data correlation precision is concerning to entire SLAM algorithm
Effect.The situation that independent compatible arest neighbors data association algorithm ICNN is suitble to environment relatively simple, have calculation amount small and
The features such as being easy storage, but error hiding is easy under the intensive environment of feature distribution.Document " is based on the matched rescue of Eigenvector
Robot builds drawing method " environment is specifically rescued for rescue robot, rotary motion and linear motion situation to robot
Under matching condition is set separately, analyze line segment feature be easy error hiding the case where, by line segment endpoint complete error hiding determine,
But the document only sets correlation threshold by the rotation angle of robot between adjacent two frame data and travel distance, works as robot
When observing same environmental entity again, it is unable to complete the association of line feature and the judgement of error hiding.The present invention will be for shifting
The line feature association under environment proposes the error hiding determination method based on ICNN to mobile robot indoors, and this method fully considers two
When line segment is parallel comprising with semi-inclusive relationship, it is conllinear when whether be the relationship of same environmental entity, and be directed to above-mentioned relation
Error hiding decision rule is set separately, when robot detects same environment again also it is possible to prevente effectively from error hiding, is improved
Association accuracy effectively prevents the error hiding when association of standard ICNN algorithm between parallel segment and conllinear line segment.
Summary of the invention
The present invention is low in order to solve the problems, such as standard ICNN algorithm association accuracy, proposes a kind of based on segment positions pass
The error hiding of system and the decision rule for preventing error hiding.
A kind of error hiding determination method of indoor environment robot line feature ICNN data correlation, this method includes following step
It is rapid:
Step 1: working sensor mode is arranged in building global coordinate system and local coordinate system；
Step 2: establishing line feature observation model；
Step 3: using improvement segmentation-polymerization extraction environment line feature；
Step 4: setting error hiding and its decision rule；
Step 5: the feature after determining data correlation retains principle.
The coordinate origin of global coordinate system described in step 1 is defined as the initial motion position of robot, and robot
Initial heading be X-direction；The local coordinate system includes laser sensor coordinate system, mobile robot local coordinate
System, the posture information of mobile robot areWherein, position coordinates of the robot under global coordinate system
For For course angle, laser sensor coordinate system xSoSyS, mobile robot local coordinate system xRoRyRAnd global ring
Border coordinate system xGoGyG, the coordinate origin o of laser sensor coordinate systemSPositioned at xRoRyRIn the x-axis direction of coordinate system, and two seats
The distance between the coordinate origin for marking system is set as a according to the actual installation position of robot geometric center and laser sensorc, swash
The posture information of optical sensor, i.e., under global coordinate system are as follows:The biography
Sense device working mode are as follows: the maximum length of perceived distance is 8m, and perception angular range is 0 °~180 °, i.e., every 0.5 ° perceives one
Range information amounts to 361 range informations.
The parameter setting of line feature described in step 2 are as follows:
Wherein, (xstart,ystart) it is coordinate of the line segment starting point under global coordinate system, (xend,yend) it is that line segment terminal exists
Coordinate under global coordinate system, L are line segment length, angular coordinate θstart、θend、θmiddleIt is line segment endpoint/midpoint and world coordinates
The line of origin and the angle of x-axis；
Line segment endpoint and the coordinate transformation relation at midpoint are expressed as following formula:
Wherein,It is polar form of the line segment starting point under robot local coordinate system；It is line
Coordinate of the Duan Qidian under global coordinate system；It is posture information of the robot under world coordinates；
The calculation formula of angular coordinate parameter:
Wherein, θstart、θend、θmiddleRespectively line segment starting point, terminal, midpoint angular coordinate；R represents line segment in world coordinates
Under polar coordinates distance parameter；L represents line segment length；
Obtained environmental data point is denoted as P={ P0,P1,...,P360, each point feature is calculate by the following formula in machine
Coordinate under people's coordinate systemEuclidean distance D between consecutive pointsiAre as follows:
Wherein i, j=1,2 ..., 360, then judge each Euclidean distance DjValue and threshold value DthSize, judged with this
Whether two adjacent points are divided into the same area；If Dj> Dth, then in point Pi-1With point PiBetween by region segmentation, then region P
It is divided into two isolated areas, then obtains subregion C0And C1:
P={ C0,C1}={ { P0,…,Pi-1},{Pi..., P360}}
Then again to region C1It is split according to above-mentioned method, until whole subset CiThe distance between midpoint and point
DiAll meet threshold value constraint condition, finally deletes points and be less than NminRegion, can finally obtain N number of area disconnected with each other
Domain { C1,C2,...CN, each region is fitted straight line, sets dynamic threshold are as follows:
In, cos β=d/r, r are the range data that laser range finder directly measures, and d arrives the range formula of straight line according to point
It is calculated, and r >=d.
It is most importantly the selection of threshold value in segmentation-polymerization described in step 3, the selection of threshold value should be with current region
The length D of first and last point lineseWith the distance D apart from longest point to first and last point linemaxIt is related, so setting ratio parameter:
Determine whether to divide current region by judging the size of scale parameter, sets ρ parameter by actual conditions
Threshold value is ρth=0.05, if calculating parameter obtains ρ > ρth, then current region is split；Conversely, not dividing；
Straight line fitting just will be carried out to each point set region after the completion of point feature segmentation, is obtained directly by least square method
Line equation determines the endpoint and line segment parameter of line segment；
A certain moment k, X in mobile robot traveling processk=(xk,yk,θk) represent the present bit of mobile robot
Appearance, the n observing environment feature that robot newly measures, is denoted as D={ D1,D2,...,Dn, have m map environment feature, note
For M={ M1,M2,...,Mm, if it is determined that being then denoted as N={ N for new feature1,N2,...,Np, and it is added to map feature concentration.
Line segment feature Di, MjPlace straight line is parallel to each other, and meets this condition there are two types of situation at this time, that is, includes and semi-inclusive
Relationship, θi,start、θi,endRespectively observational characteristic DiBeginning and end angular coordinate, θ under world coordinatesj,start、θj,endRespectively
For map feature MjThe beginning and end angular coordinate under world coordinates, error hiding and its decision rule described in step 4 are as follows:
Step 4.1: if line segment DiInclude line segment Mj, that is, meet relationship θi,start< θj,start< θj,end< θi,end, then it is assumed that
The two belongs to the same environmental characteristic；
Step 4.2: if line segment DiWith line segment MjFor semi-inclusive relationship, that is, meet relationship θi,start< θj,start< θi,end<
θj,end, then it is assumed that the two belongs to the same environmental characteristic；
Step 4.3: line segment DiAnd MjThe straight line at place is conllinear, and two line segments belong to the different piece of same environmental characteristic at this time,
If line segment DiWith line segment MjAngular coordinate meet relationship θi,end-θj,startIt is special to think that the two belongs to the same environment at this time by < Δ θ
Sign；Otherwise it is assumed that the two is for different features.
The step 5 includes:
Step 5.1: being a line segment, without over the ground if observational characteristic is almost overlapped with associated map feature
Figure feature set is updated, directly deletion observational characteristic, retains the feature in map feature set；
Step 5.2: if observational characteristic is a part of associated map feature, directly deleting observational characteristic, reservation
Feature in figure feature set；
Step 5.3: if observational characteristic partially overlaps with associated map feature, calculating separately two line segment endpoint lines
Midpoint, with two midpoint (xSm, ySm)、(xEm,yEm) be endpoint formed new line segment be added to map feature concentration, and
Delete original observational characteristic and map feature.
The beneficial effects of the present invention are:
The present invention provides the improvement of a kind of error hiding based on segment positions relationship and the decision rule for preventing error hiding
ICNN data correlation method.It is primarily based on laser sensor data extraction environment line feature, chooses line characteristic parameter again, and will
Segment positions relationship on same two-dimensional surface is summarized, and be easy to cause asking for error hiding when parallel or conllinear for line segment
Topic, provides error hiding scoring model respectively, proposes a kind of improvement ICNN algorithm on this basis.Modified hydrothermal process is calculated with standard
The method similar calculating time has obtained higher association accuracy.
Detailed description of the invention
Fig. 1 is composite coordinate system schematic of the invention.
Fig. 2 is line characteristic coordinates transition diagram of the invention.
Fig. 3 is selection of dynamic threshold schematic diagram of the invention.
Fig. 4 is dynamic threshold segmentation initial data schematic diagram of the invention.
Fig. 5 is dynamic threshold segmentation result schematic diagram of the invention.
Fig. 6 is the segment positions relationship of easy error hiding of the invention.
Fig. 7 (a) is inclusion relation schematic diagram when line segment of the invention is parallel.
Fig. 7 (b) is semi-inclusive relation schematic diagram when line segment of the invention is parallel.
Fig. 7 (c) is the collinear relationship schematic diagram of straight line where line segment of the invention.
Fig. 8 (a) is that two linked characters when feature of the invention retains principle are parallel to each other relation schematic diagram.
Fig. 8 (b) is that two linked characters when feature of the invention retains principle are inclusion relation schematic diagram.
Fig. 8 (c) is that two linked characters when feature of the invention retains principle are semi-inclusive relation schematic diagram.
Fig. 8 (d) is that endpoint when feature of the invention retains principle updates schematic diagram.
Fig. 9 is the line segment schematic diagram in test data of the invention.
Figure 10 is the error hiding schematic diagram in standard ICNN data correlation of the invention.
Figure 11 is the improvement ICNN data correlation result figure of the invention based on error hiding decision rule.
Specific embodiment
The present invention will be further described with reference to the accompanying drawing.
A kind of improvement ICNN data correlation method based on error hiding decision rule towards line feature, including following step
It is rapid:
Step 1: working sensor mode is arranged in building global coordinate system and local coordinate system；
The coordinate origin of global coordinate system is defined as the initial motion position of robot, and the initial heading of robot is
X-direction.
Local coordinate system includes laser sensor coordinate system, mobile robot local coordinate system, and wherein laser sensor is sat
The coordinate origin of mark system is located in the X-direction of robot coordinate system, and the distance between the coordinate origin of two coordinate systems is pressed
A is set as according to robot simplified modelc。
The operating mode of laser sensor are as follows: the maximum length of perceived distance is 8m, and perception angular range is 0o~180o,
I.e. every 0.5o perceives a range information, and the time for returning to a frame laser data is 26ms or 1s, amounts to 361 distance letters
Breath.
Step 2: establish line feature observation model:
Since robot is ceaselessly moving, the robot of different moments works as the sensing results of same feature with robot
Preceding position is related, just can guarantee environmental characteristic so needing for the sensing results at current time to be transformed under global coordinate system
Position is remained unchanged with respect to global coordinate system, and within the unit sampling period, and laser range finder acquires the process of a data
In, the travel distance of mobile robot is very short, this means that very big change in location will not occur suddenly for ambient enviroment feature.
So introducing endpoint-midpoint-angle parameter in online characteristic parameter in order to more accurately describe environment line feature to establish
The measurement model of line feature.
A line segment is there are certain the constraint relationship between two line segment endpoints for intuitively, and this constraint relationship is
The linear equation of straight line, further determines the positional relationship of point-to-point transmission with line segment length where line segment.The measurement of online feature
In model, starting point, terminal, midpoint and the line segment length of line segment are introduced respectively, and are defined in two endpoints and line segment of line segment
Angular coordinate of the point under polar coordinates, and determine the transformational relation between local coordinate system and global coordinate system.
Step 3: using improvement segmentation-aggregating algorithm extraction environment line feature；
The initial data that laser range finder obtains is environment point feature, indoors structured environment, and point is turned to line, is reduced
Calculation amount.Calculate Euclidean distance between points first, by compared with dynamic threshold, by point set be divided into it is multiple mutually not
The independent subregion of connection, then further segmented and cut by improved segmentation-aggregating algorithm for each subregion, finally to every
Sub-regions extract an environment line feature.Segmentation-aggregating algorithm focuses on the selection of segmentation threshold, in order to accurately extract
Environment line feature, chooses dynamic threshold segmentation herein.
Step 4: setting error hiding and its decision rule；
When carrying out the association of line feature using standard ICNN algorithm, it is associated with accuracy to improve, needs to consider that observation is special
Positional relationship between map feature of seeking peace, analysis can satisfy the positional relationship of Correlation Criteria, and line segment is easy to make by the present invention
It summarizes at the positional relationship of error hiding, and obtain error hiding decision rule for every case, when there is multiple features to fall
When entering in correlation threshold, not using and choosing the smallest feature of mahalanobis distance in standard ICNN algorithm is optimal relevance object, and
It is that optimal relevance object is selected by error hiding decision rule, prevents the error hiding between feature as far as possible, improves association accuracy.
Step 5: the feature after data correlation retains principle；
After having carried out a data correlation, observational characteristic is obtained with it and predicts being associated with pair for observation, is needed at this time over the ground
Figure feature set is updated, that is, needs to consider the reservation principle of feature.For two associated features, consider in world coordinates
The lower positional relationship of system there are a possibility that, and decision criteria is set forth for each case and retains principle, completion pair
The update of map feature collection.
1 standard ICNN data correlation result of table
The improved ICNN data correlation result of table 2
Global coordinate system and local coordinate system:
As shown in Figure 1, the posture information of mobile robot isWherein, robot is in world coordinates
System under position coordinates be For course angle.Laser sensor coordinate system xSoSyS, mobile robot local coordinate system
xRoRyRAnd global context coordinate system xGoGyG.Also, the coordinate origin o of laser sensor coordinate systemSPositioned at xRoRyRCoordinate system
X-axis direction on, and the distance between coordinate origin of two coordinate systems is according to robot geometric center and laser sensor
Actual installation position is set as ac；The coordinate origin of global coordinate system is defined as the initial motion position of robot, and robot
Initial heading be xGAxis direction.
According to the posture information of robot and the positional relationship of robot and laser range finder is combined to be known that laser
The posture information of sensor (under global coordinate system) are as follows:
Environment line feature observation model:
As shown in Fig. 2, the parameter setting of environment line feature of the invention are as follows:
Wherein, (xstart,ystart) it is coordinate of the line segment starting point under global coordinate system, (xend,yend) it is that line segment terminal exists
Coordinate under global coordinate system, L are line segment length, angular coordinate θstart、θend、θmiddle(hereinafter referred to as line segment endpoint/midpoint angle
Coordinate) it is line segment endpoint/midpoint and the line of world coordinates origin and the angle of x-axis.
Line segment endpoint and the coordinate transformation relation at midpoint are shown below (by taking line segment starting point as an example):
Wherein,It is polar form of the line segment starting point under robot local coordinate system；It is line
Coordinate of the Duan Qidian under global coordinate system；It is posture information of the robot under world coordinates.
The calculation formula of angular coordinate parameter:
Wherein, θstart、θend、θmiddleRespectively line segment starting point, terminal, midpoint angular coordinate；R represents line segment in world coordinates
Under polar coordinates distance parameter；L represents line segment length.
Extraction environment line feature:
Obtained environmental data point is denoted as P={ P0,P1,...,P360, each point feature can be calculated by following formula and existed
Coordinate under robot coordinate systemEuclidean distance D between consecutive pointsiAre as follows:
Wherein i, j=1,2 ..., 360.
Then judge each Euclidean distance DjValue and threshold value DthSize, judge whether two adjacent points can be with this
It is divided into the same area.If Dj> Dth, then in point Pi-1With point PiBetween by region segmentation, then region P is divided into two independent zones
Domain then obtains subregion C0And C1:
P={ C0,C1}={ { P0,…,Pi-1},{Pi..., P360}}
Then again to region C1It is split according to above-mentioned method, until whole subset CiThe distance between midpoint and point
DiAll meet threshold value constraint condition.It finally deletes points and is less than NminRegion, can finally obtain N number of area disconnected with each other
Domain { C1,C2,...CN, substantially each region can be fitted straight line.
Among these, as shown in figure 3, two stains are two adjacent characteristic points in raw measurement data in figure, d is sensor
To the distance (i.e. vertical line distance) of straight line where characteristic point line, DthFor region segmentation threshold value,For the scanning of laser range finder
It is spaced 0.5o, line distance of the r between laser sensor and one of data point, β is the angle corresponding to it.Based on
Upper analysis, it can be seen that the dynamic threshold for needing to set should be with the distance dependent of robot and feature, and more intuitive performance is just
Be in the variation of angle beta: when robot is closer at a distance from feature, i.e., d becomes smaller, and β is bigger at this time, then the Euclidean of point-to-point transmission
Apart from smaller.Therefore setting dynamic threshold are as follows:
Wherein, cos β=d/r, r are the range data that laser range finder directly measures, and d can arrive the distance of straight line according to point
Formula is calculated (point is robot current position coordinates, and straight line is straight line obtained by two characteristic points are connected), and r >=d,
And the measurement angle that the present invention choosesSoThen substituting into can be by upper conversion are as follows:
Dynamic threshold segmentation initial data schematic diagram of the invention is as shown in figure 4, pass through dynamic threshold segmentation result such as Fig. 5
Shown, the number in figure represents cut zone number, it can be seen that segmentation result more completely remains environmental data, is conducive to
Completely extraction environment line feature.
Characteristic point region segmentation is cut:
The feature point set that laser range finder obtains preliminarily is divided into multiple connected regions, each characteristic area by us
May there was only a line segment, it is also possible to the case where there are a plurality of line segments.Straight line then can be directly carried out if the first situation
Fitting, but if second situation, then it must further divide just the parameter that can obtain a plurality of line segment to the region.
Segmentation-aggregating algorithm main thought is that the first and last point in data point set region to be split is connected to determining one directly
Line obtains judging the size relation of this distance with the threshold value of setting apart from the maximum characteristic point of the linear distance, if the distance than
Threshold value is small, then it is assumed that whole characteristic points in data point set region belong to same straight line, are directly fitted all data points
For straight line；, whereas if the distance is greater than threshold value, it is determined that the point is cut-point, from cut-point by point set region segmentation
At two sub-regions, aforesaid operations then are repeated for two sub-regions respectively, until the characteristic point in all subregions is to directly
The distance of line is respectively less than threshold value, then stops dividing, and carries out straight line fitting to a each sub-regions.
It can be seen that being most importantly the selection of threshold value in segmentation-aggregating algorithm, the selection of threshold value should be with current region first and last
The length D of point lineseWith farthest point to the distance D of first and last point linemaxIt is related, so setting ratio parameter:
Determine whether to divide current region by judging the size of scale parameter, sets ρ parameter by actual conditions
Threshold value is ρth=0.05, if calculating parameter obtains ρ > ρth, then current region is split；Conversely, not dividing.
Straight line fitting can be carried out to each point set region after the completion of point feature is divided, obtained by least square method
To linear equation, the endpoint and other line segment parameters of line segment may further be determined.
Assuming that a certain moment k, X in mobile robot traveling processk=(xk,yk,θk) represent the current of mobile robot
Pose, the n observing environment feature that robot newly measures, is denoted as D={ D1,D2,...,Dn, have m map environment feature,
It is denoted as M={ M1,M2,...,Mm, if it is determined that being then denoted as N={ N for new feature1,N2,...,Np, and it is added to map feature collection
In.
The limitation of standard ICNN algorithm:
When carrying out the association of line feature using standard ICNN algorithm, need to consider between observational characteristic and map feature
Positional relationship, there are three types of the positional relationship possibility that can satisfy Correlation Criteria:
(1) two feature is similar to be overlapped, and is the same a part for belonging to same scanning entity；
(2) two characteristic overlappings are the adjacent parts for belonging to same scanning entity；
(3) two line segments are conllinear, are to belong to different scanning entities.
Among these, first two is correctly to match, but the third then belongs to error hiding.For this purpose, the present invention is by two line segments
It is easy to cause the positional relationship of error hiding to summarize as shown in fig. 6, it should be noted that observational characteristic DiWith map feature MjIt is affiliated straight
Line is in same two-dimensional surface.
Error hiding and its decision rule are set:
Line segment feature Di, MjPlace straight line is parallel to each other, and meets this condition there are two types of situation at this time, that is, includes and semi-inclusive
Relationship, as shown in Figure 7.
θ in figurei,start、θi,endRespectively observational characteristic DiBeginning and end angular coordinate, θ under world coordinatesj,start、
θj,endRespectively map feature MjThe beginning and end angular coordinate under world coordinates, can be with by comparing the size of corresponding angular coordinate
Judge the positional relationship of two lines section.
If (a) line segment DiInclude line segment Mj(shown in such as Fig. 7 (a)), that is, meet relationship θi,start< θj,start< θj,end<
θi,end, then it is assumed that the two belongs to the same environmental characteristic.
If (b) line segment DiWith line segment MjFor semi-inclusive relationship (shown in such as Fig. 7 (b)), that is, meet relationship θi,start<
θj,start< θi,end< θj,end, then it is assumed that the two belongs to the same environmental characteristic.
(c) line segment DiAnd MjThe straight line at place is conllinear, at this time two line segments may belong to same environmental characteristic different piece or
Person is to belong to varying environment feature, as shown in Fig. 7 (c).If line segment DiWith line segment MjAngular coordinate meet relationship θi,end-
θj,start< Δ θ thinks that the two belongs to the same environmental characteristic at this time；Otherwise it is assumed that the two is for different features.
Feature after data correlation retains principle:
After having carried out a data correlation, observational characteristic is obtained with it and predicts being associated with pair for observation, is needed at this time over the ground
Figure feature set is updated, and both has needed to consider the reservation principle of feature.For two associated features, consider in world coordinates
Positional relationship under system there is a possibility that three kinds:
(1) observational characteristic is almost overlapped with associated map feature, is approximately a line segment, such as Fig. 8 (a).
(2) observational characteristic is a part of associated map feature, such as Fig. 8 (b).
(3) observational characteristic partially overlaps with associated map feature, such as Fig. 8 (c).
Decision criteria is set forth herein for three cases above and retains principle.As shown in figure 8, line of observation feature
Endpoint be (xDS,yDS)、(xDE,yDE), the endpoint of ground topographical features is (xMS,yMS)、(xME,yME), by analyzing between them
Geometrical relationship determine the reservation principle of map feature.
(a) when the endpoint of two linked characters meets following formula:
Wherein δ is given threshold, is generally set to 0.05 times of distance R of origin to observational characteristic, i.e. δ=0.05R.
If meeting above formula, it may be considered that two lines section is almost that same line segment is not necessarily at this time as shown in Fig. 8 (a)
To map feature set is updated, directly deletion observational characteristic, retains the feature in map feature set.
(b) when the endpoint of two linked characters meets following formula:
If meeting above formula, then it is assumed that observational characteristic is a part of associated map feature, as shown in Fig. 8 (b).This
When it is identical as situation 1, directly deletion observational characteristic, retain map feature set in feature.
(c) when the endpoint of two linked characters meets following formula:
If meeting above formula, then it is assumed that observational characteristic partially overlaps with associated map feature, as shown in Fig. 8 (c).This
The parameter that Shi Chongxin updates line segment calculates separately the midpoint of two line segment endpoint lines, as shown in Fig. 8 (d) with two midpoints
(xSm, ySm)、(xEm,yEm) be that endpoint forms new line segment and is added to map feature concentration, and delete original observational characteristic and
Map feature.
The improvement ICNN line characteristic correlating method specific steps based on error hiding decision rule that the present invention designs are such as
Under:
Step1: according to the pose of current time mobile robot, gained linear equation is fitted to calculate under global coordinate system
Observational characteristic collection line characteristic parameter；
Step2: for the feature D of observational characteristic collectioniWith the feature M of map feature collectionjThe matching of ICNN data correlation is carried out,
In this process, 1 is set to corresponding incidence matrix element to the feature fallen into mahalanobis distance thresholding；
Step3: judge the feature D of observational characteristic collection in data correlation matrixiThe number NUM of " 1 " in being expert at, if NUM
=0, then this feature is added into map feature as new feature and concentrated；If NUM=1, the map feature of " 1 " column is made
To be correctly associated with；If NUM > 1 is transferred to Step4；
Step4: if NUM > 1, the association in mahalanobis distance thresholding will be fallen into, the angular coordinate successively judging line segment is closed
System chooses best match according to error hiding decision rule；
Step5: output linked character pair terminates this data correlation calculating process.
Indoors in environment, the LMS200 type laser range finder for the SICK company for utilizing " tourist No. four " robot to be equipped with
Perception environment simultaneously extracts outlet feature, and mahalanobis distance threshold value is set as dij=0.053m.In test data, the spy of map feature concentration
The line feature that observational characteristic of seeking peace is concentrated is as shown in figure 9, the line feature that wherein blue solid lines character representation map feature is concentrated, total
10 line features；Red dotted line indicates the line feature that observational characteristic is concentrated, totally 10 line features.It is carried out using standard ICNN algorithm
When data correlation, the results are shown in Table 1 for data correlation, there is shown the incidence relation of 10 observational characteristics and map feature,
It can be seen that wherein be no lack of the data correlation of mistake in conjunction with the segment positions relationship in Fig. 9, the observational characteristic number of erroneous association is 4
Item, association accuracy (association accuracy calculatings use correctly associated line segment number account for number of bus segments percentage expression) for
60%, Riming time of algorithm 0.2024s.The error hiding for wherein comparing concentration is indicated in Figure 10, blue solid lines table in figure
Pictorial map feature M, red dotted line indicate observational characteristic D, and associated feature is indicated with same color, such as feature pair
{ D10, M9 } is indicated with black.
ICNN algorithm after application enhancements re-starts data correlation, and the results are shown in Table 2 for obtained data correlation, control
Fig. 9 can be seen that improved algorithm is effectively improved in terms of being associated with accuracy, reach 100%, but the operation of algorithm
Time, improved Riming time of algorithm was 0.2103s there is no significantly extending, and increased only about 0.008s.By this hair
It is as shown in figure 11 that the error hiding decision rule of bright proposition chooses the partial association result that Optimum Matching obtains.
Claims (6)
1. a kind of error hiding determination method of indoor environment robot line feature ICNN data correlation, which is characterized in that this method
The following steps are included:
Step 1: working sensor mode is arranged in building global coordinate system and local coordinate system；
Step 2: establishing line feature observation model；
Step 3: using improvement segmentation-polymerization extraction environment line feature；
Step 4: setting error hiding and its decision rule；
Step 5: the feature after determining data correlation retains principle.
2. a kind of error hiding judgement side of indoor environment robot line feature ICNN data correlation according to claim 1
Method, it is characterised in that: the coordinate origin of global coordinate system described in step 1 is defined as the initial motion position of robot, and
The initial heading of robot is X-direction；The local coordinate system includes laser sensor coordinate system, mobile robot part
The posture information of coordinate system, mobile robot isWherein, position of the robot under global coordinate system
Coordinate is For course angle, laser sensor coordinate system xSoSyS, mobile robot local coordinate system xRoRyRAnd it is complete
Office environment coordinate system xGoGyG, the coordinate origin o of laser sensor coordinate systemSPositioned at xRoRyRIn the x-axis direction of coordinate system, and two
The distance between coordinate origin of a coordinate system is set as according to the actual installation position of robot geometric center and laser sensor
ac, the posture information of laser sensor, i.e., under global coordinate system are as follows:It is described
Working sensor mode are as follows: the maximum length of perceived distance be 8m, perception angular range be 0 °~180 °, i.e., it is every 0.5 ° perception
One range information amounts to 361 range informations.
3. a kind of error hiding judgement side of indoor environment robot line feature ICNN data correlation according to claim 1
Method, which is characterized in that the parameter setting of line feature described in step 2 are as follows:
Wherein, (xstart,ystart) it is coordinate of the line segment starting point under global coordinate system, (xend,yend) it is line segment terminal in the overall situation
Coordinate under coordinate system, L are line segment length, angular coordinate θstart、θend、θmiddleIt is line segment endpoint/midpoint and world coordinates origin
Line and x-axis angle；
Line segment endpoint and the coordinate transformation relation at midpoint are expressed as following formula:
Wherein,It is polar form of the line segment starting point under robot local coordinate system；It is that line segment rises
Coordinate of the point under global coordinate system；It is posture information of the robot under world coordinates；
The calculation formula of angular coordinate parameter:
Wherein, θstart、θend、θmiddleRespectively line segment starting point, terminal, midpoint angular coordinate；R represents line segment under world coordinates
Polar coordinates distance parameter；L represents line segment length；
Obtained environmental data point is denoted as P={ P0,P1,...,P360, it is calculate by the following formula each point feature and is sat in robot
Coordinate under mark systemEuclidean distance D between consecutive pointsiAre as follows:
Wherein i, j=1,2 ..., 360, then judge each Euclidean distance DjValue and threshold value DthSize, judged with this adjacent
Two points whether be divided into the same area；If Dj> Dth, then in point Pi-1With point PiBetween by region segmentation, then region P is divided
For two isolated areas, then subregion C is obtained0And C1:
P={ C0,C1}={ { P0,…,Pi-1},{Pi,…,P360}}
Then again to region C1It is split according to above-mentioned method, until whole subset CiThe distance between midpoint and point DiAll
Meet threshold value constraint condition, finally deletes points and be less than NminRegion, can finally obtain N number of region disconnected with each other
{C1,C2,...CN, each region is fitted straight line, sets dynamic threshold are as follows:
In, cos β=d/r, r are the range data that laser range finder directly measures, and d is calculated according to the range formula of point to straight line
It obtains, and r >=d.
4. a kind of error hiding judgement side of indoor environment robot line feature ICNN data correlation according to claim 1
Method, which is characterized in that be most importantly the selection of threshold value in segmentation-polymerization described in step 3, the selection of threshold value should with work as
The length D of forefoot area first and last point lineseWith the distance D apart from longest point to first and last point linemaxIt is related, so setting ratio is joined
Number:
Determine whether to divide current region by judging the size of scale parameter, the threshold value of ρ parameter is set by actual conditions
For ρth=0.05, if calculating parameter obtains ρ > ρth, then current region is split；Conversely, not dividing；
Straight line fitting just will be carried out to each point set region after the completion of point feature segmentation, straight line side is obtained by least square method
Journey determines the endpoint and line segment parameter of line segment；
A certain moment k, X in mobile robot traveling processk=(xk,yk,θk) represent the current pose of mobile robot, machine
The n observing environment feature that people newly measures, is denoted as D={ D1,D2,...,Dn, have m map environment feature, is denoted as M=
{M1,M2,...,Mm, if it is determined that being then denoted as N={ N for new feature1,N2,...,Np, and it is added to map feature concentration.
5. a kind of error hiding judgement side of indoor environment robot line feature ICNN data correlation according to claim 1
Method, it is characterised in that: line segment feature Di, MjPlace straight line is parallel to each other, and meets this condition there are two types of situation at this time, i.e., comprising and
Semi-inclusive relationship, θi,start、θi,endRespectively observational characteristic DiBeginning and end angular coordinate, θ under world coordinatesj,start、
θj,endRespectively map feature MjThe beginning and end angular coordinate under world coordinates, error hiding described in step 4 and its determines rule
Then are as follows:
Step 4.1: if line segment DiInclude line segment Mj, that is, meet relationship θi,start< θj,start< θj,end< θi,end, then it is assumed that the two
Belong to the same environmental characteristic；
Step 4.2: if line segment DiWith line segment MjFor semi-inclusive relationship, that is, meet relationship θi,start< θj,start< θi,end< θj,end,
Then think that the two belongs to the same environmental characteristic；
Step 4.3: line segment DiAnd MjThe straight line at place is conllinear, and two line segments belong to the different piece of same environmental characteristic at this time, if line
Section DiWith line segment MjAngular coordinate meet relationship θi,end-θj,start< Δ θ thinks that the two belongs to the same environmental characteristic at this time；It is no
Then think that the two is for different features.
6. a kind of error hiding judgement side of indoor environment robot line feature ICNN data correlation according to claim 1
Method, which is characterized in that the step 5 includes:
Step 5.1: it is a line segment if observational characteristic is almost overlapped with associated map feature, it is special without to map
Collection is updated, directly deletion observational characteristic, retains the feature in map feature set；
Step 5.2: if observational characteristic is a part of associated map feature, directly deletion observational characteristic, reservation map are special
Feature in collection；
Step 5.3: if observational characteristic partially overlaps with associated map feature, calculating separately the midpoint of two line segment endpoint lines
Position, with two midpoint (xSm, ySm)、(xEm,yEm) it is that endpoint forms new line segment and is added to map feature concentration, and it deletes
Observational characteristic and map feature originally.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910418114.0A CN110146110B (en) | 2019-05-20 | 2019-05-20 | Mismatching judgment method for ICNN data association of robot line characteristics in indoor environment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910418114.0A CN110146110B (en) | 2019-05-20 | 2019-05-20 | Mismatching judgment method for ICNN data association of robot line characteristics in indoor environment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110146110A true CN110146110A (en) | 2019-08-20 |
| CN110146110B CN110146110B (en) | 2022-11-15 |
Family
ID=67592255
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910418114.0A Active CN110146110B (en) | 2019-05-20 | 2019-05-20 | Mismatching judgment method for ICNN data association of robot line characteristics in indoor environment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110146110B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114322943A (en) * | 2021-12-15 | 2022-04-12 | 南京航空航天大学 | Target distance measuring method and device based on forward-looking image of unmanned aerial vehicle |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20080151811A1 (en) * | 2006-12-20 | 2008-06-26 | Hersham El-Damhougy | Interplanetary communications network, interplanetary communications network backbone and method of managing interplanetary communications network |
| US20080273791A1 (en) * | 2006-07-05 | 2008-11-06 | Samsung Electronics Co., Ltd. | Apparatus, method, and medium for dividing regions by using feature points and mobile robot using the same |
| WO2012173496A1 (en) * | 2011-06-14 | 2012-12-20 | Crown Equipment Limited | Method and apparatus for facilitating map data processing for industrial vehicle navigation |
| CN103198751A (en) * | 2013-03-06 | 2013-07-10 | 南京邮电大学 | Line feature map creation method of mobile robot based on laser range finder |
| CN107766818A (en) * | 2017-10-18 | 2018-03-06 | 哈尔滨工程大学 | A kind of didactic submerged structure environment line feature extraction method |
| CN108061897A (en) * | 2017-12-05 | 2018-05-22 | 哈尔滨工程大学 | A kind of submerged structure environment line feature extraction method based on Forward-Looking Sonar |
| CN109000649A (en) * | 2018-05-29 | 2018-12-14 | 重庆大学 | A kind of all directionally movable robot pose calibration method based on right angle bend feature |
-
2019
- 2019-05-20 CN CN201910418114.0A patent/CN110146110B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20080273791A1 (en) * | 2006-07-05 | 2008-11-06 | Samsung Electronics Co., Ltd. | Apparatus, method, and medium for dividing regions by using feature points and mobile robot using the same |
| US20080151811A1 (en) * | 2006-12-20 | 2008-06-26 | Hersham El-Damhougy | Interplanetary communications network, interplanetary communications network backbone and method of managing interplanetary communications network |
| WO2012173496A1 (en) * | 2011-06-14 | 2012-12-20 | Crown Equipment Limited | Method and apparatus for facilitating map data processing for industrial vehicle navigation |
| CN103198751A (en) * | 2013-03-06 | 2013-07-10 | 南京邮电大学 | Line feature map creation method of mobile robot based on laser range finder |
| CN107766818A (en) * | 2017-10-18 | 2018-03-06 | 哈尔滨工程大学 | A kind of didactic submerged structure environment line feature extraction method |
| CN108061897A (en) * | 2017-12-05 | 2018-05-22 | 哈尔滨工程大学 | A kind of submerged structure environment line feature extraction method based on Forward-Looking Sonar |
| CN109000649A (en) * | 2018-05-29 | 2018-12-14 | 重庆大学 | A kind of all directionally movable robot pose calibration method based on right angle bend feature |
Non-Patent Citations (1)
| Title |
|---|
| 肖瑶: "自主移动机器人动态环境感知与构图研究", 《中国优秀博硕士学位论文全文数据库（硕士） 信息科技辑》 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114322943A (en) * | 2021-12-15 | 2022-04-12 | 南京航空航天大学 | Target distance measuring method and device based on forward-looking image of unmanned aerial vehicle |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110146110B (en) | 2022-11-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105957342B (en) | Track grade road plotting method and system based on crowdsourcing space-time big data | |
| Ghanea et al. | Building extraction from high-resolution satellite images in urban areas: Recent methods and strategies against significant challenges | |
| CN104240266A (en) | Target object tracking method based on color-structure features | |
| CN105955258A (en) | Robot global grid map construction method based on Kinect sensor information fusion | |
| CN110378196A (en) | A kind of road vision detection method of combination laser point cloud data | |
| CN103646249B (en) | A kind of greenhouse intelligent mobile robot vision navigation path identification method | |
| CN105894502A (en) | RGBD image salience detection method based on hypergraph model | |
| CN109949326A (en) | Contour of building line drawing method based on Backpack type three-dimensional laser point cloud data | |
| CN112325883B (en) | Indoor positioning method for mobile robot with WiFi and visual multi-source integration | |
| CN103226387A (en) | Video fingertip positioning method based on Kinect | |
| CN109655805A (en) | A kind of laser radar positioning method being overlapped length estimate based on scan lines | |
| CN103389799A (en) | Method for tracking motion trail of fingertip | |
| CN104063677A (en) | Equipment used for estimating human body posture and method thereof | |
| CN108415413A (en) | A kind of intelligent forklift part obstacle-avoiding route planning method based on round region of interest | |
| CN106599760B (en) | Method for calculating running area of inspection robot of transformer substation | |
| CN103793056A (en) | Mid-air gesture roaming control method based on distance vector | |
| CN108759829A (en) | A kind of local obstacle-avoiding route planning method of intelligent forklift | |
| CN110146110A (en) | A kind of error hiding determination method of indoor environment robot line feature ICNN data correlation | |
| WO2022111723A1 (en) | Road edge detection method and robot | |
| US11847739B2 (en) | Systems and methods for pitch determination | |
| CN102073848A (en) | Intelligent optimization-based road recognition system and method | |
| Qiao et al. | Research on abnormal pedestrian trajectory detection of dynamic crowds in public scenarios | |
| Oh et al. | Dynamic EKF-based SLAM for autonomous mobile convergence platforms | |
| CN105809719A (en) | Object tracking method based on pixel multi-coding-table matching | |
| CN110763223B (en) | Sliding window based indoor three-dimensional grid map feature point extraction method |
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 |