CN109960990A - Barrier detection credibility evaluation method - Google Patents
Barrier detection credibility evaluation method Download PDFInfo
- Publication number
- CN109960990A CN109960990A CN201711431445.5A CN201711431445A CN109960990A CN 109960990 A CN109960990 A CN 109960990A CN 201711431445 A CN201711431445 A CN 201711431445A CN 109960990 A CN109960990 A CN 109960990A
- Authority
- CN
- China
- Prior art keywords
- barrier
- score
- detection result
- different
- punishment
- 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
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/56—Context or environment of the image exterior to a vehicle by using sensors mounted on the vehicle
- G06V20/58—Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
Abstract
A kind of barrier detection credibility evaluation method, implemented by processing unit, include: (A) is after receiving present image, the the first current barrier detection for indicating barrier that first classification method is detected is obtained using the first classification method as a result, and obtaining the second current barrier detection result of the barrier for indicating that second classification method is detected using the second classification method;(B) determine that first and second described current barrier detection result whether there is at least one different barrier;And (C) when determining the described first current barrier detection result, there are when at least one different barrier from the described second current barrier detection result, the at least one punishment score for respectively corresponding at least one different barrier is obtained, and confidence score is subtracted into at least one punishment score.
Description
Technical field
The present invention relates to a kind of image real time transfers, more particularly to a kind of barrier detection credibility evaluation method.
Background technique
With the universal development with computer vision field of video camera, the application of intelligent picture control is human lives
Safety and convenience are brought, such as applied to intelligent advanced driving assistance system (Advanced Driver Assistance
Systems, ADAS) detecting of barrier is carried out to image captured by video camera using image recognition techniques, to alert driving
Red route situation, and in some cases by vehicle deceleration or stopping.
However, real road belongs to more complex environment, be easy to be influenced by the intervention of external environment factor, when video camera by
Intervention to external environment factor influences (for example, the car light direct projection of the vehicle in opposite lane, backlight or camera lens pollution)
When, barrier texture and profile in captured image also will receive influence, and then lead to the detecting misalignment of barrier.So
And the barrier detection method of existing ADAS system can only export the detecting result of barrier, can't provide its detecting knot
The confidence level of fruit, the accuracy that this will lead to the detecting result that user takes for ADAS system is high, and excessively believes and rely on
The detecting result of ADAS system.
Summary of the invention
The purpose of the present invention is to provide a kind of appraisal procedures of confidence level that can provide barrier detection in due course.
Barrier detection credibility evaluation method of the invention, by the processing of electrical connection storage unit and image capturing unit
Unit is implemented, and the storage unit is stored with confidence score, described image shooting unit constantly shoots and transmits a company
Image go here and there to the processing unit, the barrier detection credibility evaluation method includes a step (A), a step (B) and one
Step (C).
In the step (A), after the processing unit receives the present image from described image shooting unit, institute
It states processing unit and is obtained and stored using the first classification method according to the present image and indicate the first classification method institute
The current barrier detection of the first of the barrier detected according to the present image using the second classification method as a result, and obtained
And store the second current barrier detection result of the barrier for indicating that second classification method is detected.
In the step (B), the processing unit determines that the first current barrier detection result is worked as with described second
Preceding barrier detection result whether there is at least one different barrier.
In the step (C), when the processing unit determines the described first current barrier detection result and described the
Two current barrier detection results there are when at least one different barrier, the processing unit according at least to this at least one not
Same barrier obtains at least one punishment score for respectively corresponding at least one different barrier, and the confidence is divided
Number subtracts at least one punishment score, to update the confidence score.
Preferably, barrier detection credibility evaluation method of the invention, step (C) includes following sub-step:
(C-1) it is deposited with the described second current barrier detection result when determining the described first current barrier detection result
In at least one different barrier, for each of at least one different barrier, the described first current barrier is determined
Hinder whether object detecting result is not present the different barrier;
(C-2) when determining the described first current barrier detection result there is no when the different barrier, at least
It is obtained according to the different barrier and punishes score as first of one in at least one punishment score, and by the confidence
Degree score subtracts the first punishment score, to update the confidence score;And
(C-3) when determining the described first current barrier detection result there are when the different barrier, at least root
It is obtained according to the different barrier and punishes score as second of one in at least one punishment score, and by the confidence
Score subtracts the second punishment score, to update the confidence score.
Preferably, barrier detection credibility evaluation method of the invention, the storage unit are also stored with barrier pair
First look-up table of score punishes weight corresponding to the first of first classification method, and corresponds to the second classification side
Second punishment weight of method, first look-up table include it is multiple be relevant to practical obstacle object and described image shooting unit away from
From obstacle distance and multiple scores for corresponding respectively to the obstacle distance:
In step (A), the first current barrier detection result and the described second current barrier detection result are all
Comprising the barrier detected, and the barrier that is detected is located at the position of the present image;
In step (C-2), the processing unit is first to be located at the present image according to the different barrier
Position estimates practical obstacle object corresponding to the different barrier at a distance from described image shooting unit, and according to
The distance estimated and first look-up table obtain the score for corresponding to the different barrier, then will be obtained
Score is multiplied by the first punishment weight to obtain the first punishment score;And
In step (C-3), the processing unit is first to be located at the present image according to the different barrier
Position estimates practical obstacle object corresponding to the different barrier at a distance from described image shooting unit, and according to
The distance estimated and first look-up table obtain the score for corresponding to the different barrier, then will be obtained
Score is multiplied by the second punishment weight to obtain the second punishment score.
Preferably, barrier detection credibility evaluation method of the invention, the storage unit are also stored with barrier pair
First look-up table of score punishes weight corresponding to the first of first classification method, and corresponds to the second classification side
Second punishment weight of method, first look-up table include multiple obstacle species and multiple correspond respectively to the type
Score:
In step (A), the first current barrier detection result and the described second current barrier detection result are all
Comprising the barrier detected, and the type of barrier detected;
In step (C-2), the processing unit is searched according to the type of the different barrier and described first
Table obtains the score for corresponding to the different barrier, then score obtained is multiplied by the first punishment weight to obtain
Obtain the first punishment score;And
In step (C-3), the processing unit is searched according to the type of the different barrier and described first
Table obtains the score for corresponding to the different barrier, then score obtained is multiplied by the second punishment weight to obtain
Obtain the second punishment score.
Preferably, barrier detection credibility evaluation method of the invention also comprises the steps of: after step (A)
(D) according to previous image it is obtained indicate barrier that first classification method is detected first before
One barrier detection result and the described first current barrier detection as a result, determine the first previous barrier detection result with
The first current barrier detection result whether there is at least one different barrier;
(E) exist with the described first current barrier detection result when determining the described first previous barrier detection result
When at least one different barrier, at least one corresponding at least one difference is obtained according at least at least one different barrier
Barrier third punish score, and by the confidence score subtract an at least third punishment score, with update described in
Confidence score.
Preferably, barrier detection credibility evaluation method of the invention, the storage unit are also stored with barrier pair
The second look-up table of score, the second look-up table include multiple obstacle levels for being relevant to barrier and being located at the position of image
It sets and multiple scores for corresponding respectively to the Obstacle Position:
In step (A), the first current barrier detection result and the described second current barrier detection result are all
Comprising the barrier detected, and the barrier that is detected is located at the position of the present image;And
In step (E), the processing unit is to be located at the present image or described according to the different barrier
The position of previous image and the second look-up table obtain the score for corresponding to the different barrier, further according to acquisition
Score obtains the third and punishes score.
Preferably, barrier detection credibility evaluation method of the invention also comprises the steps of: after step (A)
(F) according to previous image it is obtained indicate barrier that second classification method is detected second before
One barrier detection result and the described second current barrier detection as a result, determine the second previous barrier detection result with
The second current barrier detection result whether there is at least one different barrier;
(G) exist with the described second current barrier detection result when determining the described second previous barrier detection result
When at least one different barrier, at least one corresponding at least one difference is obtained according at least at least one different barrier
Barrier the 4th punishment score, and by the confidence score subtract this at least one the 4th punishment score, with update described in
Confidence score.
Preferably, barrier detection credibility evaluation method of the invention, the storage unit are also stored with barrier pair
The second look-up table of score, the second look-up table include multiple obstacle levels for being relevant to barrier and being located at the position of image
It sets and multiple scores for corresponding respectively to the Obstacle Position:
In step (A), the first current barrier detection result and the described second current barrier detection result are all
Comprising the barrier detected, and the barrier that is detected is located at the position of the present image;And
In step (G), the processing unit is to be located at the present image or described according to the different barrier
The position of previous image and the second look-up table obtain the score for corresponding to the different barrier, further according to acquisition
Score obtains the 4th punishment score.
Preferably, barrier detection credibility evaluation method of the invention, in step (A), first classification method
It is with profile texture obstruction detection, second classification method is with deep learning obstruction detection.
Preferably, barrier detection credibility evaluation method of the invention, in step (A), first classification method
It is the present image profile texture to be obtained with gradient orientation histogram and logarithm weighting pattern, and detect and hinder with support vector machines
Hinder object, second classification method is the convolutional neural networks obstruction detection with deep learning.
The beneficial effects of the present invention are: by the processing unit determine the first current barrier detection result with
The second current barrier detection result whether there is at least one different barrier, and to determine whether to obtain this, at least one is punished
Point penalty number, and after obtaining at least one punishment score, the confidence score is subtracted into at least one punishment score, to update
The confidence score.
Detailed description of the invention
Other features of the invention and effect will be clearly presented in the embodiment referring to schema, in which:
It is real for implementing the one of barrier detection credibility evaluation method of the present invention to be illustratively painted one for mono- block diagram of Fig. 1
The system for applying example;
Fig. 2 is a flow chart, illustrates the embodiment of barrier detection credibility evaluation method of the present invention;
Fig. 3 is a schematic diagram, illustrates one first current barrier detection result;
Fig. 4 is a schematic diagram, illustrates one second current barrier detection result;
Fig. 5 is a flow chart, and collocation Fig. 2 illustrates the sub-step of a step 23 of the embodiment;
Fig. 6 is a schematic diagram, illustrates a previous image;And
Fig. 7 is a schematic diagram, and collocation Fig. 6 illustrates one first previous barrier detection result;
Fig. 8 is a schematic diagram, and collocation Fig. 6 illustrates one second previous barrier detection result;
Fig. 9 is a schematic diagram, illustrates a present image;
Figure 10 is a schematic diagram, and collocation Fig. 9 illustrates the another first current barrier detection result;And
Figure 11 is a schematic diagram, and collocation Fig. 9 illustrates the another second current barrier detection result.
Specific embodiment
Refering to fig. 1, illustrate a system of the embodiment for implementing barrier detection credibility evaluation method of the present invention
100.The system 100 includes a storage unit 11, an image capturing unit 12 and the electrical connection storage unit 11 and the image
The processing unit 13 of shooting unit 12.The image capturing unit 12 constantly shoots and transmits sequence of images to the processing unit
13.The storage unit 11 is stored with a confidence score such as, 100 points, first look-up table, a barrier of the barrier to score
First punishment weight of one first classification method is corresponded to the second look-up table of score, one and one corresponds to one second classification
Second punishment weight of method.Wherein, which is relevant to practical obstacle object and the image capturing unit comprising multiple
The obstacle distance of 12 distance, multiple types for being relevant to barrier obstacle species and it is multiple correspond respectively to it is described
The score of obstacle distance and the obstacle species, 1 example of table go out first look-up table.The second look-up table includes multiple phases
About the Obstacle Position and multiple scores for corresponding respectively to the Obstacle Position of position of the barrier in an image,
2 example of table goes out the second look-up table.
Table 1
Table 2
Refering to fig. 1 and Fig. 2, illustrate how the system 100 to execute barrier detection credibility evaluation method of the present invention should
Embodiment.The step of following detailed description embodiment is included.
In step 21, which, should after receiving the present image from the image capturing unit 12
Processing unit 13, which is obtained using first classification method according to the present image and stores one, indicates the first classification method institute
The current barrier detection result (see Fig. 3) of the first of the barrier detected, and second classification is utilized according to the present image
Method obtains and stores a second current barrier detection result for indicating barrier that second classification method is detected
(see Fig. 4).It is worth noting that, in the present embodiment, which is with gradient orientation histogram (Histogram
Of oriented gradient, HOG) and logarithm weighting pattern (Logarithm Weighted Patterns, LWP) acquisition
The profile texture of the present image, and with support vector machines (support vector machine, SVM) obstruction detection, it should
Second classification method is with convolutional neural networks (the Convolutional Neural of deep learning (deep learning)
Networks, CNN) obstruction detection, but not limited to this.Whether first classification method or second classification method,
The two can not only detect barrier in obstruction detection, can also detect the position that barrier is located at the present image, and
Mark the type (see Fig. 3,4) of detected barrier, therefore the first current barrier detection result and this is second current
The barrier that barrier detection result all includes detected barrier, is detected is located at the position of the present image, with
And the type of the barrier detected.
In step 22, which determines the first current barrier detection result and the second current barrier
Detecting result whether there is at least one different barrier.When the processing unit 13 determines the first current barrier detection knot
Fruit, there are when at least one different barrier, carries out step 23 from the second current barrier detection result;And work as the processing
Unit 13 determine the first current barrier detection result and the second current barrier detection result there is no this at least one
When different barrier, step 24 is carried out.
In step 23, which obtains at least one according at least one different barrier and respectively corresponds this extremely
The punishment score of a few different barrier, and the confidence score is subtracted into at least one punishment score, to update the storage
The confidence score that unit 11 stores.In this example, it is assumed that there are N number of different barrier, N≤1, step 23 includes
Sub-step 231~236 (see Fig. 5), but not limited to this.
Referring again to Fig. 1 and Fig. 5, further illustration sub-step 231~236.
In sub-step 231, when initial, which determines that this first is worked as the 1st different barrier
Preceding barrier detection result whether there is the 1st different barrier, that is, i=1.
In sub-step 232, which determines the first current barrier detection result with the presence or absence of i-th not
Same barrier.When the processing unit 13 determines the first current barrier detection result, there is no i-th of different obstacles
When object, step 233 is carried out;And when the processing unit 13 determines the first current barrier detection result, there are i-th of differences
Barrier when, carry out step 234.
In sub-step 233, which obtains one according to this i-th different barrier and corresponds to this i-th not
Same barrier, and the first punishment score as one in at least one punishment score, and the confidence score is subtracted this
First punishment score, to update the confidence score of the storage unit 11 storage.In the present embodiment, which is
First according to this i-th different barrier be located at the present image position utilize one known to image distance evaluation method, estimate
Practical obstacle object corresponding to this i-th different barrier is calculated at a distance from the image capturing unit 12, then according to institute
The practical obstacle object corresponding to this i-th different barrier estimated at a distance from the image capturing unit 12, this i-th
The type of a different barrier cooperates first look-up table, obtains a score for corresponding to this i-th different barrier,
Score obtained is multiplied by the first punishment weight again to obtain the first punishment score;However, in other embodiments,
The processing unit 13 can also be according to other attributes of this i-th different barrier such as, and the cooperations such as Obstacle Position are relevant to it
Another look-up table of his barrier attribute and score, to obtain the score of the barrier different corresponding to this i-th, then will be obtained
The score obtained is multiplied by the first punishment weight to obtain the first punishment score;Even the processing unit 13 can also be not required to cooperate and be somebody's turn to do
First look-up table and this i-th different barrier is set as the default score, then the default score is multiplied by this and first is punished
Weight is penalized to obtain the first punishment score, is not limited thereto.It is noted that due to feature of the invention and not lying in
Image distance evaluation method known to this known those skilled in the art, for sake of simplicity, therefore there is omitted herein their details.
In sub-step 234, which obtains one according to this i-th different barrier and corresponds to this i-th not
Same barrier, and the second punishment score as one in at least one punishment score, and the confidence score is subtracted this
Second punishment score, to update the confidence score of the storage unit 11 storage.In the present embodiment, which is
First it is located at the position of the present image according to this i-th different barrier using the image distance evaluation method, estimates this
Practical obstacle object corresponding to i-th of different barrier is at a distance from the image capturing unit 12, then according to being estimated
This i-th different barrier corresponding to the practical obstacle object at a distance from the image capturing unit 12, this i-th it is different
Barrier type, cooperate first look-up table, obtain a score for corresponding to this i-th different barrier, then by institute
The score of acquisition is multiplied by the second punishment weight to obtain the second punishment score;However, in other embodiments, the processing
Unit 13 can also be according to other attributes of this i-th different barrier such as, and the cooperations such as Obstacle Position are relevant to other obstacles
Another look-up table of object attribute and score, to obtain the score of the barrier different corresponding to this i-th, then by obtained point
Number is multiplied by the second punishment weight to obtain the second punishment score;Even the processing unit 13 can also be not required to cooperate this first to look into
It looks for table and this i-th different barrier is set as the default score, then the default score is multiplied by the second punishment weight
Come obtain this second punishment score, be not limited thereto.
In sub-step 235 after connecting sub-step 233 and 234, which judges this i-th different barrier
It whether is the different barrier of n-th, that is to say, that judge whether i=N.When the processing unit 13 determines this i-th difference
Barrier when being the different barrier of n-th, carry out step 24;And when the processing unit 13 determine this i-th it is different
When barrier is not n-th different barrier, sub-step 236 is carried out.
In sub-step 236, the processing unit 13 barrier different for i+1, that is to say, that i is set as i+1.
Then, sub-step 232~236 is repeated until i=N.
In step 24, which manages unit 13 according to this and is indicated according to a previous image obtained one
First previous barrier detection of the barrier that first classification method is detected is as a result, with the first current barrier detection
As a result, determining that the first previous barrier detection result is different with the presence or absence of at least one from the first current barrier detection result
Barrier, when the processing unit 13 determines the first previous barrier detection result and the first current barrier detection knot
Fruit carries out step 25 there are when at least one different barrier;And when the processing unit 13 determines the first previous obstacle
When at least one different barrier is not present from the first current barrier detection result in object detecting result, step 26 is carried out.
In step 25, the processing unit 13 according at least one different barrier, the second look-up table and this first
Power of punishment recapture at least one corresponding at least one different barrier third punishment score, and the confidence score is subtracted
An at least third punishes score, to update the confidence score of the storage unit 11 storage.In the present embodiment, for every
One different barrier, the processing unit 13 are first according to the different barrier in the position of the present image or the previous image
Set and cooperate the second look-up table, obtain a score for corresponding to the different barrier, then by score obtained be multiplied by this
One punishment weight punishes score to obtain the third;However, in other embodiments, which can also be according to this not
Such as, the cooperations such as obstacle species or obstacle distance are relevant to other barrier attributes and score to other attributes of same barrier
Another look-up table, to obtain the score of the barrier different corresponding to this, then score obtained is multiplied by first punishment
Weight punishes score to obtain the third;Even the processing unit 13 can also be not required to cooperate the second look-up table and by each difference
Barrier be all set as same default score, then the default score is multiplied by the first punishment weight to obtain third punishment
Score is not limited thereto.For example, if the first previous barrier detection result indicates a barrier first and an obstacle
Object second, and the first current barrier detection result indicates the barrier first and a barrier third, then the processing unit 13 is sentenced
There are the barrier second and the barriers with the first current barrier detection result for the fixed first previous barrier detection result
Third two different barriers.
In step 26, which manages unit 13 according to this and is indicated according to the previous image obtained one
Second previous barrier detection of the barrier that second classification method is detected is as a result, with the second current barrier detection
As a result, determining that the second previous barrier detection result is different with the presence or absence of at least one from the second current barrier detection result
Barrier, when the processing unit 13 determines the second previous barrier detection result and the second current barrier detection knot
Fruit carries out step 27 there are when at least one different barrier;And when the processing unit 13 determines the second previous obstacle
When at least one different barrier is not present from the second current barrier detection result in object detecting result, step 28 is carried out.
In step 27, the processing unit 13 according at least one different barrier, the second look-up table and this second
Power of punishment recapture at least one corresponding at least one different barrier the 4th punishment score, and the confidence score is subtracted
At least one the 4th punishment score, to update the confidence score of the storage unit 11 storage.In the present embodiment, for every
One different barrier, the processing unit 13 are first according to the different barrier in the position of the present image or the previous image
Set and cooperate the second look-up table, obtain a score for corresponding to the different barrier, then by score obtained be multiplied by this
Two punishment weights punish score to obtain the 4th;However, in other embodiments, which can also be according to this not
Such as, the cooperations such as obstacle species or obstacle distance are relevant to other barrier attributes and score to other attributes of same barrier
Another look-up table, to obtain the score of the barrier different corresponding to this, then score obtained is multiplied by second punishment
Weight punishes score to obtain the 4th;Even the processing unit 13 can also be not required to cooperate the second look-up table and by each difference
Barrier be all set as same default score, then the default score is multiplied by the second punishment weight to obtain the 4th punishment
Score is not limited thereto.It is important to note that in the present embodiment, step 22,23 are before step 24~27, at it
In his embodiment, step 24~27 can be before step 22,23, that is to say, that when the judgement result of step 26 is negates, hold
Row step 22 executes step 28 when the judgement result of step 22 is negates.
In a step 28, which exports the confidence score, and is then reset to the confidence score
100.It is noted that in the present embodiment, if the confidence score is negative, which can be corrected to
It is exported again after 0, but not limited to this, which can also directly export the confidence score with negative sign.
Due to the third punish score and the 4th punishment score be relevant to the first previous barrier detection result with
The different barrier of the first current barrier detection result and the second previous barrier detection result and this is second current
The different barriers of barrier detection result, third punishment score and the 4th punishment score are relevant to different barriers
In the position of the previous image or the present image, to discriminate whether to pass in and out the figure in the different time for the different barrier
As there is a situation where serious mistakes for the detecting of coverage or the system 100 of shooting unit 12.And due to first classification
Method correspond to this first punishment score and the third punish score, second classification method correspond to this second punishment score and
4th punishment score, therefore the first punishment score and third punishment score correspond to the first punishment weight, second punishment
Score and the 4th punishment score correspond to the second punishment weight, which is, for example, 0.2, the second punishment weight
For example, 0.8.
Refering to fig. 1, Fig. 6 and Fig. 7, below in conjunction with an exemplary applications, to illustrate barrier detection confidence level of the present invention
The embodiment of appraisal procedure, Fig. 6 example go out the previous image, and it includes a barrier A, a barrier B and a barrier C, figures
7 examples go out the processing unit 13 and utilize the previous barrier detection of the first classification method obtained first according to the previous image
As a result, the first previous barrier detection result indicates that barrier A is located at the previous image and the centre of the present image,
Type is vehicle, and the edge of the previous image is located at apart from 120 meters of the image capturing unit and barrier B, and type is vehicle, away from
12 30 meters from the image capturing unit, Fig. 8 example goes out the processing unit 13 and utilizes the second classification side according to the previous image
The previous barrier detection of method obtained second is as a result, the second previous barrier detection result indicates that barrier A is located at
The centre of the previous image and the present image, type are vehicle, are located at apart from 120 meters of the image capturing unit, barrier B
The edge of the previous image, type are vehicle, are located at the previous figure apart from 12 30 meters of the image capturing unit and barrier C
The inclined edge in centre of picture and the present image, type is behaved, apart from 12 30 meters of the image capturing unit.Fig. 9 example goes out to deserve
Preceding image goes out the processing unit 13 it includes barrier A and barrier C, Figure 10 example and is somebody's turn to do according to present image utilization
The current barrier detection of first classification method obtained first is as a result, the first current barrier detection result indicates the barrier
Object A is hindered to be located at the present image and the centre of the present image, type is vehicle, apart from 120 meters of the image capturing unit, Figure 11
Example goes out the processing unit 13 and utilizes the current barrier detection of the second classification method obtained second according to the present image
As a result, the second current barrier detection result indicates that barrier A is located at the centre of the present image, type is vehicle, away from
120 meters from the image capturing unit and barrier C inclined edges in centre for being located at the present image, type are behaved, and distance should
12 30 meters of image capturing unit.
As shown in step 233, which shows the difference according to the first current barrier detection result meaning
Barrier (the barrier C) obstacle species and obstacle distance and first look-up table, obtain corresponding barrier C
Score (since obstacle species are behaved and obstacle distance is 30 meters, therefore the score is 70), then by score obtained
(that is, 70) be multiplied by this first punishment weight (that is, 0.2) come obtain this first punishment score (that is,
0.2 × 70=14), it is 100-14=86 which, which updates the confidence score,.As shown in step 25, the processing list
Member 13 obtains corresponding be somebody's turn to do in the position of the previous image and the second look-up table according to the different barrier (the barrier B)
(since barrier B is located at edge, therefore the score is 0), then by score obtained (that is, 0) to the score of barrier B
The first punishment weight is multiplied by obtain third punishment score (that is, 0.2 × 0=0), which updates should
Confidence score is 86-0=86.As indicated at step 27, the processing unit 13 is according to the different barrier (the barrier B)
The score of corresponding barrier B is obtained (since barrier B is located at side in the position of the previous image and the second look-up table
Edge, thus the score be 0), then by score obtained (that is, 0) be multiplied by this second punishment weight (that is, 0.8)
Score (that is, 0.8 × 0=0) is punished to obtain the 4th, and it is 86- which, which simultaneously updates the confidence score,
0=86.Finally in a step 28, the confidence score of the processing unit 13 output is 86.
In conclusion barrier detection credibility evaluation method of the present invention, determine this before first by the processing unit 13
One barrier detection result and the first current barrier detection result with the presence or absence of at least one different barrier, this is before second
One barrier detection result and the second current barrier detection result with the presence or absence of at least one different barrier and this first
Current barrier detection result whether there is at least one different barrier from the second current barrier detection result, to determine
Whether acquisition this at least one first punishment score, this at least one second punishment score, an at least third punishment score and this extremely
Few one the 4th punishment score, and after obtaining punishment score, which is subtracted into punishment score, to update and export this
Confidence score provides the confidence level of barrier detection result to steered reference, therefore can reach mesh of the invention really whereby
's.
As described above, only the embodiment of the present invention is when cannot be limited the scope of implementation of the present invention with this, i.e., all
According to simple equivalent changes and modifications made by claims of the present invention and description, all still belong to the scope of the present invention.
Claims (10)
1. a kind of barrier detection credibility evaluation method, by the processing unit of electrical connection storage unit and image capturing unit Lai
Implement, the storage unit is stored with confidence score, and described image shooting unit constantly shoots and transmits sequence of images
To the processing unit, it is characterised in that: the barrier detection credibility evaluation method comprises the steps of:
(A) after receiving the present image from described image shooting unit, the first classification is utilized according to the present image
Method obtain and store the barrier for indicating that first classification method is detected the first current barrier detection as a result,
And it is obtained and is stored using the second classification method according to the present image and indicate what second classification method was detected
The current barrier detection result of the second of barrier;
(B) determine the first current barrier detection result and the described second current barrier detection result with the presence or absence of at least
One different barrier;And
(C) when determining the described first current barrier detection result and the described second current barrier detection result there are this extremely
When a few different barrier, according at least at least one different barrier, obtains at least one and do not respectively correspond this at least one not
The punishment score of same barrier, and the confidence score is subtracted into at least one punishment score, to update the confidence
Score.
2. barrier detection credibility evaluation method according to claim 1, it is characterised in that: step (C) includes following
Sub-step:
(C-1) when determine that the described first current barrier detection result exists with the described second current barrier detection result should
When at least one different barrier, for each of at least one different barrier, the first current barrier is determined
Whether detecting result is not present the different barrier;
(C-2) when determining the described first current barrier detection result there is no when the different barrier, according at least to
The different barrier obtains the first punishment score as one in at least one punishment score, and the confidence is divided
Number subtracts the first punishment score, to update the confidence score;And
(C-3) when determining the described first current barrier detection result there are when the different barrier, according at least to institute
It states different barrier and obtains and punish score as second of one in at least one punishment score, and by the confidence score
The second punishment score is subtracted, to update the confidence score.
3. barrier detection credibility evaluation method according to claim 2, it is characterised in that: the storage unit is also deposited
First look-up table of the barrier to score, the first punishment weight corresponding to first classification method are contained, and corresponds to institute
The second punishment weight of the second classification method is stated, first look-up table is relevant to practical obstacle object and described image comprising multiple
The obstacle distance of the distance of shooting unit and multiple scores for corresponding respectively to the obstacle distance, in which:
In step (A), the first current barrier detection result all includes with the described second current barrier detection result
The barrier detected, and the barrier that is detected are located at the position of the present image;
In step (C-2), the processing unit is the position for being first located at the present image according to the different barrier,
Practical obstacle object corresponding to the different barrier is estimated at a distance from described image shooting unit, and according to being estimated
Distance and first look-up table out obtain the score for corresponding to the different barrier, then score obtained are multiplied
The upper first punishment weight is to obtain the first punishment score;And
In step (C-3), the processing unit is the position for being first located at the present image according to the different barrier,
Practical obstacle object corresponding to the different barrier is estimated at a distance from described image shooting unit, and according to being estimated
Distance and first look-up table out obtain the score for corresponding to the different barrier, then score obtained are multiplied
The upper second punishment weight is to obtain the second punishment score.
4. barrier detection credibility evaluation method according to claim 2, it is characterised in that: the storage unit is also deposited
First look-up table of the barrier to score, the first punishment weight corresponding to first classification method are contained, and corresponds to institute
The second punishment weight of the second classification method is stated, first look-up table includes multiple obstacle species and multiple respectively corresponds
In the score of the type, in which:
In step (A), the first current barrier detection result all includes with the described second current barrier detection result
The barrier detected, and the type of barrier detected;
In step (C-2), the processing unit is the type and first look-up table according to the different barrier, is obtained
Score corresponding to the different barrier, then that score obtained is multiplied by the first punishment weight is described to obtain
First punishment score;And in step (C-3), the processing unit is the type and described according to the different barrier
One look-up table obtains the score for corresponding to the different barrier, then score obtained is multiplied by second power of punishment
Weight is to obtain the second punishment score.
5. barrier detection credibility evaluation method according to claim 1, it is characterised in that: after step (A), also wrap
Containing following steps:
(D) according to previous image the first previous barrier obtained for indicating barrier that first classification method is detected
Hinder object detecting result with the described first current barrier detection as a result, determine the first previous barrier detection result with it is described
First current barrier detection result whether there is at least one different barrier;
(E) when determining the described first previous barrier detection result and the described first current barrier detection result there are this extremely
When a few different barrier, at least one corresponding at least one different barrier is obtained according at least at least one different barrier
Hinder the third of object to punish score, and the confidence score is subtracted at least third punishment score, to update the confidence
Spend score.
6. barrier detection credibility evaluation method according to claim 5, it is characterised in that: the storage unit is also deposited
Barrier is contained to the second look-up table of score, the second look-up table includes multiple positions for being relevant to barrier and being located at image
Obstacle Position and multiple scores for corresponding respectively to the Obstacle Position, in which:
In step (A), the first current barrier detection result all includes with the described second current barrier detection result
The barrier detected, and the barrier that is detected are located at the position of the present image;And
In step (E), the processing unit is to be located at the present image or described previous according to the different barrier
The position of image and the second look-up table obtain the score for corresponding to the different barrier, further according to the score of acquisition
Obtain the third punishment score.
7. barrier detection credibility evaluation method according to claim 1, it is characterised in that: after step (A), also wrap
Containing following steps:
(F) according to previous image the second previous barrier obtained for indicating barrier that second classification method is detected
Hinder object detecting result with the described second current barrier detection as a result, determine the second previous barrier detection result with it is described
Second current barrier detection result whether there is at least one different barrier;
(G) when determining the described second previous barrier detection result and the described second current barrier detection result there are this extremely
When a few different barrier, at least one corresponding at least one different barrier is obtained according at least at least one different barrier
Hinder the 4th of object to punish score, and the confidence score is subtracted at least one the 4th punishment score, to update the confidence
Spend score.
8. barrier detection credibility evaluation method according to claim 7, it is characterised in that: the storage unit is also deposited
Barrier is contained to the second look-up table of score, the second look-up table includes multiple positions for being relevant to barrier and being located at image
Obstacle Position and multiple scores for corresponding respectively to the Obstacle Position, in which:
In step (A), the first current barrier detection result all includes with the described second current barrier detection result
The barrier detected, and the barrier that is detected are located at the position of the present image;And
In step (G), the processing unit is to be located at the present image or described previous according to the different barrier
The position of image and the second look-up table obtain the score for corresponding to the different barrier, further according to the score of acquisition
Obtain the 4th punishment score.
9. barrier detection credibility evaluation method according to claim 1, it is characterised in that: described in step (A)
First classification method is with profile texture obstruction detection, and second classification method is with deep learning obstruction detection.
10. barrier detection credibility evaluation method according to claim 9, it is characterised in that: in step (A), institute
It states the first classification method and is and the present image profile texture is obtained with gradient orientation histogram and logarithm weighting pattern, and with branch
Vector machine obstruction detection is held, second classification method is the convolutional neural networks obstruction detection with deep learning.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711431445.5A CN109960990B (en) | 2017-12-26 | 2017-12-26 | Method for evaluating reliability of obstacle detection |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711431445.5A CN109960990B (en) | 2017-12-26 | 2017-12-26 | Method for evaluating reliability of obstacle detection |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109960990A true CN109960990A (en) | 2019-07-02 |
CN109960990B CN109960990B (en) | 2022-06-03 |
Family
ID=67021870
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201711431445.5A Active CN109960990B (en) | 2017-12-26 | 2017-12-26 | Method for evaluating reliability of obstacle detection |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109960990B (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110568851A (en) * | 2019-09-30 | 2019-12-13 | 重庆元韩汽车技术设计研究院有限公司 | Automobile chassis motion control system and method based on remote control |
CN113156364A (en) * | 2020-01-22 | 2021-07-23 | 西克股份公司 | Security system and method |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2005121496A (en) * | 2003-10-17 | 2005-05-12 | Toyota Motor Corp | Obstruction detecting radar installation of vehicle |
CN103043000A (en) * | 2011-10-13 | 2013-04-17 | 能晶科技股份有限公司 | Obstacle detection system and obstacle detection method thereof |
CN105095848A (en) * | 2014-12-01 | 2015-11-25 | 北京云识图信息技术有限公司 | Object identification method and system |
CN105372654A (en) * | 2015-12-14 | 2016-03-02 | 财团法人车辆研究测试中心 | Obstacle classification reliability quantification method |
CN107240299A (en) * | 2017-07-20 | 2017-10-10 | 北京纵目安驰智能科技有限公司 | Autonomous land vehicle is to mobile object identification and the method for vehicle obstacle-avoidance |
-
2017
- 2017-12-26 CN CN201711431445.5A patent/CN109960990B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2005121496A (en) * | 2003-10-17 | 2005-05-12 | Toyota Motor Corp | Obstruction detecting radar installation of vehicle |
CN103043000A (en) * | 2011-10-13 | 2013-04-17 | 能晶科技股份有限公司 | Obstacle detection system and obstacle detection method thereof |
CN105095848A (en) * | 2014-12-01 | 2015-11-25 | 北京云识图信息技术有限公司 | Object identification method and system |
CN105372654A (en) * | 2015-12-14 | 2016-03-02 | 财团法人车辆研究测试中心 | Obstacle classification reliability quantification method |
CN107240299A (en) * | 2017-07-20 | 2017-10-10 | 北京纵目安驰智能科技有限公司 | Autonomous land vehicle is to mobile object identification and the method for vehicle obstacle-avoidance |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110568851A (en) * | 2019-09-30 | 2019-12-13 | 重庆元韩汽车技术设计研究院有限公司 | Automobile chassis motion control system and method based on remote control |
CN113156364A (en) * | 2020-01-22 | 2021-07-23 | 西克股份公司 | Security system and method |
Also Published As
Publication number | Publication date |
---|---|
CN109960990B (en) | 2022-06-03 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US11854272B2 (en) | Hazard detection from a camera in a scene with moving shadows | |
CN110858295B (en) | Traffic police gesture recognition method and device, vehicle control unit and storage medium | |
US9141870B2 (en) | Three-dimensional object detection device and three-dimensional object detection method | |
CN111611853B (en) | Sensing information fusion method, device and storage medium | |
CN110501018A (en) | A kind of traffic mark board information collecting method for serving high-precision map producing | |
CN109284674A (en) | A kind of method and device of determining lane line | |
CN107273816B (en) | Traffic speed limit label detection recognition methods based on vehicle-mounted forward sight monocular camera | |
CN108776974B (en) | A kind of real-time modeling method method suitable for public transport scene | |
CN110386065A (en) | Monitoring method, device, computer equipment and the storage medium of vehicle blind zone | |
CN105608417A (en) | Traffic signal lamp detection method and device | |
EP3979196A1 (en) | Image processing method and apparatus for target detection | |
CN103593679A (en) | Visual human-hand tracking method based on online machine learning | |
CN102436578B (en) | Formation method for dog face characteristic detector as well as dog face detection method and device | |
CN109800654A (en) | Vehicle-mounted camera detection processing method, apparatus and vehicle | |
CN109960990A (en) | Barrier detection credibility evaluation method | |
CN111160132B (en) | Method and device for determining lane where obstacle is located, electronic equipment and storage medium | |
CN111738085B (en) | System construction method and device for realizing automatic driving simultaneous positioning and mapping | |
Takahashi et al. | A robust lane detection using real-time voting processor | |
Shafie et al. | Smart video surveillance system for vehicle detection and traffic flow control | |
CN106250892A (en) | Method of Automatic Recognition for Character of Lcecse Plate | |
CN109961637A (en) | Vehicle detection apparatus and system based on more subgraphs fusion and significance analysis | |
CN110334703B (en) | Ship detection and identification method in day and night image | |
CN116363628A (en) | Mark detection method and device, nonvolatile storage medium and computer equipment | |
Mitsudome et al. | Autonomous mobile robot searching for persons with specific clothing on urban walkway | |
CN109690555A (en) | Face detector based on curvature |
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 |