CN108985271B - Method for judging pedestrian of interest of magnetic model - Google Patents
Method for judging pedestrian of interest of magnetic model Download PDFInfo
- Publication number
- CN108985271B CN108985271B CN201810936629.5A CN201810936629A CN108985271B CN 108985271 B CN108985271 B CN 108985271B CN 201810936629 A CN201810936629 A CN 201810936629A CN 108985271 B CN108985271 B CN 108985271B
- Authority
- CN
- China
- Prior art keywords
- pedestrians
- pedestrian
- model
- magnetic force
- probability
- 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.)
- Expired - Fee Related
Links
- 230000005291 magnetic effect Effects 0.000 title claims abstract description 127
- 238000000034 method Methods 0.000 title claims abstract description 33
- 238000004458 analytical method Methods 0.000 claims description 5
- 238000012545 processing Methods 0.000 abstract description 7
- 238000013459 approach Methods 0.000 description 6
- 238000004364 calculation method Methods 0.000 description 6
- 238000010586 diagram Methods 0.000 description 6
- 238000001514 detection method Methods 0.000 description 5
- 238000005516 engineering process Methods 0.000 description 5
- 230000005389 magnetism Effects 0.000 description 4
- 238000004088 simulation Methods 0.000 description 4
- 230000000007 visual effect Effects 0.000 description 4
- 230000000694 effects Effects 0.000 description 3
- 241001127637 Plantago Species 0.000 description 2
- 206010039203 Road traffic accident Diseases 0.000 description 2
- 238000006073 displacement reaction Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004422 calculation algorithm Methods 0.000 description 1
- 230000005494 condensation Effects 0.000 description 1
- 238000009833 condensation Methods 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/20—Movements or behaviour, e.g. gesture recognition
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Health & Medical Sciences (AREA)
- Psychiatry (AREA)
- Social Psychology (AREA)
- Human Computer Interaction (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Traffic Control Systems (AREA)
Abstract
The invention discloses a method for judging pedestrians interested by a magnetic force model, belongs to the field of moving target tracking processing, and aims to solve the problem of classifying the pedestrians.
Description
Technical Field
The invention belongs to the field of moving target tracking processing, and particularly relates to a classification method for distinguishing potential danger degrees of pedestrians on a road by using a magnetic model.
Background
The moving target tracking processing technology is an important research subject in the field of machine vision, and with the application of autonomous automobiles and auxiliary driving systems, how to reasonably apply the target tracking processing technology to protect the safety of pedestrians and vehicles is a popular research direction at present.
At present, under the condition of only using a vehicle-mounted camera, classifying pedestrians by analyzing information such as historical moving tracks and moving speeds of pedestrians is a main approach for protecting the safety of pedestrians and vehicles by utilizing a target tracking processing technology. Firstly, the moving track and the moving speed of the pedestrian are analyzed to calculate the probability that the pedestrian collides with the vehicle, and then the pedestrian is classified into different types by utilizing the collision probability and establishing corresponding classification rules. Most of the existing road pedestrian classification methods in the target tracking processing technology calculate the probability that a pedestrian will directly collide with a vehicle, or only detect special pedestrians with specific characteristics.
The patent application number is CN201610048194.1, and the name is 'blind person detection and identification method and system based on combined features and vehicle-mounted camera', firstly, the HOG features are utilized to detect pedestrians, and then three classification features, namely a blind guiding dog, a blind stick and a blind mirror, are used near the detection result of the pedestrians to detect whether blind persons exist.
The patent application number is CN201610048233.8, entitled traffic police detection method and system based on coat-hat characteristic posture detection, and uses the police hat and fluorescent waistcoat of the traffic police in the road as classification characteristics to judge whether the pedestrian detection result has the traffic police. Although a specific target can be well identified by detecting a specific feature of the target, the method has a great limitation because only the specific target can be detected.
In the article "Analysis of Pedestrian condensation using Fuzzy Inference Model", Hariyono et al analyze the moving speed and moving direction of pedestrians and vehicles and the relative distance between the pedestrians and vehicles, calculate the Collision probability of the pedestrians and vehicles, and classify the pedestrians according to the size of the Collision probability. However, in order to accurately acquire speed and position information of pedestrians, the method uses a camera erected on the roadside to shoot the road condition in front of the vehicle from the side, so that the method is greatly influenced by the environment.
Hariyono et al in the article "Estimation of fusion Risk for Improving Driver's Safety" establishes a dangerous area in front of a vehicle, and judges pedestrians entering the dangerous area as dangerous pedestrians so as to classify the pedestrians. However, the existence of many sudden conditions on the road may cause the pedestrian to be instantly changed from a normal pedestrian to a dangerous pedestrian. For example, a pedestrian holding a mobile phone suddenly decelerates while passing through a road, causing the pedestrian who is in a safe state to suddenly become a pedestrian who may collide with the vehicle.
The road pedestrian classification method used by the invention is to establish a magnetic model, and judge the potential danger of the pedestrian by analyzing the physical logic relationship between the pedestrians, rather than estimating the danger of direct collision between the pedestrian and the vehicle. The pedestrians with high potential danger are used as main analysis objects by establishing a new road pedestrian classification rule, so that the classification results of the existing road pedestrian classification method are enriched, and the safety of vehicles and pedestrians is better protected.
Disclosure of Invention
In order to solve the classification problem of the pedestrians, the invention provides the following scheme: a method for judging pedestrians interested in a magnetic force model includes the steps of shooting pedestrians by a vehicle-mounted camera, measuring moving speed, moving direction, distance from the camera and relative positions of the pedestrians, carrying out corresponding magnetic force model analysis on measurement information, calculating the degree of coincidence of the pedestrians and classification conditions set by the magnetic force model, and classifying the pedestrians into normal pedestrians and magnetic force pedestrians.
Has the advantages that: by the scheme, the pedestrian classification method and the pedestrian classification system realize the effect of classifying pedestrians, and the classification is more in line with the actual road condition.
Drawings
FIG. 1 is a logic diagram of a magnetic force model based road pedestrian classification method;
fig. 2 is a diagram illustrating classification results of repulsive pedestrians;
FIG. 3 is a diagram illustrating classification results of suction pedestrians;
FIG. 4 is a diagram illustrating classification results of a non-magnetic pedestrian;
FIG. 5 is a diagram illustrating classification results of various magnetic pedestrians;
Detailed Description
The invention is further described with reference to the accompanying drawings and the specific classification procedures:
a method for judging pedestrians of interest of a magnetic force model is shown in a logic block diagram of fig. 1, and the method is implemented by the following steps:
step 1: inputting a moving speed v of the pedestrian, a relative distance s between the pedestrians in the image, and a relative distance z between the pedestrian and the camera;
step 2: calculating the magnetic force probability P of each pedestrian in the imageMM；
And 3, step 3: according to PMMAnd classifying the pedestrians according to the three magnetic force probabilities and outputting classification results.
To describe the above scheme in more detail and distinguish the existing direct collision probability calculation of pedestrians and vehicles, the present invention provides a more complete description of the above scheme, that is, the present invention estimates the potential danger of pedestrians by establishing a magnetic model, and is a road pedestrian classification method based on the magnetic model, which can be implemented by software, and relates to a system executing the software, which stores a plurality of instructions, which are suitable for being loaded by a processor and executing the method or algorithm.
Under the condition of only using the vehicle-mounted camera, if the information of the moving speed, the moving direction, the distance from the camera, the relative position between pedestrians and the like of the pedestrians is known, the magnetic force model can be combined for analysis, the conformity degree of the classification conditions set by the pedestrians and the magnetic force model is calculated, the pedestrians are classified into common pedestrians and magnetic force pedestrians, the classification result can further enrich the information of the pedestrians on the road obtained by the existing target tracking processing technology, and more sufficient road condition reference information is provided for the autonomous automobile and the auxiliary driving system.
The above object is achieved by the following technical solutions.
In the first step, a moving speed v of the pedestrian, a relative distance s between the pedestrians in the image, and a relative distance z between the pedestrian and the camera are input.
And secondly, establishing a magnetic model. The magnetic force model is divided into three submodels of a repulsion force model, a suction force model and a non-magnetic force model, the three submodels are independent of each other and do not influence each other, and respectively correspond to a magnetic force probability: pr(probability of repulsion), Ps(suction probability), Pm(probability of no magnetic force). The total magnetic probability P of the pedestrian is the collection of the three magnetic probabilitiesMMAnd is and
PMM＝{Pr,Ps,Pm} (1)
let the probability of magnetism δ be 0.7, δ may be a constant coefficient. When P of a certain pedestrianr、PsOr PmWhen any one of the magnetic force probabilities exceeds delta (for the non-magnetic force probability P)mI.e. Pm1), the pedestrian is determined to be a magnetic pedestrian. The magnetic force model in the invention sets the pedestrian to move from the left side to the right side of the image as positive direction and from the right side to the left side as negative direction. To the value of delta, take the repulsion pedestrian as an example, when two pedestrians that go in opposite directions and the travel route can collide are relatively close to each other, these two pedestrians can be judged as the repulsion pedestrian, this application is according to human social space distanceThe general distance (more than 360cm from others) in the social space of the human body is used as a judgment basis and is converted into a probability threshold value of about 0.7, and similarly, when two rows are in the same direction and the distance is close, the general distance (more than 360cm from others) in the social space of the human body can also be used as a judgment basis and is converted into a probability threshold value of about 0.7. Therefore, when the pedestrians are relatively close to each other, the magnetic pedestrian probability determination threshold is used, and of course, the social space distance may be adaptively adjusted by using the current scheme as a reference in the implementation of the scheme, so that the general distance is adaptively modified, or the probability threshold may be set by itself.
(1) Definition of repulsion model: in the pedestrians which run in opposite directions and collide in a plurality of traveling routes, the moving speed of the pedestrians can be reduced or even stopped when the pedestrians meet each other until the original speed is recovered after meeting is finished, the characteristic that the moving directions of the pedestrians are opposite is similar to the same-polarity magnetic repulsion effect, and the situation is called as a repulsion model.
The number of pedestrians in the repulsion model is at least two, and pedestrians which run in opposite directions and collide with each other in the traveling route form the repulsion model. The determination of the repulsive force model takes the distance between pedestrians and the distance between a pedestrian and a vehicle, which are traveling in opposite directions, as the main basis: firstly, comparing the distances between two pedestrians walking in opposite directions and a vehicle to judge whether the traveling routes of the two pedestrians collide; if the traveling routes of the pedestrians walking in the opposite directions collide, the distance between the two pedestrians walking in the opposite directions is compared to judge whether the two pedestrians are about to collide. The formula (2) is a repulsive force probability calculation formula,
in the formula PriIs the probability of repulsion of the pedestrian i. sijIs the distance between pedestrians i and j. dijIs the difference between the distances of the pedestrians i and j and the camera respectively and dij＝|zi-zjL, wherein ziIs the distance between the pedestrian i and the camera, zjIs the distance between the pedestrian j and the camera. m and n are constant coefficients, and m, n ∈ (0, 1). v. ofiAnd vjThe moving speeds, v, of pedestrians i and j, respectivelyivj< 0 indicates that the pedestrians i and j are traveling in opposite directions. The repulsion probability only exists among pedestrians in opposite directions, and when the pedestrians in opposite directions approach each other, the repulsion probability is gradually increased; after the opposite pedestrians are staggered, the probability of repulsion is gradually reduced. In the formula (2), m and n are constant coefficients, and m, n belongs to (0,1), in an embodiment of the present invention, m is 0.4, and n is 0.4.
For pedestrian encounters on the road, pedestrians i and j move at a speed of 1.5m/s in a 640 x 480 image, where pedestrian i is (30, 90) and remains moving positively and pedestrian j is (600, 100) and remains moving negatively. When pedestrians i and j approach each other, the probability of repulsion PrAnd they are gradually increased, and their repulsive probability exceeds δ when they approach to a certain distance, and thus they are determined as magnetic pedestrians.
On the road, the pedestrian meets the condition very commonly, and the possibility that the pedestrian collides and then accidents occur may occur due to the difference of the conditions of the pedestrian, for example, the pedestrian who holds a mobile phone or hurried and hurried to the road exists in the meeting pedestrian, the former easily collides with other people due to the distraction, and the latter easily collides with other people due to the faster moving speed. The encountered ones of the pedestrians are thus distinguished by the repulsive force model.
(2) Defining a suction model: among a plurality of pedestrians walking in the same direction, two or more pedestrians walk in an overlapped or adjacent state, and the characteristics that the moving speed is similar in direction and magnitude are similar to the magnetic opposite attraction effect, and the condition is called as an attraction model.
The number of pedestrians in the suction model is at least two, and all the pedestrians in the same direction which keep overlapping or walking at an adjacent distance form the suction model. Whether the pedestrian is walking with the accompanying is the main basis for judging the suction model, and whether the pedestrian is walking with the accompanying is judged by comparing the distance between the pedestrians in the same direction and the distance between the pedestrian and the vehicle. The formula (3) is a suction probability calculation formula,
in the formula
Probability of attraction for pedestrian i, sijDistance between pedestrians i and j, dijIs the difference between the distances of the pedestrians i and j and the camera and dij＝|zi-zjL, wherein ziIs the distance between the pedestrian i and the camera, zjIs the distance between the pedestrian j and the camera. v. ofiAnd vjThe moving speeds, v, of pedestrians i and j, respectivelyivj> 0 indicates that pedestrians i and j are traveling in the same direction. k is a radical ofijIs a parameter of suction, and kij＝-[min(T|vi|,T|vj|)/sij]Where T is the elapsed time, min (T | v)i|,T|vj|) represents a smaller displacement value of the pedestrians i and j after the time T. The longer the time for maintaining a close and fixed relative distance between pedestrians traveling in the same direction, the greater the probability of attraction of the pedestrians. In the formula (3), m, n and l are constant coefficients, and m, n, l belongs to (0,1), in an embodiment of the present invention, m is 0.4, n is 0.4, and l is 0.2.
For pedestrian companions on the road, pedestrians i and j move in a 640 x 480 image at a speed of 1.5m/s, where pedestrian i is located at (30, 90) and pedestrian j is located at (45, 100), and they both maintain forward movement. When the pedestrians i and j are in the process of advancing and the relative distance between the pedestrians i and j does not change greatly, the probability P of suction force of the pedestrians i and jsWill gradually increase. When the two move forward for a certain distance, the probability of attraction of the two exceeds delta, and the pedestrian is judged to be a magnetic pedestrian.
On the road, the condition that pedestrians walk together is common, and the walking together can shield the sight line, so that the attention of the pedestrians is weakened, and the danger of the pedestrians is improved. Meanwhile, the old and the children often accompany people to walk on the street, wherein the traffic accidents on the road are easily caused by the children, and the old is one of people who need to be cared and given. The accompanying one of the pedestrians is thus distinguished by the suction model.
(3) Definition of a non-magnetic force model: in a plurality of pedestrians, the moving speed of a certain pedestrian is obviously different from that of other pedestrians but does not necessarily satisfy a repulsion model or a suction model, and the situation that the certain pedestrian does not necessarily satisfy the repulsion or the suction is similar to the non-magnetic effect, and the situation is called a non-magnetic model.
At least three pedestrians exist in the non-magnetic force model, and when the speed of one pedestrian is higher or lower than that of at least two other pedestrians, the non-magnetic force model is formed. The determination of the non-magnetic force model is mainly based on the moving speed value of the pedestrian. The formula (4) is a nonmagnetic probability calculation formula,
in the formula
Is the nonmagnetic probability of the pedestrian i. v. ofi、vjAnd vhThe moving speeds of the pedestrians i, j, and h, respectively. Only when the speed of the pedestrian i is higher or lower than that of the other at least two pedestrians, and the speeds of the three pedestrians satisfy the formula
(vi 2-vj 2)+(vi 2-vh 2)＞ε2 (5)
For a particular group of pedestrians on the road, their movement speed is generally slow, and in a 640 x 480 image, pedestrian i moves at 0.5m/s, and pedestrians j and h move at 1.5 m/s. Wherein pedestrian i is located at (30, 90), pedestrian j is located at (170, 240), pedestrian h is located at (310, 420), and they both remain moving in the forward direction. The non-magnetic probability P of the pedestrian i when the speeds of the three pedestrians remain unchangedmIt exceeds the set delta and is determined to be a magnetic pedestrian. For pedestrians moving faster on the road, in one 640 × 480 image, the pedestrian i moves at a speed of 2m/s, and the pedestrians j and h move at a speed of 1.5 m/s. Wherein the pedestrian i is located at (310, 420), the pedestrian j is located at (170, 240), the pedestrian h is located at (30, 90), and they both keep moving forward. The non-magnetic probability P of the pedestrian i when the speeds of the three pedestrians remain unchangedmDelta is exceeded and it is determined as a magnetic pedestrian.
On roads, special groups and pedestrians overtaking roads are also common, wherein the special groups are one of the pedestrians needing attention and care, the special groups tend to move at a slower speed, and the probability of causing traffic accidents due to the faster moving speed of the pedestrians is much higher than that of the pedestrians walking normally. The two pedestrians are therefore distinguished by a no-magnetic model.
The magnetic force probability of the same pedestrian may be different from that of different pedestrians through calculation, and the larger magnetic force probability is taken as the standard.
And thirdly, specifically classifying the pedestrians. After the calculation of the magnetic force model, each pedestrian can have respective magnetic force probability, and the pedestrians are divided into four types according to the difference of the magnetic force probability: the pedestrian comprises a common pedestrian, a repulsion pedestrian, a suction pedestrian and a non-magnetic pedestrian, wherein the repulsion pedestrian, the suction pedestrian and the non-magnetic pedestrian are three types of the magnetic pedestrian. Setting as PrThe pedestrian is a repulsive pedestrian when the delta is exceeded; when PsWhen delta is exceeded, the pedestrian is a suction pedestrian; (when P)mWhen delta is exceeded, the pedestrian is a non-magnetic pedestrian; four kinds of magnetic force probability PmOf highest priority, PsIs centered on the priority of PrIs lowest. The probability of magnetism of the known pedestrian i is
When the probability of no magnetic force exceeds delta, the pedestrian i is a common pedestrian; when only one magnetic force probability exceeds δ, the pedestrian i is determined as a corresponding kind of magnetic force pedestrian according to the above setting; when there is aWhen two or three kinds of the magnetic probabilities exceed δ, the pedestrian i is determined as a magnetic pedestrian with a higher priority according to the priority of the magnetic probabilities.
By the technical scheme, the road pedestrian classification method based on the magnetic force model has the beneficial effects that:
the existing road pedestrian classification method mainly analyzes and calculates the direct collision probability between pedestrians and vehicles, ignores the complexity and variability of the actual road, cannot well analyze the road condition, and is greatly influenced by the environment. The method classifies the pedestrians on the road by adopting a magnetic model, considers three common pedestrian conditions which are easy to cause danger, analyzes partial potential dangers of the pedestrians, and realizes attention to partial special groups on the road. The invention finds out the pedestrians which are easy to cause the accident by analyzing the common road condition environment, and the pedestrians are not necessarily collided with the vehicle, so the pedestrians cannot be classified by the existing method. Compared with the existing method, the pedestrian classification result can further enrich the pedestrian information obtained by the existing road pedestrian classification method, and provide more sufficient road condition information for the autonomous automobile and the assistant driving system. The method comprises the steps of firstly using the existing road pedestrian classification method to find out pedestrians which can directly collide with vehicles, and then using the method to find out other pedestrians which are easy to cause danger, so that a safer path scheme can be obtained when an autonomous automobile or an auxiliary driving system carries out obstacle avoidance and path planning. In the routing of the autonomous vehicle, different classes of magnetic pedestrians have different basic routing. The repulsion pedestrians and the attraction pedestrians can influence the path selection of the vehicles only when being positioned on the driving path of the vehicles, and for the repulsion pedestrians, the vehicles can preferentially select to pass through the pedestrians after the repulsion pedestrians are mutually staggered; for a suction pedestrian, the vehicle will preferentially choose to pass in the opposite direction of the direction of movement of the suction pedestrian. When a pedestrian without magnetism appears in the front of the vehicle, the vehicle can preferentially select to stop driving no matter whether the pedestrian is positioned on the driving route of the vehicle or not, and the pedestrian without magnetism is waited to leave the front visual angle and then continue driving. Because the pedestrian without magnetic force has two states of high moving speed or low moving speed, the pedestrian may be a special crowd and needs to be given a gift for the pedestrian with low moving speed; even if a pedestrian moving at a high speed is not on the driving route of the vehicle at the present time, the pedestrian may move to the vicinity of the front of the vehicle and collide with the vehicle at a later time, and therefore the vehicle may choose to stop when a pedestrian without magnetic force is present. The problem of selecting the path of the vehicle when the single magnetic force pedestrian exists is described above, when a plurality of magnetic force pedestrians occur and the basic paths of the magnetic force pedestrians conflict with each other, the vehicle can choose to stop traveling, otherwise, the vehicle can select to travel according to the basic paths. For example, non-magnetic pedestrians and repulsive pedestrians exist in the front of the vehicle, and the vehicle can choose to stop; repulsion pedestrian and suction pedestrian exist in the plantago, and the suction pedestrian walks to the left side from the right side of plantago visual angle, and the suction pedestrian is on repulsion pedestrian's right side, and the route selection conflict between them, and the vehicle can select to park this moment. The influence of the magnetic force pedestrian on the vehicle routing is described above, when a dangerous pedestrian and a magnetic force pedestrian which collide with the vehicle appear simultaneously, the vehicle considers the route capable of avoiding the dangerous pedestrian first, then judges whether the route accords with the routing of the magnetic force pedestrian, and stops traveling if the route does not accord with the routing of the magnetic force pedestrian. Take the example of a dangerous pedestrian in the center of the view in front of the vehicle, and the pedestrian walks to the right of the view in front of the vehicle. In order to avoid the pedestrian, the vehicle can choose to pass through the left side of the pedestrian, if the repulsive force pedestrian exists on the right side of the front of the vehicle, the route selection of the vehicle is not influenced, because the repulsive force pedestrian is not on the driving route, if the repulsive force pedestrian is on the left side of the front of the vehicle, the vehicle can choose to wait for the passage of the repulsive force pedestrian after being staggered; if a suction pedestrian exists on the front right side of the vehicle, no matter how the movement direction of the suction pedestrian exists, the path selection of the vehicle is not influenced, if the suction pedestrian walks on the front left side of the vehicle and to the right side of the front visual angle, the vehicle can choose to run from the left side of the suction pedestrian, if the suction pedestrian walks on the front left side of the vehicle and to the left side of the front visual angle, the vehicle can choose to pass between the suction pedestrian and a dangerous pedestrian, and meanwhile, the suction pedestrian and the dangerous pedestrian can meet the judgment requirement of a repulsion pedestrian at the previous moment, so the running route also meets the basic path selection of the repulsion of the pedestrian; if no-magnetic force pedestrians appear in the front of the vehicle, the vehicle can stop moving forward no matter whether the no-magnetic force pedestrians are located on the left side or the right side in front of the vehicle.
A road pedestrian classification system storing a plurality of instructions adapted to be loaded and executed by a processor: the vehicle-mounted camera shoots images of the pedestrians on the roads, and the pedestrians on the roads are classified according to the magnetic force relation existing among the pedestrians on the roads displayed by the images; wherein: the magnetic force relationship is characterized by the moving speed of the pedestrians, the relative distance between the pedestrians in the image and the relative distance between the pedestrians and the camera.
The magnetic force relation is represented by a magnetic force model, the magnetic force model comprises a repulsion force model, an attraction force model and a non-magnetic force model, the probability of each specific model in the magnetic force relation displayed by the current image is calculated, and if the probability of the specific model exceeds a probability threshold value, the magnetic force relation of the road pedestrian in the current image is reflected by the current model.
The repulsion model is defined as follows: the pedestrians who run in opposite directions have collision routes in a plurality of traveling routes, and the moving speed of the pedestrians is reduced or even stopped when the pedestrians meet each other until the speed is recovered after meeting is finished.
The magnetic force relationship of the pedestrian displayed by the current image is the probability of the repulsive force model calculated by the following formula:
when the pedestrians i and j move in opposite directions and approach each other continuously, the probability P of repulsion isrWill increase gradually until the probability of repulsion exceeds δ and is judged as a magnetic pedestrian; wherein: priProbability of repulsion for pedestrian i; sijIs the distance between pedestrians i and j; dijIs the difference between the distances of the pedestrians i and j, respectively, from the camera, and dij＝|zi-zjL, wherein ziIs the distance between the pedestrian i and the camera, zjIs the distance between the pedestrian j and the camera; m and n are constant coefficients, and m, n belongs to (0, 1); v. ofiAnd vjThe speeds of movement of pedestrians i and j, respectivelyDegree, vivj< 0 indicates that the pedestrians i and j are traveling in opposite directions.
The suction model is defined as follows: among a plurality of pedestrians walking in the same direction, two or more pedestrians walk in an overlapped or adjacent state, and the moving speeds of the pedestrians are similar in direction and magnitude.
The magnetic force relationship of the pedestrian displayed by the current image is that the probability of the suction model is calculated by the following formula:
when pedestrians i and j continuously walk in the same direction and overlap or in the adjacent state, the probability of suction force
Gradually increasing until the suction probability exceeds delta so as to be judged as a suction pedestrian;
wherein:
the probability of suction being a pedestrian i; sijIs the distance between pedestrians i and j; dijIs the difference between the distances of the pedestrians i and j and the camera, and dij＝|zi-zjL, wherein ziIs the distance between the pedestrian i and the camera, zjIs the distance between the pedestrian j and the camera; v. ofiAnd vjThe moving speeds, v, of pedestrians i and j, respectivelyivj> 0 indicates that pedestrians i and j are traveling in the same direction; k is a radical ofijIs a parameter of suction, and kij＝-[min(T|vi|,T|vj|)/sij]Where T is the elapsed time, min (T | v)i|,T|vj|) represents a smaller displacement value of the pedestrians i and j after the time T.
The definition of the nonmagnetic model is as follows: among the several pedestrians, there are at least three pedestrians, one of which has a higher or lower speed than the other at least two pedestrians.
The probability that the magnetic force relationship of a pedestrian displayed by the current image is a no-magnetic force model is calculated by the following formula:
the pedestrians i, j and h walk in the same direction, the speed of one pedestrian is obviously higher than or lower than the speeds of the other two pedestrians, the speeds of the other two pedestrians are equal, the current speeds of the three pedestrians are kept unchanged, and the magnetic probability P is not generatedmExceeds δ to be determined as a no-magnetic pedestrian; wherein:
the magnetic force free probability of the pedestrian i; v. ofi、vjAnd vhThe moving speeds of the pedestrians i, j and h respectively, the speed of the pedestrian i is obviously higher than or lower than the speeds of the other two pedestrians, and the speeds of the three pedestrians meet the following conditions:
(vi 2-vj 2)+(vi 2-vh 2)＞ε2
then
Equal to 1, otherwise 0;
ε is a constant coefficient.
The specific method for classifying the pedestrians is as follows: calculating the magnetic force probability of a certain pedestrian in the current image:
PMM＝{Pr,Ps,Pm}
when no model probability exceeds a probability threshold value delta, the pedestrian i is a common pedestrian; when only one model probability exceeds a probability threshold value delta, the pedestrian i is judged as a corresponding type of magnetic pedestrian; when more than two model probabilities exceed the probability threshold value delta, the pedestrian i is judged as a magnetic pedestrian with higher priority according to the priority of the model probabilities;
setting: when P is presentrThe pedestrian is a repulsive pedestrian when the delta is exceeded; when P is presentsWhen delta is exceeded, the pedestrian is a suction pedestrian; when P is presentmWhen delta is exceeded, the pedestrian is a non-magnetic pedestrian; three kinds of magnetic force probability PmOf highest priority, PsIs centered on the priority of PrIs lowest.
Example 1:
repulsive force pedestrian classification situation
The simulation results for the classification of repulsive pedestrians in this example are shown in fig. 2. Fig. 2 lists three frames of images in consecutive video frames and the pedestrian classification results for the frames, wherein the magnetic force probability of a pedestrian satisfies the determination requirement of a magnetic force pedestrian only by the repulsive force probability. Three pedestrian targets move at the speed of about 1.2m/s in the video, wherein two pedestrians move in the positive direction, one pedestrian moves in the negative direction, and the three pedestrian targets move in a straight line without changing the moving speed. From frame 8 to frame 33, pedestrians B and C approach each other. The repulsive probabilities of the pedestrians B and C until the 33 th frame exceed δ, and are determined as repulsive pedestrians. Similarly, at the 72 th frame, the repulsive probability of the pedestrians a and C exceeds δ, and is determined as a repulsive pedestrian, at this time, the pedestrians B and C have finished the meeting process, the repulsive probability is decreased, and the pedestrian B is determined as a normal pedestrian.
Example 2:
suction force pedestrian classification situation
The simulation results for the classification of repulsive pedestrians in this example are shown in fig. 3. Fig. 3 lists three images in consecutive video frames and the pedestrian classification results for the frames, wherein the magnetic force probability and the attraction probability of the pedestrian meet the determination requirement of the magnetic force pedestrian. Three pedestrian objects move in the forward direction at a speed of about 1.2m/s in the video, and all keep moving straight without changing the moving speed. From frame 11 to frame 39, pedestrians B and C keep walking concomitantly. The probability of the suction force of the pedestrians B and C exceeds δ until the 39 th frame is determined as a suction pedestrian. The pedestrians B and C maintain the determination result of the sucking pedestrian at the 75 th frame thereafter.
Example 3:
non-magnetic pedestrian classification condition
The simulation result of the present example for the classification of the non-magnetic pedestrian is shown in fig. 4. Fig. 4 lists three frames of images in consecutive video frames and the pedestrian classification results for the frames, wherein the magnetic force probability of the pedestrian satisfies the determination requirement of the magnetic force pedestrian only without the magnetic force probability. In the video, three pedestrian targets move along the negative direction at different speeds, and all move linearly without changing the moving speed, wherein the speed of a pedestrian A is about 0.5m/s, and the speeds of pedestrians B and C are about 1.3 m/s. At the time of the 9 th frame, the calculated non-magnetic probability of the pedestrian a is 1, the pedestrian is determined to be a non-magnetic pedestrian, and the pedestrian a maintains the determination result of the non-magnetic pedestrian at the following 42 th frame and 103 th frame.
Example 4:
hybrid magnetic pedestrian classification scenario
The simulation results for the classification of the mixed magnetic force pedestrian are shown in fig. 5. Fig. 5 lists three frames of images in consecutive video frames and the pedestrian classification results of the frames, wherein the magnetic force probability of the pedestrian with repulsive force probability and the magnetic force probability without magnetic force satisfy the determination requirements of the magnetic force pedestrian. In the video, the pedestrian a moves in the positive direction at a speed of 1.19m/s, the pedestrian B moves in the positive direction at a speed of 1.83m/s, the pedestrian C moves in the negative direction at a speed of 1.21m/s, and the three pedestrians keep moving straight and do not change the moving speed. At frame 6, the probability of no magnetic force of the pedestrian B is calculated to be 1, and the pedestrian is determined to be a no magnetic force pedestrian. By frame 54, the probability of repulsion for pedestrians B and C exceeds δ, where the probability of no magnetic force for pedestrian B also exceeds δ, thus pedestrian B is determined to be a no magnetic pedestrian and pedestrian C is determined to be a repulsive pedestrian. By the 109 th frame, the pedestrian B and the pedestrian C end the meeting process early, the repulsive probability of the pedestrian B is reduced to be lower than δ, but the non-magnetic probability of the pedestrian B is not changed and is still determined as the non-magnetic pedestrian. And the pedestrian C and the pedestrian A are close to each other, so that the probability of repulsion of the pedestrians C and A is increased and exceeds delta, and the pedestrians C and A are judged to be repulsive pedestrians.
Claims (1)
1. A method for judging pedestrians interested in a magnetic force model is characterized in that a vehicle-mounted camera is used for shooting pedestrians, the moving speed, the moving direction, the distance from the camera and the relative position between the pedestrians of the vehicle-mounted camera are measured, corresponding magnetic force model analysis is carried out on the measured information, the degree of coincidence of the pedestrians and classification conditions set by the magnetic force model is calculated, and the pedestrians are divided into normal pedestrians and magnetic force pedestrians;
the magnetic force model comprises a repulsion force model, a suction force model and a non-magnetic force model, the probability of each specific model in the magnetic force relation displayed by the current image is calculated, and if the probability of the specific model exceeds a probability threshold value, the magnetic force relation of the road pedestrians in the current image is reflected by the current model; when more than two model probabilities exceed the probability threshold, the pedestrian is judged as a magnetic pedestrian with higher priority according to the priority of the model probabilities;
the repulsion model is defined as follows: pedestrians walking in opposite directions have collision routes in a plurality of advancing routes, and the moving speed of the pedestrians is reduced or even stopped when meeting is carried out until the speed is recovered after the meeting is finished;
the suction model is defined as follows: among a plurality of pedestrians walking in the same direction, two or more pedestrians walk in an overlapped or adjacent state, and the moving speeds of the pedestrians are close in direction and size;
the definition of the nonmagnetic model is as follows: among the several pedestrians, there are at least three pedestrians, one of which has a higher or lower speed than the other at least two pedestrians.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810936629.5A CN108985271B (en) | 2018-08-16 | 2018-08-16 | Method for judging pedestrian of interest of magnetic model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810936629.5A CN108985271B (en) | 2018-08-16 | 2018-08-16 | Method for judging pedestrian of interest of magnetic model |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108985271A CN108985271A (en) | 2018-12-11 |
| CN108985271B true CN108985271B (en) | 2021-10-08 |
Family
ID=64553379
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810936629.5A Expired - Fee Related CN108985271B (en) | 2018-08-16 | 2018-08-16 | Method for judging pedestrian of interest of magnetic model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108985271B (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102096803A (en) * | 2010-11-29 | 2011-06-15 | 吉林大学 | Safe state recognition system for people on basis of machine vision |
| CN102765365A (en) * | 2011-05-06 | 2012-11-07 | 香港生产力促进局 | Pedestrian detection method based on machine vision and pedestrian anti-collision warning system based on machine vision |
| CN104802793A (en) * | 2014-01-23 | 2015-07-29 | 罗伯特·博世有限公司 | Method and device for classifying a behavior of a pedestrian when crossing a roadway of a vehicle as well as passenger protection system of a vehicle |
| CN108122431A (en) * | 2016-11-30 | 2018-06-05 | 中国移动通信有限公司研究院 | A kind of traffic accident method for early warning and device |
Family Cites Families (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| FR2896070B1 (en) * | 2006-01-11 | 2008-02-15 | Commissariat Energie Atomique | MAGNETIC SYSTEM FOR CONTROLLING TRAFFIC |
| WO2010099416A1 (en) * | 2009-02-27 | 2010-09-02 | Magna Electronics | Alert system for vehicle |
| KR101611194B1 (en) * | 2014-04-04 | 2016-04-11 | 주식회사 와이즈오토모티브 | Apparatus and method for peripheral image generation of vehicle |
| CN108803626B (en) * | 2018-08-16 | 2021-01-26 | 大连民族大学 | System for planning a route for an autonomous vehicle or a driver assistance system |
| CN109145980B (en) * | 2018-08-16 | 2020-09-22 | 大连民族大学 | Magnetic force probability algorithm and system for road pedestrian classification |
| CN108961838B (en) * | 2018-08-16 | 2020-09-22 | 大连民族大学 | Road pedestrian classification system |
| KR20200054657A (en) * | 2018-11-12 | 2020-05-20 | (주) 에코투모로우코리아 | Automotive pulse radar device |
-
2018
- 2018-08-16 CN CN201810936629.5A patent/CN108985271B/en not_active Expired - Fee Related
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102096803A (en) * | 2010-11-29 | 2011-06-15 | 吉林大学 | Safe state recognition system for people on basis of machine vision |
| CN102765365A (en) * | 2011-05-06 | 2012-11-07 | 香港生产力促进局 | Pedestrian detection method based on machine vision and pedestrian anti-collision warning system based on machine vision |
| CN104802793A (en) * | 2014-01-23 | 2015-07-29 | 罗伯特·博世有限公司 | Method and device for classifying a behavior of a pedestrian when crossing a roadway of a vehicle as well as passenger protection system of a vehicle |
| CN108122431A (en) * | 2016-11-30 | 2018-06-05 | 中国移动通信有限公司研究院 | A kind of traffic accident method for early warning and device |
Non-Patent Citations (1)
| Title |
|---|
| 一种基于智能手机的行人航位推算室内定位方法;徐龙阳;《电脑知识与技术》;20171231;第13卷(第36期);第198-221页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108985271A (en) | 2018-12-11 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| KR102098140B1 (en) | Method for monotoring blind spot of vehicle and blind spot monitor using the same | |
| US10796574B2 (en) | Driving assistance method and device | |
| LU101647B1 (en) | Road pedestrian classification method and top-view pedestrian risk quantitative method in two-dimensional world coordinate system | |
| JP6679152B1 (en) | Accident analysis device, accident analysis method and program | |
| US11120690B2 (en) | Method and device for providing an environmental image of an environment of a mobile apparatus and motor vehicle with such a device | |
| CN113243029B (en) | Other vehicle behavior prediction method and other vehicle behavior prediction device | |
| CN105620489A (en) | Driving assistance system and real-time warning and prompting method for vehicle | |
| JP2000090243A (en) | Periphery monitoring device and method therefor | |
| JP2016001170A (en) | Processing unit, processing program and processing method | |
| JP2008009843A (en) | Obstacle discrimination device | |
| CN108803626B (en) | System for planning a route for an autonomous vehicle or a driver assistance system | |
| CN109145980B (en) | Magnetic force probability algorithm and system for road pedestrian classification | |
| CN108961838B (en) | Road pedestrian classification system | |
| CN109165591B (en) | Road pedestrian classification method | |
| CN113706901A (en) | Intelligent prevention, control and early warning system for accidents at entrance section of highway tunnel | |
| CN108985271B (en) | Method for judging pedestrian of interest of magnetic model | |
| CN108909708B (en) | Method and system for judging road pedestrian repulsive force relationship | |
| CN109145807B (en) | Method and system for judging no-magnetic relation of road pedestrians | |
| CN109147388B (en) | Method and system for judging road pedestrian attraction relationship | |
| CN109147389B (en) | Method for planning route by autonomous automobile or auxiliary driving system | |
| Zhao et al. | Moving object classification using horizontal laser scan data | |
| Seo et al. | Use of a monocular camera to analyze a ground vehicle’s lateral movements for reliable autonomous city driving | |
| Hernández et al. | Laser based collision warning system for high conflict vehicle-pedestrian zones | |
| CN114475588A (en) | Automobile unmanned recognition system and automobile unmanned recognition method | |
| Jyothi et al. | Driver assistance for safe navigation under unstructured traffic environment |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20211008 |