CN111985386A - Method for identifying pedestrian illegal-passing behavior based on planned behavior theory - Google Patents

Abstract

The invention discloses a method for identifying pedestrians' illegal passing behavior based on the theory of planned behavior. The method includes: step 1, collecting relevant perception information in a current traffic scene; step 2, according to the individual characteristics of pedestrians in the relevant perception information, Identify the radical degree of pedestrians; step 3, identify the crowd effect of pedestrians according to the pedestrian group characteristics in the relevant perception information; step 4, according to the historical illegal pedestrian data of the current traffic scene or the same traffic scene, learn that the pedestrian violated regulations The main influencing factors of pedestrian behavior; step 5, through the theory of planned behavior, the results obtained in steps 2, 3 and 4 are integrated to obtain the pedestrian's illegal crossing intention, and then identify the illegal crossing behavior of pedestrians, and output the identification result. The present invention can accurately identify the pedestrian's illegal crossing intention in real time, and can effectively identify the illegal crossing behavior of the pedestrian, thereby supporting driving decision-making.

Description

A Pedestrian Illegal Walking Behavior Recognition Method Based on the Theory of Planned Behavior

technical field

The invention relates to the technical field of safe driving of intelligent traffic systems, in particular to a method for identifying pedestrians' illegal passing behavior based on the theory of planned behavior.

Background technique

In recent years, with the gradual development of smart cars, the number of traffic accidents caused by cars has continued to increase, and the most serious traffic accidents caused by people and vehicles have gradually become the focus of social attention. The identification of pedestrians' illegal crossing behaviors is of great significance to ensure the safety of vehicles and reduce the traffic risks of drivers and road pedestrians. Therefore, it is very necessary to identify the intention of pedestrians' illegal crossing behaviors and their illegal crossing behaviors. The identification of pedestrian illegal crossing behavior needs to comprehensively consider the coupling relationship between people, vehicles and road elements, so as to identify pedestrian crossing intentions and provide support for driving decision-making.

In the prior art, Zhou et al. proposed a method for predicting pedestrian illegally crossing behavior based on the theory of planned behavior. This method obtains the pedestrian's behavior concept, subjective norm, observed behavior control, objective norm, scene danger degree and other variables by means of questionnaire survey, and uses the theory of planned behavior to explain the pedestrian's illegal crossing behavior. However, the questionnaire is too idealized, the actual vehicle cannot obtain relevant information, and the interaction between people, vehicles and roads in the actual scene is not considered, so it cannot be used in the actual scene, but can only be used in the field of accident analysis. Dommes et al. tried to find the subjective and objective reasons for pedestrians crossing the rules by analyzing the factors that affect pedestrians' decision to run a red light. The study extracted 13 behavioral indicators (12 before and while crossing the road, and red light running indicators), and studied the role of some demographic, environmental, and behavior-related variables, which can guide the research on pedestrian crossing decision-making process significance. However, its research stops at the analysis of influencing factors, does not explore its internal reasons and connections, and focuses on specific scenes, which has a certain deviation from the actual scene pedestrians' decision to cross the rules and regulations, so it is difficult to identify the actual behavior of pedestrians, which is not suitable for practical applications. . Therefore, it is necessary to develop a method for identifying pedestrian illegal crossing behavior based on the theory of planned behavior.

SUMMARY OF THE INVENTION

The purpose of the present invention is to provide a pedestrian illegal crossing behavior identification method based on the theory of planned behavior, which can accurately identify the pedestrian illegal crossing intention in real time, and effectively identify the illegal crossing behavior of pedestrians, thereby supporting driving decision-making.

In order to achieve the above object, the present invention provides a method for identifying pedestrian illegally crossing behavior based on the theory of planned behavior, and the method includes:

Step 1, collect relevant perception information in the current traffic scene;

Step 2: Identify the radicalness of the pedestrian according to the individual characteristics of the pedestrian in the relevant perception information;

Step 3: Identify the crowd effect of pedestrians according to the pedestrian group characteristics in the relevant perception information;

Step 4: Learn the main influencing factors of pedestrians' illegal crossing behaviors according to the historical illegal crossing pedestrian data of the traffic scene similar to or the same as the current traffic scene;

Step 5: Through the theory of planned behavior, the results obtained in steps 2, 3 and 4 are integrated to obtain the pedestrian's illegal crossing intention, and then identify the illegal crossing behavior of the pedestrian, and output the identification result.

Further, the radical degree of the pedestrian in the step 2 can be specifically expressed as formula (1):

A _norm =S(A _v +A _a +A _w +A _b +A _d +A _o ) (1)

In formula (1), S represents the sigmoid function, A _v represents the degree of aggression caused by the average speed, A _a represents the degree of aggression caused by the maximum acceleration, A _w represents the degree of aggression caused by the waiting time, and A _b represents the degree of aggression caused by the retreat, A _d represents the degree of aggression caused by the distractor, and A _o represents the degree of aggression caused by the observed traffic flow.

Further, A _v , A _a , A _w , _Ab , _Ad , and A _o are continuous variables with values in the range of 0 to 1, or discrete variables with values of 0 or 1.

Further, in the step 3, the acquisition method of the identified pedestrian's herd effect CE is expressed as formula (2):

和人群数量N的二元函数；In formula (2), the conformity effect CE of pedestrians is defined as the proportion of illegal pedestrians

and a binary function of the number of people N;

The conformity effect CE is standardized by the following formula (9) to obtain the conformity effect CE _norm :

CE _norm = S(CE) (9)

In formula (9), S is a sigmoid function, which converts the crowd effect C into a standard variable between 0 and 1.

Further, the step 4 specifically includes:

Step 4.1, using the data classification method, divide the historical pedestrian data of the same or similar traffic scene as the current traffic scene into non-violation pedestrians and illegal pedestrians, and select the traffic scenes that may violate the rules, and divide the training according to the preset ratio. Set and test set to learn the main scene influencing factors of the differential behavior of the above two groups of people;

Step 4.2, after obtaining the n-dimensional main scene influencing factors of the differential behavior of the non-violent pedestrians and the illegal pedestrians through Step 4.1, further, using the principal component analysis method, obtain the principal components of the crowd's differential behavior scene influence factors, which are: Expressed as formula (10):

In formula (10), Y _i represents the ith principal component, C _i represents the ith scene influencing factor; p represents the dimension of the principal component, n represents the dimension of the scene influencing factor, and A is the principal component of p×n dimension. matrix, whose matrix elements are a ₁₁ to a _pn , and the principal components of the final scene influencing factors are expressed as formula (11):

Y=[Y ₁ , Y ₂ , . . . , Y _p ] ^T (11).

Further, in step 4.1, the main scene influencing factors of the differential behaviors of the above-mentioned two types of people are obtained through the analysis of variance method, and the non-violation pedestrians are defined as class A, the illegal pedestrians are defined as class B, and all the pedestrians are defined as class B. The impact factor of the above-mentioned scenario is defined as the C factor;

Case 1, the C factor has significant variance difference between the A and B populations. When the confidence level is [0.95, 1.00], the hypothesis test value P-value is [0, 0.01], then the factor is considered to be It is one of the main scene influencing factors of the differential behavior of the two groups of people;

In case 2, the C factor does not contain significant variance difference between the A and B groups. When the confidence level is [0.95, 1.00], the hypothesis test value P-value is [0.1, 1.0], then it is considered that the The factor is not one of the main scene influencing factors of the differential behavior of the two groups of people;

In case 3, the C factor has significant variance difference between the A and B groups, but when the confidence level is [0.95, 1.00], its hypothesis test value P-value is (0.01, 0.1), it needs more Divide the training set and the test set to test the significance of this factor.

Further, the step 5 specifically includes:

Step 5.1, according to the pedestrian's radical degree A _norm obtained in step 2, step 3 and step 4, the conformity effect CE _norm , and the principal component Y of the scene influencing factor, the pedestrian's illegal crossing intention I is obtained through the theory of planned behavior, which is expressed as formula (12) to formula (14), and then solve the unknown parameters {a, b, c, d}:

I _o =a+bA _{norm_0} +cCE _{norm_0} +dY ₀ (12)

I _t =αI _t-1 +(1-α)(a+bA _{norm_t} +cCE _{norm_t} +dY _t ) (13)

I _t ∈ [0, 1] (14)

In the formula, I _o represents the initial value of the pedestrian’s intention to cross the rules, A _{norm_0} , CE _{norm_0} , and Y ₀ represent the pedestrian’s radical degree, herd effect, and the value of the principal component of the scene influencing factor at the initial moment, and I _t represents the pedestrian’s The intention of illegal crossing at time _t , α is a preset experience value, It = 0 means that the pedestrian does not have the intention of crossing the rule at all _, It = 1 represents that the pedestrian has a certain intention to cross the rule, It = (0, ₁ ) represents Uncertain intention of illegal crossing, which partially conforms to the probability distribution of pedestrian crossing illegally;

Step 5.2, according to the pedestrian's illegal crossing intention It obtained in step 5.1 and the principal component Y _t of the scene influencing factors, according to the planned behavior theory, the pedestrian's illegal crossing behavior ACT _{t is} expressed as formula (18):

ACT _t =Rule(I _t , Y _t ) (18)

Among them, Rule(I _t , Y _t ) represents a predefined walking rule.

The present invention has the following advantages due to the adoption of the above technical solutions: 1. The present invention establishes a pedestrian illegal crossing behavior identification method that comprehensively considers the coupling relationship between people, vehicles and road elements, and the method analyzes the radical degree and conformity of pedestrians. The influence brought by the effect, taking into account the principal components of the factors influencing pedestrian behavior decision-making by the interaction environment of people, vehicles and roads, can objectively and truly describe the pedestrian's illegal crossing process. 2. The present invention generates a method for identifying illegal passing behavior that supports different types of pedestrians through the aggressiveness degree identification module, which lays a foundation for realizing vehicle differentiated decision-making. 3. The present invention is based on the identification of the herd effect, which reflects the supervision effect brought about by crowd gathering and the shielding effect when the proportion of illegal pedestrians is too large, and provides a feasible theoretical direction for studying the effect of crowd on individuals. 4. Based on the real historical data of the traffic environment, the present invention analyzes the main components of the traffic environment influencing factors for pedestrians who do not violate regulations and pedestrians who violate regulations, reflects the objective impact of pedestrians in the coupled environment of people, vehicles and roads, and based on the theory of planned behavior, Combining the influence of pedestrian internal factors, it realizes the separation and integration of internal and external influencing factors of pedestrian illegal crossing behavior. Compared with the existing pedestrian violation behavior identification method, it can more comprehensively reflect the decision-making mechanism of the pedestrian violation process, and also proves the effectiveness, feasibility and scientificity of the method of the present invention.

Description of drawings

Fig. 1 is the schematic flow chart of illegal behavior identification provided by the embodiment of the present invention;

2 is an information flow diagram provided by an embodiment of the present invention;

3 is a schematic diagram of an algorithm result in which a clustering algorithm provided by an embodiment of the present invention divides the crowd into two categories in advance;

4 is a schematic diagram of an application scenario provided by an embodiment of the present invention;

5 is a schematic diagram of information fusion and illegal behavior identification based on the theory of planned behavior provided by an embodiment of the present invention;

FIG. 6 is a schematic diagram of an information prompt on a driving assistance screen for a manually driven vehicle provided by an embodiment of the present invention.

Detailed ways

The present invention will be described in detail below with reference to the accompanying drawings.

For example, when a smart vehicle enters an intersection with traffic lights and zebra crossings from the side of the road, the pedestrian indicator light is red, and the pedestrian is on the curb. The vehicle needs to judge whether the pedestrian will violate the regulations to pass through, so as to decelerate and avoid the behavior. The specific scene is shown in Figure 4.

As shown in Figure 1 and Figure 2, the method for identifying pedestrian illegally crossing behavior based on the theory of planned behavior provided by the present invention includes:

Step 1: Collect relevant perception information in the current traffic scene. Among them, the "related perception information" is collected by the vehicle perception system through sensors such as vehicle cameras and radars. This information can be understood as time-related and continuously updated data.

Step 2: Identify the radicalness of the pedestrian according to the individual characteristics of the pedestrian in the relevant perceptual information. Among them, the "personal characteristics of pedestrians" include six items of information: the average speed of pedestrians, the maximum acceleration, the waiting time, whether to retreat, whether there is interference, and whether to observe the traffic flow. Among them, the average speed information and acceleration information can be obtained directly through the lidar, the waiting time information can be calculated through the timer, whether the backward information is calculated through the pedestrian position information, whether there is interference information can be obtained through the camera, and whether the traffic flow information is observed through the camera. Get the pedestrian's head orientation. Each item of information is specific to each pedestrian, and this item is the "individual" characteristic of the pedestrian. Specifically, for example: average speed, in order to avoid errors in the data collection of this frame, the data of the current frame and the previous five or six frames are used for average processing to avoid errors.

Step 3: Identify the crowd effect of pedestrians according to the pedestrian group characteristics in the relevant perceptual information. Among them, the "pedestrian group characteristics" include the number of crowds and the proportion of crowd violations. The "conformity effect" can be a binary function of crowd size and crowd violation ratio. "Number of crowds" can be understood as the number of pedestrians contained in the crowd.

Step 4: According to the historical data of illegal pedestrians passing through the traffic scene similar to or the same as the current traffic scene, the main influencing factors of the illegal passing behavior of the pedestrians are known. Among them, the "scenario influencing factors" refers to the principal components of the factors that affect pedestrians' illegal crossing through the means of variance analysis and principal component analysis. Among them, "historical illegal travel data" includes a large number of public data sets, videos obtained by surveillance cameras, and data obtained by aerial cameras.

Step 5: Through the theory of planned behavior, the results obtained in Step 2, Step 3 and Step 4 are integrated to obtain the pedestrian's illegal crossing intention, and then identify the illegal crossing behavior of pedestrians, and provide corresponding information on the driving assistance screen for manually driven vehicles. information.

This example comprehensively considers the inherent characteristics of pedestrians, crowd characteristics and environmental influence factors in the process of pedestrian illegally passing through.

As a preferred implementation method for identifying the aggressiveness of pedestrians in step 2, it can be expressed as formula (1):

A _norm =S(A _v +A _a +A _w +A _b +A _d +A _o ) (1)

In formula (1), S represents the sigmoid function, A _v represents the degree of aggression caused by the average speed, A _a represents the degree of aggression caused by the maximum acceleration, A _w represents the degree of aggression caused by the waiting time, and A _b represents the degree of aggression caused by the retreat, A _d represents the degree of aggression caused by the distractor, and A _o represents the degree of aggression caused by the observed traffic flow. The values of _Av , A _a , A _w , _{Ab , Ad , and A o may be} continuous variables in the range of 0 to 1, or may be discrete variables, such as: 0 or 1. The specific value methods of these A with different subscripts include:

First, define whether A _v , A _a , A _w , _{Ab , A d ,} and A _o are discrete or continuous variables; then, according to the historical real data of pedestrians, the aggressiveness of pedestrians and A _v , A _a , A _w , The values of _Ab , _{Ad , and A o are} subjected to deep learning training to obtain the true values of _Av , A _a , A _w , _Ab , _{Ad , and A o .} The historical real data of pedestrians include the historical real data of individual characteristics of pedestrians and the real value of historical pedestrian radicalness (calculated through questionnaires or the degree of danger when they pass through).

The sigmoid function in this embodiment can also be replaced with a selu function or a tanh function, etc., the purpose is to map the defined aggressiveness between (0, 1), thereby eliminating the difference of the input.

Of course, the parameters in formula (1) can also be represented by other factors, such as: the degree of aggression brought by age, the degree of aggression brought by gender, and/or the degree of aggression brought by distractors, etc. However, the parameters selected in Equation (1) can be obtained through vehicle perception and are easily applied in practical scenarios. And through literature research, it can be proved that the parameters selected in the sigmoid function have a significant impact on the pedestrian's aggressiveness.

In one embodiment, the acquisition mode of the herd effect CE identified in step 3 is expressed as formula (2):

和人群数量N的二元函数。随着人群数量N的增加，根据违章行人比例

的不同，人群产生互相监督或互相包庇的效应。“监督/包庇效应”实际是Conformity effect，也就是“从众效应”的一种说法，指的是单个行人的行为随着其他人的行为产生变化的倾向，目前一般通过问卷调查获得行人的从众效应倾向，但本实施例欲采用人群数量和人群违章比例拟合，用真实数据训练出对应的二元函数。“监督/包庇效应”根据人群的违章比例发生变化。In formula (2), the conformity effect CE of pedestrians is defined as the proportion of illegal pedestrians

and a binary function of the number of people N. As the number of people N increases, according to the proportion of illegal pedestrians

different, the crowds have the effect of monitoring or shielding each other. The "supervision/protection effect" is actually the Conformity effect, which is a term for the "conformity effect", which refers to the tendency of a pedestrian's behavior to change with the behavior of others. At present, the conformity effect of pedestrians is generally obtained through questionnaires. However, this embodiment intends to use the number of crowds and the proportion of crowd violations to fit, and use real data to train a corresponding binary function. The "surveillance/cover-up effect" varies according to the percentage of violations in the population.

的二元函数可以表示为下式(3)、(4)和(5)：Specifically, the number of people N and the proportion of illegal pedestrians defined by the conformity effect CE of pedestrians

The binary function of can be expressed as the following equations (3), (4) and (5):

时，无从众效应。当然，也可以根据需要，预先定义a、b、c、d、e、f的具体数值。In the formula, a, b, c, d, e, and f are all predefined positive coefficients, for example: a=0.04, b=1.1, d=f=1, c=e=2, when

, there is no conformity effect. Of course, the specific numerical values of a, b, c, d, e, and f can also be pre-defined as required.

以1/2为分界点，在

的范围内，有人违章破坏了从众效应，从众效应的值降低；在

时，形成了破坏规则的从众效应，故从众效应的值反而上升。It can be seen from this that the number of people N takes 4 as the dividing point, and within the range of 1≤N≤4, with the increase of the number of people N, the supervision effect is enhanced, and the value of the conformity effect increases; when N>4, The value of the herd effect no longer changes. Percentage of illegal pedestrians

Taking 1/2 as the dividing point, at

Within the range of , someone violates the rules and destroys the conformity effect, and the value of the conformity effect decreases;

When , the conformity effect that breaks the rules is formed, so the value of the conformity effect increases instead.

的二元函数可以表示为下式(6)和(7)：Specifically, the number of people N and the proportion of illegal pedestrians defined by the conformity effect CE of pedestrians

The binary function of can be expressed as the following equations (6) and (7):

In the formula, abs(*) is the absolute value function, and the trend is the same.

In this embodiment, the sigmoid activation function is used, and the following formula (9) is used to standardize the conformity effect CE to obtain the conformity effect CE _norm :

CE _norm = S(CE) (9)

In formula (9), S is a sigmoid function, which converts the crowd effect CE into a standard variable between 0 and 1 for subsequent processing.

The sigmoid function in this embodiment may also be replaced by a selu function or a tanh function, or the like. The parameters selected in Equation (8) can be obtained through vehicle perception and can be easily applied in practical scenarios. Through literature research, it is proved that the parameters selected in formulas (6)-(8) can find evidence and corresponding data that have a significant impact on the conformity effect.

Standardizing the herd effect by formula (9) can eliminate the error caused by the input scale to the final neural network training to obtain illegal behavior. Without normalization, if some parameters are too large (such as herd effect) and some parameters are too small (such as aggressiveness), the effect of aggressiveness may be directly ignored (the value is too small).

In one embodiment, the above-mentioned conformity effect CE can also be obtained by means of a questionnaire survey, and the participants reflect their tendency to conform by scoring multiple questions raised. That is to say, the conformity effect can also be expressed as a scoring function: CE(i)=grade(pedes(i)), where i represents the ith pedestrian. However, this method can only be used as prior knowledge, and it is difficult to be applied in practice, because pedestrians encountered on actual roads cannot be scored.

The herd effect CE can also be defined using rules, such as: CE(i)=CE(j)=f(N), where i represents the ith and jth pedestrians in a group of pedestrians. That is, when the i-th pedestrian is the leader and violates the law, and the follower j-th pedestrian’s conformity effect CE is greater than the threshold, it will also violate the law. This also reflects the herd effect, but it is difficult to identify leaders and followers, the threshold is difficult to find, and the defined rules are often difficult to achieve perfection.

In one embodiment, step 4 specifically includes:

Step 4.1, as shown in Figure 3, use the data classification method to divide the historical pedestrian data of the same or similar traffic scene as the current traffic scene shown in Figure 4 into: define non-violation pedestrians as Class A, and illegal pedestrians Pedestrians are defined as class B. Only select the traffic scenes that may violate the regulations, divide the training set and the test set with a preset ratio (such as a ratio of 3:1), and use the variance analysis method to learn the main factors influencing the differential behavior of the above two groups of people. The data classification method may be a clustering algorithm, a support vector machine, a synovial algorithm, or the like.

Case 1, the C factor has significant variance difference between the A and B populations. When the confidence level is [0.95, 1.00], the hypothesis test value P-value is [0, 0.01], then the factor is considered to be It is one of the main scene influencing factors of the differential behavior of the two groups of people.

In case 2, the C factor does not contain significant variance difference between the A and B groups. When the confidence level is [0.95, 1.00], the hypothesis test value P-value is [0.1, 1.0], then it is considered that the Factors are not one of the main scene influence factors of the differential behavior of the two groups of people.

In case 3, the C factor has significant variance difference between the A and B groups, but when the confidence level is [0.95, 1.00], its hypothesis test value P-value is (0.01, 0.1), it needs more Divide the training set and the test set to test the significance of this factor.

Step 4.2, after obtaining the n-dimensional main scene influencing factors of the differential behavior of the A and B groups through step 4.1, further, using the principal component analysis method, obtain the principal components of the scene influencing factors of the crowd differential behavior, which is expressed as formula (10) :

In formula (10), Y _i represents the ith principal component, C _i represents the ith scene influencing factor; p represents the dimension of the principal component, n represents the dimension of the scene influencing factor, and A is the principal component of p×n dimension. A matrix whose matrix elements are a ₁₁ to a _pn . The principal components of the final scene influencing factors are expressed as formula (11):

Y=[Y ₁ , Y ₂ , ..., Y _p ] ^T (11)

The variance analysis method in step 4.1 can also be replaced by clustering, t-test, or z-test method.

The principal component analysis method in step 4.2 can also be replaced by a factor analysis method or the like.

In one embodiment, step 5 specifically includes:

Step 5.1, according to the pedestrian radical degree A _norm obtained in step 2, step 3, and step 4, the conformity effect CE _norm , and the principal component Y of the scene influencing factor, convert it into three variables: behavioral concept, subjective norm, and observational behavior control, such as As shown in Figure 5, the pedestrian's illegal crossing intention I is obtained through the theory of planned behavior, which is expressed as formula (12) to formula (14), and then the unknown parameters {a, b, c, d} are solved:

I _o =a+bA _{norm_0} +cCE _{norm_0} +dY ₀ (12)

I _t =αI _t-1 +(1-α)(a+bA _{norm_t} +cCE _{norm_t} +dY _t ) (13)

I _t ∈ [0, 1] (14)

In formula (12), I _o represents the initial value of the pedestrian's illegal crossing intention, which is calculated according to the pedestrian's aggressiveness, herd effect and scene principal component of the initial time frame, or directly set an arbitrary value of 0-1. Yes, because it will be iterated over time, approaching the true value; A _{norm_0} , CE _{norm_0} , Y ₀ respectively represent the radical degree of pedestrians, the herd effect, and the value of the principal component of the scene influencing factor at the initial moment, The initial moment is the moment when the vehicle starts to observe the pedestrian, and the end moment is the moment when the sensor cannot perceive the pedestrian. The specific values of A _{norm_0} , CE _{norm_0} , and Y ₀ are calculated according to the pedestrian characteristics, group characteristics, and scene information perceived by the sensor in the current frame through the formula and model previously proposed.

In formula (13), I _t represents the pedestrian's illegal crossing intention at time t, which is affected by the illegal crossing intention It _-1 at the previous moment and the pedestrian's aggressiveness at time t, herd effect, and the principal components of scene influencing factors A _{norm_t} , CE _{norm_t} , Y _t , and its influence ratio is α:(1-α); α is a preset empirical value, such as 0.5.

In formula (14), It = 0 means that the pedestrian has no intention to pass through _illegally at all, It = 1 _means that the pedestrian has a certain intention to pass through _illegally , It = (0, 1) represents the intention to pass through indefinitely, and partially It conforms to the probability distribution of pedestrians crossing illegally.

It is also possible to obtain the pedestrian's illegal crossing intention I through the first-order quadratic terms represented by equations (15) to (17), and then solve the unknown parameters {a ₁ , b ₁ , b ₂ , c ₁ , c ₂ , d ₁ , d ₂ }:

I _t ∈ [0, 1] (17)

In the same way, you can also obtain the pedestrian's illegal crossing intention I through the third and fourth items.

In step 5.2, according to the pedestrian's illegal crossing intention It and the principal component Y _t of the scene influencing factors obtained in step 5.1 _, as shown in Figure 5, according to the theory of planned behavior, the pedestrian's illegal crossing behavior ACT _t is expressed as formula (18):

ACT _t =Rule(I _t , Y _t ) (18)

Among them, Rule(I _t , Y _t ) represents a predefined walking rule. like:

That is, if and only when the pedestrian's intention to walk illegally is greater than half, and the risk component of the principal component of the scene influencing factor is also less than half, it is considered that the pedestrian will walk illegally, otherwise, it is considered that the pedestrian will not walk illegally.

Of course, the ACT _t table of pedestrian's illegal walking behavior can also make simple judgments according to the actual situation, but the correct rate is difficult to reach more than 80%. The intention It and the scene factor Y _t can also be used for training by using a neural network method _. Similar methods include support vector machine SVM, k-means classification method, and neural network neutral network.

In one embodiment, as shown in FIG. 6 , for a manually driven vehicle, it is necessary to display the current suggested vehicle speed, signal light status, pedestrian violation probability and suggested decision-making information on the driving assistance screen to assist the driver in making reasonable decisions and avoid pedestrians violating regulations. resulting in a collision.

Finally, it should be pointed out that the above embodiments are only used to illustrate the technical solutions of the present invention, but not to limit them. Those of ordinary skill in the art should understand that: the technical solutions described in the foregoing embodiments can be modified, or some technical features thereof can be equivalently replaced; these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the various aspects of the present invention. The spirit and scope of the technical solutions of the embodiments.

Claims (7)

1. A pedestrian violation passing behavior identification method based on a planned behavior theory is characterized by comprising the following steps:

step 1, collecting related perception information in a current traffic scene;

step 2, identifying the degree of the pedestrian according to the individual characteristics of the pedestrian in the relevant perception information;

step 3, identifying the crowd effect of the pedestrian according to the pedestrian population characteristics in the relevant perception information;

step 4, acquiring main influence factors of the pedestrian violation trafficking behaviors according to historical pedestrian violation trafficking data similar to or identical to the current traffic scene;

and 5, fusing the results obtained in the step 2, the step 3 and the step 4 through a planned behavior theory to obtain the violation passing intention of the pedestrian, further identifying the violation passing behavior of the pedestrian, and outputting an identification result.

2. The method for identifying the pedestrian violation passing behavior based on the planning behavior theory as claimed in claim 1, wherein the degree of aggressiveness of the pedestrian in the step 2 can be specifically expressed as formula (1):

A_norm＝S(A_v+A_a+A_w+A_b+A_d+A_o) (1)

in the formula (1), S represents sigmoidFunction, A_vIndicates the degree of acceleration due to average velocity, A_aIndicates the degree of acceleration due to maximum acceleration, A_wIndicating the degree of aggressiveness due to latency, A_bIndicates the degree of acceleration due to backward movement, A_dIndicates the degree of invasiveness caused by the interfering substance, A_oIndicating the degree of aggressiveness resulting from observing traffic.

3. The method for identifying pedestrian violation traversal behavior based on planning behavior theory as claimed in claim 2, wherein A is_v、A_a、A_w、A_b、A_dAnd A_oThe variable is a continuous variable with a value in the range of 0-1, or a discrete variable with a value of 0 or 1.

4. The method for identifying the pedestrian violation passing behavior based on the planning behavior theory as claimed in claim 1, wherein in the step 3, the acquisition mode of the identified pedestrian subordinate effect CE is represented by formula (2):

in the formula (2), the pedestrian dependent effect CE is defined as the proportion of the pedestrian violating the regulations

And a binary function of the number of people N;

the dominant effect CE is obtained by normalizing the dominant effect CE by the following formula (9)_norm：

CE_norm＝S(CE) (9)

In the formula (9), S is a sigmoid function and converts the conotoxic effect C into a standard variable between 0 and 1.

5. The pedestrian violation passing behavior identification method based on the planning behavior theory as claimed in claim 1, wherein the step 4 specifically comprises:

step 4.1, using a data classification method to divide historical pedestrian data of the same or similar traffic scene as the current traffic scene into non-violation-passing pedestrians and violation-passing pedestrians, selecting a traffic scene which is possible to violate the traffic scene, and dividing a training set and a testing set according to a preset proportion to obtain main scene influence factors of the differential behavior of the two classes of people;

step 4.2, after the n-dimensional main scene influence factors of the group difference behaviors of the pedestrian who does not pass the violation and the pedestrian who passes the violation are obtained through the step 4.1, further, a main component analysis method is used to obtain main components of the group difference behavior scene influence factors, wherein the main components are expressed as a formula (10):

in the formula (10), Y_iRepresents the ith principal component, C_iRepresenting the ith scene influencing factor; p denotes the principal component dimension, n denotes the dimension of the scene influencing factor, A is a principal component matrix of dimension p x n, the matrix elements of which are a₁₁To a_pnThe final scene influencing factor principal component is expressed by equation (11):

Y＝[Y₁,Y₂,…,Y_p]^T (11)。

6. the method for identifying the pedestrian violation trafficking behavior based on the planning behavior theory as claimed in claim 5, wherein in step 4.1, the main scene influencing factors of the differential behaviors of the two classes of people are obtained through a variance analysis method, the pedestrian class which does not violate the regulations is defined as class A, the pedestrian class which does violate the regulations is defined as class B, and the scene influencing factors are defined as factor C;

case one, the factor C contains a significant variance difference in A, B people, and in the case of a confidence level of [0.95,1.00], it assumes that the test value P-value is [0,0.01], and it is considered to be one of the main scene influencing factors of the differentiated behavior of the two people;

in case II, the factor C does not contain a significant variance difference in A, B people, and in case that the confidence level is [0.95,1.00], the factor C assumes that the test value P-value is [0.1,1.0], and then the factor C is considered not to be one of the main scene influence factors of the difference behaviors of the two people;

case three, the factor C contains significant variance difference in the population of class a and B, but in the case of confidence level at [0.95,1.00], which assumes the test value P-value at (0.01,0.1), it is necessary to divide the training set and the test set multiple times to test the significance of the factor.

7. The pedestrian violation passing behavior identification method based on the planning behavior theory as claimed in claim 1, wherein the step 5 specifically comprises:

step 5.1, obtaining the pedestrian excitation degree A according to the step 2, the step 3 and the step 4_normFrom the dominant effect CE_normAnd obtaining the violation passing intention I of the pedestrian through a planning behavior theory by using the main component Y of the scene influence factor, expressing the violation passing intention I as formulas (12) to (14), and further solving unknown parameters { a, b, c and d }:

I_o＝a+bA_{norm_0}+cCE_{norm_0}+dY₀ (12)

I_t＝αI_t-1+(1-α)(a+bA_{norm_t}+cCE_{norm_t}+dY_t) (13)

I_t∈[0,1] (14)

in the formula I_oInitial value representing pedestrian's intention to pass violation, A_{norm_0}、CE_{norm_0}、Y₀Respectively representing the value of the main components of the radical degree, the slave effect and the scene influence factor of the pedestrian at the initial moment, I_tRepresenting the illegal passing intention of the pedestrian at the time t, alpha is a preset empirical value, I_t0 represents that the pedestrian has no illegal passing intention at all, I_t1 represents the pedestrian's confident intention to pass through the traffic violation, I_tThe violation passing intention is uncertain and is partially in line with the probability distribution of the violation passing of the pedestrian;

step 5.2, according to step 5.1 the pedestrian passing violation intention I_tAnd scene influencing factor principal component Y_tAccording to the plan behavior theory, ACT is used for the illegal walk-through behavior of the pedestrian_tRepresented by formula (18):

ACT_t＝Rule(I_t，Y_t) (18)

wherein, Rule (I)_t，Y_t) Representing predefined traversal rules.

Images