CN104700475A - Self-adaptive passenger flow counting algorithm based on video analysis - Google Patents

Abstract

The invention discloses a self-adaptive passenger flow counting algorithm based on video analysis. The passenger flow counting algorithm mainly aims at the condition of passenger flow congestion during the rush hour. The statistic analysis of historical video data and passenger flow counting data proves that under the condition that the number of people who get on a bus once exceeds a certain value (20 people/time), the passenger flow counting accuracy is relatively low, and less counting is mainly caused. Therefore, the passenger flow counting data with larger error is correspondingly corrected according to the passenger flow counting time and the historical passenger flow data, so that the passenger flow counting data is closer to the truth, and the whole accuracy of the passenger flow data can be improved.

Description

Based on the self-adaptation passenger flow counting algorithm of video analysis

Technical field

The present invention relates to passenger flow counting algorithm field, specifically a kind of self-adaptation passenger flow counting algorithm based on video analysis.

Background technology

Along with the development of computer vision counting, the passenger flow counting technology based on video is more and more paid attention to.Mainly by camera acquisition view data, video processing technique is then utilized to obtain passenger flow data.By video passenger flow calculating instrument, can obtain a large amount of bus passenger flow data reliably, stability and the accuracy rate of statistics reach practical purpose completely.The accuracy rate how improving passenger flow counting is the emphasis of research.

Find carrying out actual statistical study to history video record data, when single get on the bus number higher than certain number (20 people/time), counting accuracy rate is relatively low, mainly counts less.。

Summary of the invention

The object of this invention is to provide a kind of self-adaptation passenger flow counting algorithm based on video analysis, to solve the not high problem of accuracy rate under the passenger flow congested conditions of prior art peak period.。

In order to achieve the above object, the technical solution adopted in the present invention is:

Based on the self-adaptation passenger flow counting algorithm of video analysis, it is characterized in that: comprise the following steps:

(1), by history video data certain hour N is counted, the actual value U of the number of getting on the bus that vehicle every day is each in N>=1 month _ii=1,2 ..., N ₁, N ₁for the data volume counted on;

(2) the number W that gets on the bus that every day in the N time, passenger flow counting device statistics went out, is counted _i, and extract when every day gone up visitor by vehicle at every turn and open the door the T.T. of closing the door, namely count passenger's single and to get on the bus T.T. T _i;

(3), from step (1), (2) gained passenger flow counting data W _iin, choose W _i>=20 people/time time corresponding all data, comprise the number W that gets on the bus of passenger flow counting _j, video statistics to the number U that truly gets on the bus _j, passenger's single gets on the bus T.T. T _j, j=1,2 ..., N ₂, N ₂for meeting W _i>=20 people/time data volume, then can be met W _i>=20 people/time time passenger flow counting total number of getting on the bus of obtaining actual total number of getting on the bus total pick-up time the single pick-up time of single and then can obtain at W _i>=20 people/time time passenger flow counting accuracy rate be the mean value of the single pick-up time of single $\stackrel{\‾}{t}=\frac{\underset{j=1}{\overset{N_2}{\mathrm{\Σ}}}{t}_j}{N_2};$

(4), choose the data needing to correct according to people's logarithmic data of getting on the bus of passenger flow counting, comprise the following steps:

(4.1), according to the mean value of the single pick-up time of single choose the enumeration data needing to correct: by t _j? outside Data correction be if desired the data volume corrected is N ₃, then each enumeration data after correcting is respectively k=1,2 ... N ₃, after correcting, total count value is the accuracy rate obtaining correcting rear passenger flow counting is

(4.2), according to total pick-up time T and reality always get on the bus number U ratio namely choose the enumeration data needing to correct: by t _jdata correction outside (t-t*10%, t+t*10%) is t, and the data volume if desired corrected is N ₄, then each enumeration data after correcting is respectively r=1,2 ..., N ₄, then the total count value after correcting is the accuracy rate obtaining correcting rear passenger flow counting is

(5), by actual passenger flow video data and passenger flow counting data, utilize above-mentioned correcting algorithm, compare the accuracy rate of passenger flow counting before and after correcting, the feasibility of verification algorithm;

(6), due to when 20 people/when time to get on the bus, can there is relatively large deviation in passenger flow counting, can select when enumeration data be 18 people/time or 16 people/time more than situation correct, if according to averaging time correct and utilize people/time or the time point of 16 people/time a to draw bound * 18 people/time or 16 people/time, with people/time or 16 people/time, at (t _under, t _on) data outside scope according to T.T. divided by draw the numerical value of correction second.

Algorithm of the present invention to occur that the situation of very large deviation has carried out corresponding correction, make it more close to real situation, from the globality of data, be eliminate larger error to correct in the situation according to normality, improve the accuracy rate of passenger flow data whole body counting.

Accompanying drawing explanation

Fig. 1 is process flow diagram of the present invention.

Embodiment

Shown in Figure 1, based on the self-adaptation passenger flow counting algorithm of video analysis, comprise the following steps:

(1), by history video data certain hour N is counted, the actual value U of the number of getting on the bus that vehicle every day is each in N>=1 month _ii=1,2 ..., N ₁, N ₁for the data volume counted on;

(2) the number W that gets on the bus that every day in the N time, passenger flow counting device statistics went out, is counted _i, and extract when every day gone up visitor by vehicle at every turn and open the door the T.T. of closing the door, namely count passenger's single and to get on the bus T.T. T _i;

(3), from step (1), (2) gained passenger flow counting data W _iin, choose W _i>=20 people/time time corresponding all data, comprise the number W that gets on the bus of passenger flow counting _j, video statistics to the number U that truly gets on the bus _j, passenger's single gets on the bus T.T. T _j, j=1,2 ..., N ₂, N ₂for meeting W _i>=20 people/time data volume, then can be met W _i>=20 people/time time passenger flow counting total number of getting on the bus of obtaining actual total number of getting on the bus total pick-up time the single pick-up time of single and then can obtain at W _i>=20 people/time time passenger flow counting accuracy rate be the mean value of the single pick-up time of single $\stackrel{\‾}{t}=\frac{\underset{j=1}{\overset{N_2}{\mathrm{\Σ}}}{t}_j}{N_2};$

(4), choose the data needing to correct according to people's logarithmic data of getting on the bus of passenger flow counting, comprise the following steps:

(4.1), according to the mean value of the single pick-up time of single choose the enumeration data needing to correct: by t _j? outside Data correction be if desired the data volume corrected is N ₃, then each enumeration data after correcting is respectively k=1,2 ... N ₃, after correcting, total count value is the accuracy rate obtaining correcting rear passenger flow counting is

(4.2), according to total pick-up time T and reality always get on the bus number U ratio namely choose the enumeration data needing to correct: by t _jdata correction outside (t-t*10%, t+t*10%) is t, and the data volume if desired corrected is N ₄, then each enumeration data after correcting is respectively r=1,2 ..., N ₄, then the total count value after correcting is the accuracy rate obtaining correcting rear passenger flow counting is

(5), by actual passenger flow video data and passenger flow counting data, utilize above-mentioned correcting algorithm, compare the accuracy rate of passenger flow counting before and after correcting, the feasibility of verification algorithm;

(6), due to when 20 people/when time to get on the bus, can there is relatively large deviation in passenger flow counting, can select when enumeration data be 18 people/time or 16 people/time more than situation correct, if according to averaging time correct and utilize people/time or the time point of 16 people/time a to draw bound * 18 people/time or 16 people/time, with people/time or 16 people/time, at (t _under, t _on) data outside scope are according to the numerical value drawing correction T.T. divided by t second.

Claims (1)

1., based on the self-adaptation passenger flow counting algorithm of video analysis, it is characterized in that: comprise the following steps:

(1), by history video data certain hour N is counted, the actual value U of the number of getting on the bus that vehicle every day is each in N>=1 month _ii=1,2 ..., N ₁, N ₁for the data volume counted on;

(2) the number W that gets on the bus that every day in the N time, passenger flow counting device statistics went out, is counted _i, and extract when every day gone up visitor by vehicle at every turn and open the door the T.T. of closing the door, namely count passenger's single and to get on the bus T.T. T _i;

(3), from step (1), (2) gained passenger flow counting data W _iin, choose W _i>=20 people/time time corresponding all data, comprise the number W that gets on the bus of passenger flow counting _j, video statistics to the number U that truly gets on the bus _j, passenger's single gets on the bus T.T. T _j, j=1,2 ..., N ₂, N ₂for meeting W _i>=20 people/time data volume, then can be met W _i>=20 people/time time passenger flow counting total number of getting on the bus of obtaining actual total number of getting on the bus total pick-up time the single pick-up time of single and then can obtain at W _i>=20 people/time time passenger flow counting accuracy rate be the mean value of the single pick-up time of single

(4), choose the data needing to correct according to people's logarithmic data of getting on the bus of passenger flow counting, comprise the following steps:

(4.1), according to the mean value of the single pick-up time of single choose the enumeration data needing to correct: by t _j? outside Data correction be if desired the data volume corrected is N ₃, then each enumeration data after correcting is respectively k=1,2 ... N ₃, after correcting, total count value is the accuracy rate obtaining correcting rear passenger flow counting is

(4.2), according to total pick-up time T and reality always get on the bus number U ratio namely choose the enumeration data needing to correct: by t _jdata correction outside (t-t*10%, t+t*10%) is t, and the data volume if desired corrected is N ₄, then each enumeration data after correcting is respectively r=1,2 ..., N ₄, then the total count value after correcting is the accuracy rate obtaining correcting rear passenger flow counting is

(5), by actual passenger flow video data and passenger flow counting data, utilize above-mentioned correcting algorithm, compare the accuracy rate of passenger flow counting before and after correcting, the feasibility of verification algorithm;

(6), due to when 20 people/when time to get on the bus, can there is relatively large deviation in passenger flow counting, can select when enumeration data be 18 people/time or 16 people/time more than situation correct, if according to averaging time correct and utilize * 18 people/time or the time point of 16 people/time a to draw bound * 18 people/time or 16 people/time, with * 18 people/time or 16 people/time, at (t _under, t _on) data outside scope according to T.T. divided by draw the numerical value of correction second.