JP6272158B2 - Human flow analysis system - Google Patents

Description

  The present technology relates to a human flow analysis system. More specifically, the present invention relates to a human flow analysis system that estimates a person's flow line from point cloud data.

  A human detection technique using an apparatus using an infrared laser beam or the like that scans the periphery of the apparatus and measures the position of an object around the apparatus is generally used.

  Patent Document 1 describes a technique for comparing the background image of the intrusion detection area and the monitoring image to determine the presence of the shielding object. This shielding object is determined to be an intruder when a predetermined condition is satisfied, and is determined to be an article (acceptable shielding object) placed by a regular user when the condition is not satisfied. Japanese Patent Laid-Open No. 2004-228561 keeps monitoring with a monitoring camera in mind, but a laser sensor or the like can be used instead of the monitoring camera.

  Patent Document 2 describes a technique for tracking an object (a creature and a vehicle) using a laser sensor device. In particular, it is described that when an object cannot be tracked due to reasons such as passing by another object, the trajectory of the object is estimated by a stochastic method.

JP2012-215983 Special table 2005-513636

  In the prior art, a shield (such as a pillar) fixed in a measurement range is measured in advance, and the movement of a person can be detected based on the difference between the measured objects and a moving person can be passed. However, the actual operational environment may not always exist like a wall or a pillar, but may exist for a very short time such as a passing person (for example, installed temporarily) Signboards, exhibition instructors, panels for special events, etc.). Such a thing shields the measurement by a laser etc. and causes tracking to fail. Such a shield that is not fixed at a certain point but is almost stationary is hereinafter referred to as a static shield. Static shields are generally stationary (but not always), but their existence is not known before the start of operation of the measurement system, and measures cannot be taken in advance. Further, the method for dealing with the passing between moving objects has a problem that it is not always possible to deal with things that are close to the sensor and shield.

  Based on the above, an object of the present invention is to provide a technique for improving the accuracy of human flow measurement in an environment where a static shield exists.

  A representative example of means for solving the problems according to the present invention is exemplified by a human flow measurement system for estimating a human flow line from point cloud data measured by a sensor arranged in an environment and indicating a distance and an angle with respect to the sensor. A calibration unit for detecting a point cloud corresponding to an environmental structure fixed to the environment from the point cloud data, a process for extracting a point cloud corresponding to a person from the point cloud data, and a point corresponding to a person A moving object detection unit that performs a process of estimating a flow line of the same person from the group, and a shielding that is a distance and an angle from the point group data that the shielding object is more frequently present on the sensor side than the environmental structure. A moving object detection unit that extracts a distance and a shielding angle range, and the moving object detection unit has a point within the range of the shielding angle range and larger than the shielding distance when estimating the flow line of the same person. Versus group For persons, and changes the conditions for determining the same person.

  According to the present invention, it is possible to improve the accuracy of human flow measurement in a human flow measurement system in an environment where a static shield exists.

The schematic diagram which shows an example of the system configuration | structure of a human flow analysis system. The schematic diagram which shows an example of the hardware constitutions of each apparatus which comprises a human flow analysis system. The schematic diagram explaining arrangement | positioning of a laser measuring device and the objective of this invention. The schematic diagram explaining arrangement | positioning of a laser measuring device and the objective of this invention. The flowchart which shows a calibration process. The flowchart which shows a human flow measurement process. The flowchart which shows the flow of a process of a mobile body detection part. The schematic diagram which shows the output result of a mobile body candidate extraction process. The schematic diagram which shows the specific example of the same person determination process. Data structure of occlusion area DB. The schematic diagram which shows the conditions of shielding. The schematic diagram which shows the outline | summary of a static shielding object extraction process. Human flow data of human flow database and human flow correspondence table. The schematic diagram which shows the specific example of a human flow analysis. The schematic diagram which shows an example of the system configuration | structure of a human flow analysis system. The flowchart which shows the flow of the totalization process by a human flow totalization apparatus. The schematic diagram which shows the outline | summary of a static shielding object extraction process. The schematic diagram which shows an example of the screen displayed on an analysis result presentation part.

  FIG. 1 shows an example of a system configuration of a human flow analysis system according to the first embodiment. This system includes a laser measurement device (101), a human flow estimation device (110), a human flow totalization device (120), and an analysis result presentation device (130).

  The laser measuring apparatus (101) according to the first embodiment has a function of irradiating an infrared laser while changing an irradiation direction and a function of receiving the reflected light, and the phase of the reflected light of the laser irradiated in each direction. It is possible to measure the distance of an object at each angle from and to transmit the measurement result to the human flow estimation device (110) as point cloud data. The human flow estimation device (110) in the present embodiment is generally a moving object detection unit (111) that finds a moving object from point cloud data obtained from the laser measurement device (101), and is not a building structure. A static shield detection unit (112) for finding a shield that exists regularly, a calibration unit (113) for finding a building structure, etc., and a measurement start instruction reception unit (114) for accepting a user instruction ), And as data, a moving object candidate DB (117) that is data of an object that appears to be a moving object that is intermediate data of the moving object detection unit (111), and a shielding area DB that indicates an area shielded by a static shielding object (115), the structure DB (116) indicating the shape of walls, pillars, etc., and the final output of the moving object detection unit (111), and the human flow DB (118), which is measured moving body data, are stored. It is possible to do. In addition, the human flow totaling device (120) includes a human flow totaling unit (121) that totals the contents of the human flow DB (118) that is the output of the human flow estimation device (110), and a human flow DB (118) that is a duplicate of the human flow DB (118). 122). The analysis result presentation device (130) includes an analysis result display unit (131) that presents a result output from the human flow totalization unit (121) of the human flow estimation device (110).

  FIG. 2 shows a hardware configuration of each device constituting the system according to the first embodiment. The laser measuring device (101) of the first embodiment includes a laser oscillator (201) that emits laser light, a laser receiver (202) that reads reflected light of the laser, laser oscillation, time taken for light reception, and the like. An arithmetic unit (203) that obtains a distance to an object around the device (101), and a NIC (204) that is a network interface for delivering a measurement result and receiving a measurement start / end instruction.

  In addition, the human flow estimation apparatus (110) of the first embodiment uses a processor (205) having computing performance, a DRAM (208) which is a volatile temporary storage area capable of reading and writing at high speed, an HDD, a flash memory, and the like. A storage device (207) that is a permanent storage area, a NIC (206) that is a network interface card for communication, an input device (210), and a monitor (209) that displays analysis results are provided. These sections can be realized by the processor (205) executing the program recorded in the storage area (207). Also, the constituent elements of the human flow aggregation device (120) and the analysis result presentation device (130) that have the same numbers as those of the human flow analysis device (110) have the same functions, and thus detailed description thereof is omitted. .

  The arrangement of the laser measuring device (101) of this embodiment and the object of the present invention will be described with reference to FIGS. 3A and 3B. Two white circles shown in (301) of FIG. 3A indicate the installation location of the laser measuring device (101). The laser transmitter (201) of the laser measuring device (101) emits laser light to the surroundings, and the reflected light is received by the laser receiver (202). The computing unit (203) uses a known calculation method based on the difference in phase between light emission and light reception, etc., and any object is in the range of the distance r and direction θ from the laser measurement device (101) (302). Can be measured.

  Here, when the laser measuring device (101) is indoors, it is shielded by environmental structures (303) such as walls and pillars, so an object such as a moving line of a moving person ( 305) can be detected. However, as shown in FIG. 3B, if there is any static shield (306), the back side (307) cannot be measured. As a result, the flow line (308) and the flow line (309) are measured when the flow line (305) of one person should be measured. If the static shield (306) exists in the environment in advance, it is possible to take measures such as the arrangement of the laser measuring device (101) in consideration of the static shield (306). ) Appears after the start of the operation of the system, for example, if it is movable but almost stationary, such as an exhibitor at the exhibition hall, a guidance display board, a promotional panel, etc. I can't take it. There is also a problem that it is necessary to take countermeasures each time the static shield (306) is removed or moved. The present invention aims to obtain a flow line as close as possible to the flow line (305) even in such a case.

  4A and 4B show the flow of processing in this embodiment. The processing in the present embodiment is mainly composed of two types of calibration processing (FIG. 4A) executed in advance and human flow measurement processing (FIG. 4B) executed when measuring the human flow. The calibration process (FIG. 4A) is a process to be executed when the laser measuring device (101) is installed. The user (401) of this system checks the situation where no one is around the laser measuring device (101), and operates the measurement start instruction receiving unit (114) to execute the calibration process. Then, the laser measuring device (101) receives the instruction from the measurement start instruction receiving unit (114) and measures the surroundings. In addition, the calibration unit (113) receives the instruction from the measurement start instruction receiving unit (114), extracts the shape of the wall, pillar, etc. existing around the laser measurement device (101) as the background, and the structure An environmental structure extraction process (402) for storage in the DB (116) is executed.

  Thereafter, when actually performing the human flow measurement (FIG. 4B), the user (401) inputs to the measurement start instruction receiving unit (114) that the human flow measurement is started. Then, the moving object detection unit (111) compares the measurement result of the laser measuring device (101) with the contents of the structure DB (116) generated by the calibration process (FIG. 4A), The mobile candidate extraction process (403) is performed to extract. This result is stored in the moving object candidate DB (117). Thereafter, the moving object detection unit (111) connects the elements of the moving object candidate DB (117) by the same person determination process (404) to estimate the flow line of the person, and the result is the human flow DB (118). And then sent to the human flow aggregation device (120). A feature of the present invention is a static shielding object extraction process (405) executed by the static shielding object detection unit (112) in parallel with these processes. The static shielding object extraction process (405) detects a semi-stationary thing in the environment from the moving object candidate DB (117), and uses it as a candidate for a static shielding object so that it is well shielded. The area is stored in the shielding area DB (115). When there is data in the shielding area DB (115), the influence of the static shielding object can be reduced by changing the determination method of the same person in the same person determination process (404).

  Details of the calibration process (FIG. 4A) will be described below. The measurement start instruction receiving unit (114) receives an instruction to start calibration by the user (401). As a receiving method, a specific button or the like of the input device (210) may be pressed, or an operation instruction may be issued from the monitor (209). When the measurement start instruction receiving unit (114) receives the instruction, the calibration unit (113) starts the environmental structure extraction process (402). In the environmental structure extraction process (402), the current state of the laser measuring device (101) is measured, and it is a situation where there is no person or object, that is, the structure DB is assumed to be information on the background structure. Record in (116). First, it instructs the laser measuring device (101) to start scanning. As a result, point cloud data is obtained as a scan result from the laser measuring device (101). This point cloud data is data indicating the angle at which the laser is irradiated and the distance to the object at that angle. As a result of measurement at every fixed angle, the distance r to the object and the object are present. The angle θ and the measured time (measurement time) are stored. This point cloud data means a point cloud along a wall or column, and this point cloud is converted into (x, y) = (r ・ cosθ, r ・ sinθ) in the structure DB (115). The obtained xy coordinate value is stored.

  The measurement start instruction receiving unit (114) receives an instruction from the user (401), and sends a measurement start instruction to the laser measurement device (101). Upon receiving this instruction, the laser measuring device (101) periodically performs scanning, and sends the point cloud data as a result of the scanning to the moving object detection unit (111) each time scanning is completed. As shown in FIG. 5, the moving object detection unit (111) performs a moving object candidate extraction process (403) for extracting a moving object candidate from the point cloud, and thereafter, the same person determination process (404) for connecting the moving object candidates. Execute. In the moving object candidate extraction process (403), the background difference process (502) for extracting only the point cloud that is not the background by comparison with the structure DB (115), and the same moving object for the point group other than the background The moving object point group cluster extraction process (503) for extracting the cluster of the measured points is executed, and data indicating the moving object is recorded in the moving object candidate DB (117). In the same person determination process, based on the result of the prediction, the moving body traveling direction prediction process (504) for predicting the traveling direction of each moving body from the movement history of the past moving body stored in the human flow DB (118). The mobile object matching process (505) for associating which mobile object candidate included in the mobile object candidate DB (117) corresponds to which mobile object stored in the human flow DB (118) is executed, and the processing result is Store in DB (118). The data stored in the human flow DB (118) is later sent to the human flow totalization device (120) and totaled. In the process of predicting the moving direction of this moving object and associating it with a moving object candidate, the information of the static shielding object obtained from the shielding object region DB (115) is used, and the processing method is changed on the back side and the front side of the shielding object. This is one of the features of the present application.

  Details of the moving object candidate extraction process (403) will be described below. First, the background difference process (502) separates the moving object candidate from the structure such as a wall or a pillar. For this purpose, first acquire the point cloud data of the laser measuring device (101) and the data of the structure DB (116), and then remove the points close to the points contained in the structure DB (116) as a background. To do. This proximity determination can be performed using a known method, for example, that the Euclidean distance calculated by converting polar coordinates into the Euclidean coordinate system is equal to or less than a threshold value.

  After the background difference process (502), a cluster that is considered to be a point obtained by measuring the same moving object is determined by a moving object point group cluster extraction process (503). First, the point cloud is classified into clusters. For this, a known method such as k-means can be used. Next, it is determined what kind of moving body each cluster measures by a known method, and a feature amount is determined. For this, for example, it is possible to apply a method of fitting assuming that the moving body is circular or elliptical, setting the radius as the size of the moving body, and setting the center as the position of the moving body. Finally, the center position and feature amount (for example, radius) of this moving object are stored in the moving object candidate DB (117).

  FIG. 6 shows an output result of the moving object candidate extraction process (403). Among these, the moving object candidate ID is assigned the serial number of the moving object candidate detected in the moving object candidate extraction process (403), and the time when the data is obtained from the laser measuring device is used as the time.

  Next, details of the same person determination process (404) will be described. The same person determination process (404) includes a moving object traveling direction prediction process (504) and an inter-moving object adaptation process (505). The moving body advancing direction prediction process (504) predicts movement for each data stored in the human flow DB (118) and predicts the position at the measurement time stored in the point cloud data. . As a prediction method, any known method can be used as long as it can predict a position at an arbitrary time from past data, such as obtaining speed and acceleration from data of past several points and predicting that constant acceleration motion is continued. May be used. At this time, data that is larger (older) than the threshold T1 in the time difference from the measurement time is not subject to prediction. Here, T1 means the upper limit time for maintaining the human flow measurement for the moving body passing through the shielded area after the moving body becomes invisible.

  An example is shown in FIG. In this example, the time up to t1, t2, t3, t4, and t5 is the past, and the measurement time of the latest point cloud data is t6. (701) is a time t5, that is, a case where data up to immediately before can be acquired, and is naturally a prediction target. (702) is an example in which data could be acquired before time t4, that is, the data could not be acquired immediately before. If t6-t4 <T1, this (702) is also a prediction target. Similarly, (703) is an example in which data has not been acquired since the data was acquired three times before, and if t6-t3 <T1, it is not subject to prediction. Next, using the prediction result and the shielding area DB (118), a removal process is performed based on the relationship with the shielding area from the predicted moving body.

  FIG. 8 shows the data structure (801) of the shielding area data stored in the shielding area DB (118). The shielding area DB (118) stores, for each laser measuring instrument (101), the time zone during which shielding by a static shielding object is likely to occur, the range of angles at which shielding is performed, and the distance from the laser measuring instrument (101). Has been. By using this, it can be determined whether or not the predicted trajectory of the moving object is shielded by the static shield.

  FIG. 9 illustrates this shielding condition. In the figure, a region surrounded by a broken line (901) is a shielding region. First, as shown in (902), for a moving object whose prediction result does not pass through the shielding area, use another threshold T2 (satisfy T2 <T1) instead of T1, and then predict again. It is determined whether or not it is excluded from the prediction target. For mobile objects whose prediction results pass through the shielding area, such as (903), the moving time is determined in such a way that the larger the difference between the time when entering the shielding area (905) and the time when leaving the shielding area (904) the larger the difference is. Similar processing is performed using the threshold value T ′ (T1> T ′> T2). Note that at this time, since the moving object passes through the farthest area in the shielding area as in (906), even if it exceeds T1, data of T1 or higher is not subject to prediction, so T1 is substantially It is an upper limit. As an example of this process, for the difference dT between the time (905) when entering the shielded area and the time (904) leaving the shielded area, T1 is set as a threshold value when T1 <dT, and T2 is set as a threshold value when T2> dT. Otherwise, the method of using dT itself as a threshold can be applied.

  As described above, the human flow measurement system according to the present embodiment is a human flow measurement system (110) that estimates a person's flow line from point cloud data that is measured by a sensor (101) arranged in the environment and indicates a distance and an angle with respect to the sensor. A calibration unit (113) for detecting a point cloud corresponding to an environmental structure fixed to the environment from the point cloud data, a process for extracting a point cloud corresponding to a person from the point cloud data, and a person From the point cloud corresponding to the moving object detection unit (111) that performs the process of estimating the flow line of the same person, and from the point cloud data, the frequency of the presence of a shield on the sensor side relative to the environmental structure And a shielding object detection unit (112) that extracts a shielding distance and a shielding angle region that are high distances and angles, and the moving object detection unit is configured to estimate a flow line of the same person when the range of the shielding angle region is estimated. For point clouds with a distance greater than the shielding distance For the responding person, the condition (T ′) for determining the same person is changed. With this configuration, the human flow measurement system according to the present embodiment can estimate a flow line corresponding to a static shield that was not known at the start of operation of the system, and can further improve the accuracy of human flow measurement. Is possible. More specifically, the change of the condition means a threshold value (T ′) of the time difference between the point groups when determining the person corresponding to the point group within the shielding angle range and having a distance larger than the shielding distance as the same person. ) Should be larger than the threshold (T2) for other persons. Further, the threshold value for a person with a larger distance to the sensor may be increased. Hereinafter, the result of this prediction will be referred to as a human flow prediction position.

  The static shielding object extraction process (405) for constructing the shielding area DB (115) will be described with reference to FIG. In the static shielding object extraction process (405), first, the moving object candidate data (601) is acquired from the moving object candidate DB (117). Here, data may be acquired at appropriate time zones (however, with a sufficiently long time width compared to the measurement target such as 1 hour), or all data may be acquired without limiting the time zone. . This data is illustrated as (1000). In this figure, the black circle (1001) indicates the moving object candidate data (601) around the laser measuring instrument (101). If there is no shielding object, it is considered that the moving object candidate data (601) is generated evenly on the measurement line of the laser measuring device (101), for example, (1002). On the other hand, in the measurement line when there is a static shield, for example (1003), the moving object candidate data (601) should be measured extremely biased around the static shield (1004). Therefore, the occurrence frequency of the moving object candidate data (601) for each distance r from the laser measuring device (101) at each angle, that is, (number of moving object candidate data (601) of a certain r) / (on the line of the angle The total number of moving object candidate data (601)). (1010) shows an example of a graph in which the horizontal axis is r and the vertical axis is the occurrence frequency. If the static shield is at an angle on the line, the frequency of occurrence increases around the static shield as in (1011). On the other hand, when the angle is such that there is no static shield on the line, a flat graph is obtained as shown in (1012). Therefore, the angle and position of the static shield can be specified by, for example, searching for the most frequently occurring r and determining that there is a static shield if it exceeds the threshold. . This process is performed while changing the angle at a predetermined step size, and one or a plurality of sections having a static shielding object at the same position are continuously found. Laser measuring instrument ID that identifies the current laser measuring device (101) with the starting angle as the starting angle, the ending angle as the ending angle, and the distance with the static shield as the distance. And the data stored in the data (801) of the shielding area DB (115) can be constructed using the period (which is the shielding time zone) when the moving object candidate data (601) is acquired.

  Returning to FIG. 5, the inter-moving body adaptation process (505) will be described. In the inter-moving body adaptation process (505), the moving body candidate DB (117) and the human flow DB (118) are associated with each other using the position predicted by the moving body advancing direction prediction process (504). This association can be performed by, for example, a method of associating a human current predicted position at the measured time of the moving object candidate DB (117) with a distance between the moving object candidate DB (117) that is equal to or less than a threshold value. At this time, when two human flow prediction positions correspond to one mobile object candidate DB (117) (distance is equal to or less than the threshold value), one to one for all the mobile object candidate DBs (117). It is possible to apply a method such as trying all combinations that associate human flow predicted positions and using the one having the smallest sum of distances in each combination. When the combination is obtained, the human flow data (1101) of the human flow DB (118) shown in FIG. 11 can be updated. For the moving object candidate DB (117) for which the combination destination is determined, the human ID that is an identifier for uniquely identifying a person included in the human flow DB (118), the measurement time and coordinates x of the moving object candidate DB (117) , y can be used to add data to the human flow DB (118) as the position of a new person. If there is a moving object candidate DB (117) (hereinafter referred to as an independent moving object candidate) that is not associated with any predicted human flow position, a new human ID number is assigned to the human flow DB. Register at (118). Thereby, the human flow data DB (1101) can be updated.

  Alternatively, it is also possible to apply a method of separately outputting only combination information as attached information for a moving object that passes behind a static shield while obtaining all combinations in the same manner as described above. In that case, when calculating the predicted human flow position, T2 is used as a threshold value to obtain a one-to-one combination by the above-mentioned method. Output as correspondence table (1102). At this time, a new human ID is assigned to each independent mobile candidate, and is added to the human flow DB (118) as human flow data (1101). In the person correspondence table (1102), this new person ID, the corresponding person ID of the predicted human flow position, and the fitness determined from the sum of the distances (assuming that the greater the distance, the less the fitness, for example, the distance Can be stored). As a result, it is possible to provide supplementary information when analyzing data in the human flow DB.

  The human flow tabulation device (120) uses the information of this human flow DB (1101), for example, tabulates the number of people who passed a certain passage, the number of people who turned right at the intersection, etc., and the result is analyzed by the analysis result presentation device (130) The result of the human flow analysis is presented to the user (401) by displaying it on the result display unit (131).

  If there is a person correspondence table (1102), this analysis can be corrected. As a specific example of this correction, an example in which the ratio of the number of people who go straight at the intersection and the number of people who turn right is obtained is shown in FIG. In this example, it is determined whether a person coming from the left passage in the figure goes straight at this intersection or turns right and left, and obtains the ratio. For this purpose, a method of drawing a line near the entrance of the intersection and examining the intersection of the line and the human flow can be applied. For example, a person who passes the entrance line (1201) entering the intersection from the left passage and passes the upper passage line (1202) is a left turn. A person passing through the line (1201) and passing through the line (1203) goes straight, and a person passing through the line (1201) and passing through the line (1204) turns right. By counting the number of people flow lines where such intersections occur, the ratio of straight turn to left and right can be calculated. However, considering the case where there is a static shield (1202) to block the intersection, a person who turns left does not need to pass behind this shield, but a person who turns right or goes straight behind the shield. Is likely to pass. There is a high probability of measurement failure behind the shielding object, and in extreme cases, tracking may always fail when passing through the shielding area. In this example, if the tracking fails for all but the left turn, an incorrect result of 100% left turn is calculated.

  On the other hand, if the prediction in the shielding area by the static shielding object operates correctly and the same person determination process (404) functions correctly, the correspondence between people can be corrected. However, for example, if the person from the left (1205) from the left in the figure and (1206) from the top enter the intersection almost simultaneously and exit from the intersection on the route (1207) (1208) almost simultaneously, the correspondence May not always be correct. However, since the statistics for a large number of people are required in the present embodiment, the process of calculating the number of straight turns can be substituted by obtaining the expected value of the number. Therefore, the number of cases can be sorted using the ratio of the fitness of the person correspondence table (1102). For example, if the fitness of (1205) and (1207) is 2, and the fitness of (1205) and (1208) is 8, this person is counted as 0.2 who goes straight and 0.8 who turns right. Thereby, even when it cannot be correctly totaled by a static shielding object, the total result as a suboptimal measure can be obtained.

  According to the present embodiment, it is possible to aggregate human flow with reduced influence even on a static shield whose existence is not known until the actual operation is started.

  In the second embodiment, the human flow counting device (120) in the first embodiment is configured to detect a static shield. FIG. 13 shows an example of a configuration diagram in the embodiment of the second embodiment.

  Similar to the first embodiment, this system includes a laser measurement device (101), a human flow estimation device (110), a human flow totaling device (1300), and an analysis result presentation device (1310). The laser measuring device (101) is the same as that of the first embodiment. The human flow estimation device (110) is the same as that of the first embodiment, except that the static shielding object detection unit (112) and the shielding region DB (115) are not included. In addition to the human flow totaling unit (121) and the human flow DB (122) similar to those in the first embodiment, the human flow totaling device (1300) includes a static shield detection unit (1301) for newly finding a static shield, It has a human flow correction unit (1302) that corrects the human flow DB (122) using information on the target shield, and a shielding region DB (1303) that indicates a region shielded by the static shield. In addition to the analysis result display unit (131) similar to that in the first embodiment, the analysis result presentation device (130) is used to set whether or not to correct the influence of static shielding on the human flow aggregation device (1300). The correction condition setting unit (1311) is provided.

  FIG. 14 shows a flow in which the human flow totalization apparatus (1300) performs totalization in consideration of the shielding area by the static shielding object. Data sent from the human flow estimation device (110) is stored in the human flow DB (122). The area shielded by the static shield is extracted by the static shield extraction process (1402) using the information stored in the human flow DB (122), and stored in the shield area DB (1303). Since this process is performed using past data included in the human flow DB (122), it may be performed in advance. The human flow correction unit (1302) corrects the human flow by the human flow correction processing (1401) using the shielding area DB (1303), and sends the result to the human flow totalization unit (121). The human flow totalization unit (121) counts the number of people and sends the result to the analysis result presentation device (1310). The analysis result presentation device (1310) includes a correction condition setting unit (1311) composed of arbitrary input / output capable parts, and executes human flow correction processing (1401) according to a user instruction. It is also possible to switch whether to analyze the original human flow DB (122).

  The static shielding object extraction process (1402) will be described with reference to FIG. In the occluded region extraction process, first, the human flow data (same as (1101)) is acquired from the human flow DB (122). Among them, for each person ID data, the first appearing point (hereinafter referred to as a flow line appearance point) and the disappearing point (hereinafter referred to as a flow line disappearance point) are extracted. This is illustrated in (1500). The black circle (1501) in the figure means the flow line appearance point and the flow line vanishing point. In this figure, it is assumed that there is a static shielding object (1502) at the position of the dotted line circle (1502) in the figure, and the lines (1504) and (1505) are the boundaries between the shielding area and the other areas. As shown in the figure, the number of black circles (1501) inevitably increases because the human flow line is interrupted or restored near the boundary of the shielding area. Therefore, the rate of occurrence of the flow line appearance point and the flow line vanishing point at a fixed step angle, that is, `` the flow line appearance point and the flow line vanishing point at the angle are determined by Find the amount divided by the number. The result is shown in graph (1510). In this graph, it is assumed that there are boundaries of the shielding area at (1511) and (1512) where the number of flow line appearance points and flow line disappearance points is large, and the angle that is equal to or greater than the threshold value is the boundary.

  Also, the ratio of the flow line appearance point and the flow line vanishing point for each distance r from the laser measuring device (101) at that angle, that is, (number of flow line appearance points and flow line vanishing points of a certain r) / (the angle concerned) (Total number of flow line appearance points and flow line vanishing points on the line). (1520) shows an example of a graph in which the horizontal axis is r and the vertical axis is the generation ratio. Since a lot of flow line appearance points and flow line vanishing points should appear at a distance (1521) farther than the static shielding object, it can be said that the area after r that is equal to or greater than the threshold is the shielding area.

  As a result, it is possible to specify that a range equal to or greater than a specific r on a specific θ line is a shielding region by a static shielding object. For these boundary lines, take adjacent combinations, check the number of human flow DB (122) in the area between them, and if the number of cases above the threshold is found, that area is identified as a shielded area. The

  Next, the human flow correction process (1401) will be described. In the human flow correction process (1401), first, the human flow data (1101) of the person ID without the latest data is extracted, and the same method as the moving body traveling direction prediction process (504) of the first embodiment is applied to them. The movement of each person is predicted using a plurality of threshold values T1, T2, and dT (the predicted result of the position at the latest time is taken as the predicted human flow position). If there is a flow line appearance point at a distance below the threshold from the predicted human flow position, it can be inferred that the person ID and the human ID at the predicted human flow position are actually the same person. Therefore, thereafter, the process of replacing the person ID at the flow line appearance point with the person ID at the person flow prediction position is performed. If there are a plurality of flow line appearance points at a distance equal to or less than the threshold, a more appropriate result can be obtained by obtaining a combination that minimizes the sum of the distances, as in the first embodiment.

  FIG. 16 shows an example of a screen displayed on the analysis result display unit (131) of the analysis result presentation device (1310). In this example, the analysis result display unit (131) receives the ratio of straight-ahead and left-right turns of the intersection as the totaling result sent from the human flow totaling unit (1300), and the numerical value is shown in the balloon (1601). In this balloon, the result of correcting the influence of the static shield and the result before correction are displayed at the same time. As a result, information that can be used as a reference for determining whether or not to confirm the site can be presented to the user, for example, when extreme correction is applied. If correction processing is not required, it is possible to specify whether or not correction is necessary in (1602). This setting is received by the correction condition setting unit (1311) and simultaneously sent to the analysis condition setting unit (111) of the human flow totalization apparatus (1300). Based on this result, it is possible to switch whether or not the correction by the human flow correction unit (1302) is performed.

  According to the second embodiment, the function of the present invention is simply implemented in the human flow totalization apparatus (1300) that receives the output of the known laser measurement apparatus (101) and the human flow estimation apparatus (110) without any modification. The effect of reducing the influence of the static shield can be obtained.

  101: Laser measurement device, 110: Human flow estimation device, 111: Moving object detection unit, 112: Static shielding object detection unit, 113: Calibration unit, 114: Measurement start instruction reception unit, 115: Shielding region DB, 116: Structure DB, 117: moving object candidate data, 118: human flow DB, 120: human flow totaling device, 121: human flow totaling unit, 122: human flow totaling unit, 130: analysis result presentation device, 131: analysis result display unit, 201: laser Oscillator, 202: Laser receiver, 203: Calculator, 204: NIC, 205: Processor, 206: NIC, 207: Storage device, 208: DRAM, 209: Monitor, 210: Input device, 301: Installation location, 302: Range, 303: Environmental structure, 304: Area, 305: Flow line, 306: Static shield, 307: Back side, 308: Flow line, 309: Flow line, 401: User, 402: Environmental structure extraction process 403: Moving object candidate extraction processing 404: Same person determination processing 405: Static shielding object extraction processing 502: Background difference processing 503: Moving object point group cluster Outgoing processing, 504: Mobile body travel direction prediction processing, 505: Inter-mobile body adaptation processing, 601: Moving body candidate data, 701: Data up to time t5 can be acquired, 702: Data up to time t4 can be acquired 703: When data up to time t3 can be acquired, 801: Data structure of the occluded area DB, 901: Occluded area, 902: Mobile object whose prediction result does not pass through the occluded area, 903: Prediction Mobile object whose result passes through the shielded area, 904: Time to exit from the shielded area, 905: Time to enter the shielded area, 906: Mobile object through the farthest area in the shielded area, 1000: Mobile object candidate data, 1001 : Moving object candidate data, 1002: Measurement line, 1003: Measurement line, 1004: Around the static shield, 1010: Graph, 1011: Static angle on the line, 1012: Static on the line 1101: Human flow data, 1102: Person correspondence table, 1201-1204: Line, 1205 1208: Traffic line, 1300: Human flow analysis device, 1301: Static shielding object detection unit, 1302: Human flow correction unit, 1303: Shielding area DB, 1310: Analysis result display device, 1311: Correction condition setting unit, 1401: Human flow correction Processing, 1402: Static shield extraction processing, 1500: Human flow data, 1501: Flow line appearance point and flow line disappearance point, 1502: Static shield, 1504-1505: Boundary, 1510: Graph, 1511-1512: Boundary , 1520: Graph, 1521: Shielding area, 1601: Speech balloon, 1602: Specification of necessity of correction.

Claims (5)

A human flow measurement system for estimating a flow line of a person from point cloud data measured by a sensor disposed in an environment and indicating a distance and an angle with respect to the sensor,
A calibration unit for detecting a point cloud corresponding to an environmental structure fixed to the environment from the point cloud data;
A moving object detection unit that executes a process of extracting a point cloud corresponding to a person from the point cloud data, and a process of estimating a flow line of the same person from the point cloud corresponding to the person;
A shielding object detecting unit that extracts a shielding distance and a shielding angle region, which are distances and angles at which the shielding object is present more frequently on the sensor side than the environmental structure from the point cloud data,
When the moving object detection unit estimates the flow line of the same person, a condition corresponding to a point group within a range of the shielding angle range and a distance greater than the shielding distance is determined to be the same person. Human flow measurement system characterized by changing In claim 1,
The moving object detection unit sets a threshold value of a time difference between the point groups when determining a person corresponding to a point group within a range of the shielding angle range and a distance greater than the shielding distance as the same person to another person. A human flow measurement system characterized in that it is larger than the threshold value for. In claim 2,
The moving object detection unit increases a threshold value for a person having a larger distance to the sensor among persons corresponding to a point group within the shielding angle range and having a distance larger than the shielding distance. Characteristic human flow measurement system. In claim 1,
The moving body detection unit
When estimating the flow line of the same person and the first ID given to the person corresponding to the point group that is within the shielding angle range and larger than the shielding distance, the person ID is associated with any person. A human flow measurement system that outputs a correspondence table indicating combinations that can be associated with a second ID assigned to a person who has not been made. In claim 4,
In the correspondence table, the fitness level for each of the combinations is further indicated.

Images