JP2006163616A - Traffic signal controlling system, traffic signal controlling method and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a traffic signal controlling system, traffic signal controlling method and program for controlling the traffic signal by recognizing information precisely on a subject near the traffic signal. <P>SOLUTION: An action pattern model storing section 13 of the traffic signal controlling system 10 stores a human behavior model 104a, animal behavior model 104b, automobile model 104c and environmental factor model 104d, an analyzing section 11 for analyzing a signal outputted from a pyroelectric sensor 101 arranged near the traffic signal 20, a subject recognition section 14 for recognizing behavior of the subject based on the analytic result of the analyzing section 11 and the action pattern model stored in the action pattern model storing section 13, and a traffic signal controlling section 15 that controls the traffic signal 20 based on the result of recognition. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to a traffic signal control device, a traffic signal control method, and a program for controlling a traffic signal.

Traditionally, roads are provided with pedestrian crossings and traffic lights for crossers to cross the road. The traffic signal is automatically switched at a predetermined cycle. However, because elderly people and children walk relatively slowly, traffic signals sometimes switched from green light to red light before crossing the pedestrian crossing.
There is a traffic signal with a push button as a traffic signal that enables a pedestrian to cross a pedestrian crossing. In a traffic signal with a push button, the switching timing of the traffic signal can be controlled or a sound can be produced by pressing a “push button” attached to the traffic signal.

However, for passers-by, there was annoyance of having to press a push button. Also, especially for visually impaired people, there is an inconvenience that the push buttons must be searched for.
In addition, it has been studied to control a traffic signal by recognizing whether or not the detected object is a crossing person using a sensor (see, for example, Patent Document 1). In the crossing person detection method described in Patent Document 1, the moving speed of the detected object passing through the fixed area is determined by transmitting an ultrasonic wave toward the fixed area, and the moving speed is equal to or lower than the reference speed. For example, the detected object is recognized as a crossing person.
JP 2000-346938 A

However, in the method of recognizing a crosser only by the moving speed, a car in a traffic jam or an animal such as a dog or a cat is also recognized as a crosser. In addition, when a person riding a bicycle stops near a traffic signal, the bicycle may be recognized as a car depending on the speed of the bicycle. As described above, the recognition based on the moving speed cannot accurately determine whether or not the detected object is a crossing person. Furthermore, since it is impossible to determine whether the traverser is a child or an adult, it is impossible to perform control according to the traverser's attribute.
The present invention has been made paying attention to an unsolved problem of the prior art, and is a traffic signal control capable of accurately controlling information on a detected object in the vicinity of a traffic signal and controlling a traffic signal. An object is to provide a device, a traffic signal control method, and a program.

  In order to solve the above-mentioned problem, the invention described in claim 1 is directed to a sensor that outputs a signal according to an operation of a detected object existing in a predetermined area near a traffic signal, and a signal output from the sensor. Based on the motion pattern model storage means for storing the motion pattern model of the detected object generated based on the signal output from the sensor and the motion pattern model stored in the motion pattern model storage means, A traffic signal controller comprising: detected object recognition means for recognizing information related to the detected object; and traffic signal control means for controlling the traffic signal based on a recognition result by the detected object recognition means. I will provide a.

According to this configuration, since the traffic signal control device recognizes information about the detected object based on the motion pattern model, the traffic signal control device can accurately recognize and control the traffic signal based on the accurate recognition result. Is possible.
The invention according to claim 2 is the traffic signal control apparatus according to claim 1, wherein the information on the detected object recognized by the detected object recognition means includes the attribute of the detected object and the detected object. It includes at least one of the operation contents.

  According to this configuration, since the information about the detected object includes at least one of the attribute of the detected object and the operation content of the detected object, for example, whether the detected object is an adult or a child, It is possible to accurately recognize whether the detected object is about to cross a pedestrian crossing or whether it is about to pass through a predetermined area. It becomes possible.

According to a third aspect of the present invention, in the traffic signal control device according to the first or second aspect, the traffic signal control means has recognized that the detected object is a crossing person by the detected object recognition means. In this case, the traffic signal is controlled.
According to this configuration, when it is recognized that the detected object is merely a passerby or an animal, it is possible to control the traffic signal only for the crossing person without controlling the traffic signal.

According to a fourth aspect of the present invention, in the traffic signal control device according to any one of the first to third aspects, the traffic signal control means includes control of switching timing of the traffic signal, output control of sound, It is characterized in that at least one of the control is performed.
According to this configuration, the traffic signal control device selects at least one of the traffic signal switching timing control and the sound output control, and selects a traffic signal control method according to the person to be detected. And the detected person can safely cross the road.

According to a fifth aspect of the present invention, in the traffic signal control apparatus according to any one of the first to fourth aspects, the traffic signal control means has an attribute of the detected object by the detected object recognition means. Delaying the timing at which the lighting state of the traffic signal is switched to a state indicating prohibition of crossing when recognized as at least one of a child, a visually impaired person, and a person in a wheelchair. Features.
According to this configuration, a child, a visually impaired person, and a person on a wheelchair can cross the road slowly and safely.

The invention according to claim 6 is the traffic signal control device according to any one of claims 1 to 5, wherein the traffic signal control means is configured to cause the detected object recognition means to remove the detected object from the ground. When it is recognized that the height is equal to or less than a predetermined value, a timing at which the lighting state of the traffic signal switches to a state indicating a crossing prohibition is delayed.
According to this configuration, the detected object whose height from the ground is a predetermined value or less can cross the road slowly and safely.

The invention according to claim 7 is the traffic signal control device according to any one of claims 1 to 6, wherein the traffic signal control means is configured such that the attribute of the detected object is visually recognized by the detected object recognition means. When it is recognized that the person is a handicapped person, the traffic signal is controlled to be output at a predetermined timing.
According to this configuration, the visually handicapped person can recognize the lighting state of the traffic signal by sound without performing a special operation only by entering the predetermined area.

The invention according to claim 8 is the traffic signal controller according to any one of claims 1 to 7, wherein the detected object recognition means recognizes a moving direction of the detected object, and the traffic signal The control means controls a traffic signal existing in the moving direction recognized by the detected object recognition means.
According to this configuration, it is possible to control the traffic signal in the vicinity of the road that the person to be detected is approaching to cross.

According to a ninth aspect of the present invention, in the traffic signal control apparatus according to any one of the first to eighth aspects, the motion pattern model storage means is provided for an adult, a child, a visually impaired person, and a wheelchair. It is characterized in that at least one operation pattern model is stored among a rider, an animal, a car, and environmental factors.
According to this configuration, by storing many types of motion pattern models and recognizing the detected object, the misrecognition rate can be reduced, and information on the detected person can be accurately recognized. Can do.

According to the tenth aspect of the present invention, there is provided an operation pattern model storing step for storing an operation pattern model obtained by modeling an operation pattern of a detected object existing in a predetermined area near a traffic signal, and an object existing in the predetermined area. A detected object recognition step for recognizing information on the detected object based on a signal output from a sensor that has detected a detected object and the operation pattern model stored in the operation pattern model storage step; and the detected object And a traffic signal control step of controlling the traffic signal based on a recognition result in the body recognition step.
According to this method, since the information about the detected object is recognized based on the motion pattern model, it can be recognized accurately, and the traffic signal can be controlled based on the accurate recognition result.

According to an eleventh aspect of the present invention, an operation pattern model storing step of storing an operation pattern model obtained by modeling an operation pattern of a detected object existing in a predetermined area near a traffic signal in a computer, and the predetermined area A detected object recognition step for recognizing information about the detected object based on a signal output from a sensor that detects an existing detected object and the operation pattern model stored in the operation pattern model storage step; There is provided a program for executing a traffic signal control step for controlling the traffic signal based on a recognition result in the detected object recognition step.
The effect of this program is the same as the effect of the invention described in claim 10.

  According to the present invention, since the traffic signal control device recognizes information about the detected object based on the motion pattern model, the traffic signal control device can accurately perform the recognition and control the traffic signal based on the accurate recognition result. Is possible.

Hereinafter, embodiments of the present invention will be described with reference to the drawings. FIG. 1 shows a hardware configuration of a traffic signal control apparatus 10 according to an embodiment of the present invention. As shown in the figure, the traffic signal control device 10 includes a pyroelectric sensor 101, a Fresnel lens 102, an A / D converter 103, a storage device 104, and a processor 105.
The traffic signal control device 10 is used to detect that a detection object such as a human 40, a car 50, or an animal 60 has entered a predetermined area 30 near the traffic signal 20 shown in FIG. The traffic signal control device 10 is installed in the traffic signal 20 or in the vicinity of the traffic signal 20, and is connected to the traffic signal 20 so that a control signal can be output. In FIG. 2, the traffic signal control device 10 is installed on the signal pole 21 of the traffic signal 20.

  The pyroelectric sensor 101 detects infrared rays emitted from the detection target (human 40, animal 60, car 50, etc.) using the pyroelectric effect, and outputs a signal corresponding to the operation pattern of the detection target. The Fresnel lens 102 is a plastic lens for enlarging the area 30 where infrared rays are detected by the pyroelectric sensor 101. It is possible to replace the Fresnel lens 102 with a mirror. The A / D converter 103 converts the analog signal output from the pyroelectric sensor 101 into a digital signal.

  The processor 105 includes a CPU and a memory, and controls the traffic signal controller 10 as a whole. Various kinds of software are stored in the memory. The various software includes a program for analyzing the signal output from the pyroelectric sensor 101 and a program for recognizing information about the detected object as programs characteristic of the present invention. . The processor 105 implements various functions of the traffic signal control apparatus 10 by executing these programs.

  The storage device 104 is a recording medium for storing data, and includes a hard disk device, a nonvolatile memory, or the like. The storage device 104 stores an operation pattern model obtained by modeling the operation pattern of the detected object. The storage device 104 stores mode selection information as information for selecting a mode of processing performed by the traffic light control device 10. The mode selection information represents one of a model generation mode and a detected object recognition mode. The mode selection information can be input from outside the traffic signal control device 10.

The functional configuration characteristic of the present invention shown in FIG. 3 is realized in the traffic signal control device 10 by the hardware and software included in the traffic signal control device 10 described above.
The analysis unit 11 includes an A / D converter 103 and a function realized by the processor 105 executing a program. The analysis unit 11 analyzes a signal output from the pyroelectric sensor 101 and outputs an analysis result. Specifically, the analysis unit 11 first causes the A / D converter 103 to output an analog signal output from the pyroelectric sensor 101 according to the operation of the detected object that has entered the area 30 near the traffic signal 20. Convert to digital signal. And the analysis part 11 extracts the feature-value data, such as an amplitude and a frequency, by performing a sampling process, a FFT (Fast Fourie Transform) process, etc. with respect to a digital signal, and outputs the said feature-value data as an analysis result To do.

  The operation pattern model generation unit 12 is a function realized by the processor 105 executing a program. The motion pattern model generation unit 12 models the feature amount data as the analysis result of the analysis unit 11 using a predetermined modeling method, and generates a motion pattern model. As the predetermined modeling method, an HMM (Hidden Markov Model) that is a probability model of a time series signal may be used, or a neural network may be used.

In the present embodiment, the motion pattern model generation unit 12 generates a human behavior model 104a, an animal behavior model 104b, an automobile model 104c, and an environmental factor model 104d as motion pattern models.
The human behavior model 104 a is an operation pattern model for the human 40. The motion pattern model of the human 40 is composed of motion pattern models of adults, children, visually-impaired persons, and people on wheelchairs. Furthermore, each motion pattern model includes a motion pattern model of a person who approaches the traffic signal 20 near the area 30 and stops near the traffic signal 20 (that is, a crossing person), and a motion pattern model of a person who passes the area 30. Composed.

  The animal behavior model 104 b is an operation pattern model when an animal 60 such as a dog or a cat stops in the area 30 or passes through the area 30. The automobile model 104c is an operation pattern model by an automobile 50 that stops in front of a pedestrian crossing near the area 30, an automobile 50 that passes the area 30, a headlight of the automobile 50, and the like. The environmental factor model 104d is an operation pattern model based on environmental factors such as light, heat, rain, snow, wind, and tree shaking.

The animal behavior model 104b, the car model 104c, and the environmental factor model 104d include an animal 60 such as a dog and a cat, weather conditions such as wind, rain, and temperature, a headlight of the car 50, and the like. Used to prevent misrecognition.
The motion pattern model storage unit 13 stores the motion pattern models (human behavior model 104a, animal behavior model 104b, automobile model 104c, and environmental factor model 104d) generated by the motion pattern model generation unit 12 in the storage device 104.

The detected object recognition unit 14 recognizes information about the detected object by matching the feature amount data extracted by the analysis unit 11 with each of the motion pattern models stored in the motion pattern model storage unit 13. .
Here, the “information about the detected object” includes the attribute of the detected object, the operation content, the height of the detected object from the ground, the moving direction / moving speed of the detected object, and the like.

The attributes of the detected object include “human (more detailed attributes: adults, children, visually impaired, people in wheelchairs)”, “animals (more detailed attributes: dogs, cats)”, “Automobile” is included. Further, the operation content of the detected object includes “stop” in the vicinity of the traffic signal 20 and “pass” in the area 30.
The detected object recognition unit 14 does not need to control the traffic signal device 20 when the attribute of the detected object is recognized as the animal 60 or the automobile 50 or when the operation content is recognized as “passing”. Therefore, it is rejected (that is, the recognition result is not output to the traffic signal controller 15). If the attribute is “human (adult, child, blind person, or person in a wheelchair)” and the action content is recognized as “stop”, the detected object is Since it is a crossing person and it is necessary to control the traffic signal 20, the detected object recognition unit 14 outputs a recognition result to the traffic signal control unit 15.

The traffic signal controller 15 controls the traffic signal 20 by transmitting a traffic signal control signal to the traffic signal 20 based on the recognition result by the detected object recognition unit 14.
The types of control of the traffic signal 20 include control of switching timing of the traffic signal 20 and output control of sound from the traffic signal 20. Here, the control of the switching timing of the traffic signal 20 is to control the timing at which the three lamps (red lamp, blue lamp, yellow lamp) 22 included in the traffic signal 20 are turned on and off. Further, the “sound” output from the traffic signal 20 may be music or voice guidance.

Examples of the control of the traffic signal 20 performed by the traffic signal controller 15 include the following.
(1) When the detected object recognition unit 14 recognizes that the attribute of the detected object is at least one of a child, a visually impaired person, and a person on a wheelchair, it can walk slowly and safely. As described above, the timing at which the lighting state of the traffic signal 20 is switched to the state indicating the prohibition of crossing is delayed.
Here, the “timing at which the traffic light 20 is switched to a state that indicates crossing prohibition” switches from a state in which the red light of the traffic signal 20 for pedestrians is turned off to a state in which it is lit. Timing (hereinafter referred to as “switching timing to red signal”).

(2) When the detected object recognition unit 14 recognizes that the height of the detected object from the ground is a predetermined value (for example, 1 m) or less, the switching timing to the red signal is delayed. This is because when the height of the detected body from the ground is equal to or less than a predetermined value, the probability that the detected body is a child, an elderly person, or a person on a wheelchair is high.
(3) When the detected object recognition unit 14 recognizes that the attribute of the detected object is a visually impaired person, a sound is transmitted from the traffic signal 20 at a predetermined timing in order to notify the lighting state of the traffic signal 20 by sound. Is output. Here, as the predetermined timing, five seconds before the traffic signal 20 switches to a red signal or the time when the traffic signal 20 switches to a green signal can be considered.
(4) Control the traffic signal 20 present in the moving direction of the detected object. Alternatively, the traffic signal 20 existing in the vicinity of the predicted future moving location is controlled.

[Analysis processing]
Next, the flow of analysis processing performed by the analysis unit 11 of the traffic signal control device 10 will be described with reference to the flowchart shown in FIG.
First, the analysis unit 11 receives a signal output from the pyroelectric sensor 101 (step S101). Since the signal output from the pyroelectric sensor 101 is an analog signal, the analysis unit 11 uses the A / D converter 103 to convert the analog signal into a digital signal. The analysis unit 11 samples the converted signal at a predetermined time interval (for example, 100 ms) (step S102).

  The analysis unit 11 determines whether or not the sampled signal has an output change greater than or equal to a predetermined value, and whether or not a predetermined time has elapsed (step S103). If it is determined that the predetermined time has not elapsed and the output has changed (step S103; Yes), it is detected that the detected object has entered the area 30 and moved. Therefore, the analysis unit 11 stores the signal as data in the storage device 104 (step S104), and repeats the processing from step S101.

On the other hand, when it is determined that there is no output change in the sampled signal, or when a predetermined time has elapsed (step S103; No), the analysis unit 11 stores the data stored in the storage device 104 in step S104. It is determined whether or not there is (step S105). If it is determined that there is stored data (step S105; Yes), the analysis unit 11 divides the data into frames in a predetermined time unit (for example, 1.6 s) in order to analyze the stored data. (Step S106).
Next, the analysis unit 11 performs analysis processing for each divided frame. Specifically, the analysis unit 11 performs an FFT process on the frame, and calculates a spectrum of each harmonic from the result of the FFT process. Further, the analysis unit 11 calculates the average frequency and amplitude of the frame (step S107).

Next, the analysis unit 11 determines whether or not the mode selection information stored in the storage device 104 represents a model generation mode (step S108). When it is determined that the mode selection information represents the model generation mode (step S108; Yes), the analysis unit 11 inputs the analysis result in step S107 to the operation pattern model generation unit 12 (step S109).
On the other hand, when it is determined that the mode selection information represents the detected object recognition mode (step S108; No), the analysis unit 11 inputs the analysis result in step S107 to the detected object recognition unit 14 ( Step S110).

[Operation pattern model generation processing]
Next, the flow of the operation pattern model generation process performed by the operation pattern model generation unit 12 will be described with reference to the flowchart of FIG.
First, the motion pattern model generation unit 12 determines whether an analysis result has been acquired from the analysis unit 11 (step S201). When it is determined that the analysis result is not acquired (step S201; No), the process waits until acquisition. On the other hand, when it is determined that the analysis result has been acquired (step S201; Yes), the motion pattern model generation unit 12 generates a motion pattern model using the HMM based on the acquired analysis result (step S202). .
The motion pattern model generation unit 12 associates the attribute and motion content of the detected object with the generated motion pattern model (step S203). The motion pattern model generation unit 12 stores a motion pattern model in which attributes and motion contents are associated with each other in the motion pattern model storage unit 13 (step S204).

[Detected object recognition processing and traffic signal control processing]
Next, the flow of the detected object recognition process and the traffic signal control process performed by the detected object recognition unit 14 and the traffic signal control unit 15 of the traffic signal control device 10 will be described with reference to the flowchart shown in FIG. First, the detected object recognition unit 14 determines whether an analysis result is acquired from the analysis unit 11 (step S301). When it is determined not to acquire the analysis result (step S301; No), the process waits until acquisition. On the other hand, when it is determined that the analysis result has been acquired (step S301; Yes), the detected object recognition unit 14 reads the motion pattern model from the motion pattern model storage unit 13 (step S302).

Next, the detected object recognition unit 14 performs pattern recognition by matching the read operation pattern model with the acquired analysis result. Specifically, the detected object recognizing unit 14 detects an action pattern model having a state transition sequence having a maximum probability using the Viterbi algorithm (step S303).
The detected object recognition unit 14 recognizes the attribute and operation content of the detected object based on the detection result (step S304). Specifically, the detected object recognizing unit 14 reads out the attribute and the operation content associated with the operation pattern model detected in step S303 from the operation pattern model storage unit 13.

  Next, the to-be-detected body recognition part 14 determines whether it is necessary to control the traffic signal apparatus 20 based on a recognition result (step S305). Specifically, the recognized attribute is “human (adult, child, visually handicapped person or person in wheelchair)”, and the operation content is “stop” near the traffic signal 20. In this case, the detected object recognition unit 14 determines that the traffic signal 20 needs to be controlled (step S305; Yes), and outputs the recognition result to the traffic signal control unit 15.

The traffic signal control unit 15 determines how to control the traffic signal 20 based on the recognition result (attribute and operation content), and outputs a traffic signal control signal representing the content to be controlled to the traffic signal 20. (Step S306).
On the other hand, if the recognized attribute is other than “human”, or the operation content is “pass” (step S305; No), the process ends.
Here, FIG. 7 shows a misrecognition rate of attributes when a plurality of motion pattern models are generated by the above-described motion pattern model generation processing and the above-described detected object recognition processing is performed using the generated motion pattern model. It is a graph.

  (A) is a misrecognition rate when the detected object is recognized as an adult or a child using three motion pattern models of an adult, a child, and a dog. As shown in the figure, when the detected object is an adult, the rate of not being recognized as an adult (misrecognition rate) is 0.277%, and when the detected object is a child, it is not recognized as a child. The ratio (misrecognition rate) was 1.302%.

  (B) is a misrecognition rate when the detected object is recognized as an adult or a child by using four motion pattern models of an adult, a child, a large dog, and a small dog. As shown in the figure, the misrecognition rate when the detected body is an adult is 0.408%, and the erroneous recognition rate when the detected body is a child is 1.615%. In both (a) and (b), the misrecognition rate is 2% or less, indicating a very low misrecognition rate.

  FIG. 8 shows the recognition result of the moving direction of the detected object using the same operation pattern model generation process and the detected object recognition process. Here, the motion pattern model is generated using data obtained when the 17 objects to be detected perform the motion of moving in eight directions shown in (1) to (8) of FIG. 8 five times. Then, as a result of recognizing the motion direction when the detected object passes the area 30 using the generated motion pattern model, the average recognition rate is 73. It was 78.7%, and ignoring the same line error, it was 88.7%. This recognition result is a recognition result when the detected object has completely passed through the area 30, but by applying this, either of the objects before entering and passing through the area 30 or before stopping in the area 30 can be used. In principle, it is also possible to recognize whether the detected object has walked from the direction. Therefore, it is also possible to recognize the direction in which the detected object has entered the area 30 and to predict and recognize which traffic signal device 20 should be controlled based on the recognized direction.

  In addition, since wheelchairs (including manual and electric) are moved by wheels, the movement is close to a constant speed movement. Therefore, when the wheelchair crosses the sensor detection area 30 diffused in a mesh shape by the Fresnel lens 102 at a constant speed, the waveform of the signal output from the pyroelectric sensor 101 is close to a sine wave. This is because, in the sensor detection area 30, positive and negative areas are alternately arranged, and the wheelchair passes through the area 30 at a constant speed. Since the wheelchair does not suddenly enter the middle of the area 30, the waveform of the output signal changes little by little from plus to minus or from minus to plus, alternating waveform pattern of plus and minus Because it becomes.

On the other hand, the movement of adults and children is not as constant-velocity movement as the movement by wheels, and the output waveform fluctuates. Moreover, since an adult is higher from the ground than a person on a wheelchair, the amplitude of the waveform increases.
For visually handicapped persons, the movement of the cane carried by the visually handicapped person is very fine, so the waveform of the output signal has a unique high-frequency pattern.

In summary, wheelchairs and adults can be distinguished by how the waveform fluctuates and the difference in waveform amplitude. That is, the wheelchair has a stable waveform close to a sine wave, and the adult has a height, so the amplitude of the waveform increases.
In addition, a wheelchair and a child can be distinguished by a difference in waveform fluctuation. Wheelchairs and children have little difference in height from the ground because of the small height difference from the ground, but wheelchairs can be distinguished because they have a stable waveform that is closer to a sine wave than children. .
For children and dogs, waveform fluctuations and amplitudes are approximated, but it is possible to distinguish them by collecting models of children and dogs and creating models and increasing the accuracy of the models. Become.

  Further, the height of the detection object from the ground can be identified by the amplitude of the waveform. This is because the closer the distance between the detected object and the pyroelectric sensor 101, the larger the amplitude of the output waveform. For example, the object to be detected is divided into a group whose height from the ground is less than 100 cm, a group that is 100 cm or more and less than 150 cm, and a group that is 150 cm or more, and a model is created for each group to detect the sensor. It is possible to recognize the height of the detected object that has passed through the area 30 from the ground. Note that the number of models to be created is not limited to three. However, if a model is created in which the height of the detected object is divided into units that are too fine (for example, in units of 5 cm), the recognition rate is considered to decrease. .

Next, an operation example in the above configuration will be described with reference to FIGS.
First, signals output from the pyroelectric sensor 101 are collected in advance according to the operations of humans 40, animals 60, automobiles 50, and environmental factors that have entered the area 30. By extracting feature data from these signals and learning the correspondence between the attributes of the detected object and the motion content, a human behavior model 104a, an animal behavior model 104b, an automobile model 104c, Then, the environmental factor model 104d is generated and stored in the storage device 104 of the traffic signal control device 10 in advance.

Next, for example, as shown in FIG. 10, the traffic signal control device 10 (traffic signal control device 10A, traffic signal control device 10B,..., Traffic signal control device 10G) is installed near the pedestrian crossing at the intersection.
The pyroelectric sensor 101 of the traffic signal controller 10 detects infrared rays emitted from the detected object in accordance with the operation of the detected object that has entered the area 30, and outputs an analog signal (process P1 in FIG. 8). The A / D converter 103 converts the signal output from the pyroelectric sensor 101 from analog to digital (process P2). Next, the processor 105 extracts feature amount data such as amplitude and frequency by analyzing the digital signal (process P3). The processor 105 performs pattern recognition by matching the extracted feature amount data with the motion pattern model stored in the storage device 104 in advance (process P4).

For example, as shown in FIG. 10, when it is recognized that the person 40 moves in the direction of the traffic signal 20A and stops in the area 30A that the traffic signal control device 10A can detect. In order to control the traffic signal 20A, the processor 105 outputs a traffic signal control signal to the traffic signal 20A (process P5 in FIG. 8).
Also, as shown in FIG. 10, even when the automobile 50 enters the area 30B, the recognition result of the automobile 50 is rejected by preparing the automobile model 104c, and no traffic signal control signal is output. Thus, the control of the traffic signal 20A does not change.

In the above-described embodiment, by preparing the vehicle model 104c, the operation of the vehicle 50 is prevented from being erroneously recognized as the operation of the human 40. However, the vehicle model 104c is used for controlling the traffic signal device 20. It is also possible to use it actively.
For example, on a narrow road, as shown in FIG. 11, it may be desired to control the traffic signal 20 by detecting the approach of the automobile 50a. In this case, in order to prevent misrecognition by the automobile 50b traveling in a direction perpendicular to the traveling direction of the automobile 50a, a “direction-specific model” of the automobile 50 in this area 30 (the automobile 50a traveling in the direction D1) A model 104c1 and an automobile model 104c2) of the automobile 50b traveling in the direction D2 are prepared in advance. The automobile model 104c1 is used for outputting a traffic signal control signal. On the other hand, the automobile model 104c2 is used to prevent the output of a traffic signal control signal.

That is, the detected object recognition unit 14 matches the feature amount data extracted from the signal output from the pyroelectric sensor 101 with the car model 104c1 (the motion pattern model having the state transition series with the maximum probability is When the vehicle model 104c1), the switching timing of the traffic signal 20 is controlled.
As described above, the traffic signal control device 10 recognizes information about the detected object using the motion pattern model, and controls the traffic signal 20 based on the recognition result. Therefore, the traffic signal control apparatus 10 relates to the detected object near the traffic signal 20. The traffic signal 20 can be controlled by accurately recognizing information. Moreover, since the traffic signal control apparatus 10 can recognize the attribute and operation | movement content of a to-be-detected body in detail, it can control the traffic signal 20 finely according to the attribute and the operation | movement content of a to-be-detected body.

Further, since the pyroelectric sensor 101 operates only when the operation of the detection target is detected, it is low in consumption and economical. And since the hardware which comprises the traffic signal control apparatus 10 is mainly the pyroelectric sensor 101 and the processor (arithmetic apparatus) 105, the traffic signal control apparatus 10 can be produced cheaply and easily. .
In addition, the animal behavior model 104b, the car model 104c, and the environmental factor model 104d are generated, and when matched with these models, the animals 60 such as dogs and cats, wind, rain, temperature, etc. are rejected. It is possible not to react to the weather conditions, the headlight of the automobile 50, and the like.

  In the above-described embodiment, the pyroelectric sensor 101 that detects heat is used as the sensor that detects the detected object. However, the type of sensor is not limited to this, and the sensor is separated from the detected object. Any sensor may be used as long as it is installed at a position and can detect the operation of the object to be detected. For example, an ultrasonic sensor or a microwave sensor may be used. Further, as a sensor for detecting heat other than the pyroelectric sensor 101, a quantum sensor using a photovoltaic effect or a photoconductive effect, or a thermal sensor using a thermoelectromotive effect or a thermal conductive effect may be used. Good.

  In the above-described embodiment, the detected object recognition unit 14 outputs information about the detected person to the traffic signal control unit 15 only when the detected person is a crossing person. Without being limited thereto, for example, the detected object recognition unit 14 outputs information on all detected persons to the traffic signal controller 15, and the traffic signal controller 15 determines traffic signals based on the information on the detected persons. It may be determined whether it is necessary to output a control signal.

  Moreover, in embodiment mentioned above, although the traffic signal control apparatus 10 demonstrated as performing both a motion pattern model production | generation process and a to-be-detected body recognition process, it is not necessary to necessarily perform both processes with one apparatus. . For example, the operation pattern model generation process may be performed by a device other than the traffic signal control device 10, and the generated operation pattern model may be stored in the traffic signal control device 10. And the traffic signal control apparatus 10 may be made to perform only to-be-detected body recognition processing.

In the above-described embodiment, it has been described that the action pattern model storage unit 13 stores the human action model 104a, the animal action model 104b, the car model 104c, and the environmental factor model 104d. If the environment is low, only the human behavior model 104a may be stored in the motion pattern model storage unit 13.
In the above-described embodiment, the traffic signal control device 10 is described as being installed near the traffic signal 20 or the traffic signal 20, but only the pyroelectric sensor 101 and the Fresnel lens 102 are located near the traffic signal 20 or the traffic signal 20. The processor 105 and the like constituting the other traffic signal control device 10 may be installed in another place.

  It is possible to accurately recognize information related to the detected person and to control the traffic signal according to the attribute and the operation content of the detected person.

It is a figure which shows the hardware constitutions of the traffic signal control apparatus which concerns on embodiment of this invention. It is a figure which shows the example of installation of the traffic signal control apparatus which concerns on the embodiment. It is a figure which shows the function structure of the traffic signal control apparatus which concerns on the same embodiment. It is a flowchart which shows the flow of the analysis process which concerns on the same embodiment. It is a flowchart which shows the flow of the operation | movement pattern model production | generation process which concerns on the embodiment. It is a flowchart which shows the flow of the to-be-detected body recognition process and traffic signal control process which concern on the embodiment. (A) is a graph which shows the misrecognition rate at the time of making it recognize whether a to-be-detected body is an adult or a child using three motion pattern models of an adult, a child, and a dog. (B) is a graph which shows the misrecognition rate at the time of making it recognize whether a to-be-detected body is an adult or a child using the four motion pattern models of an adult, a child, a large dog, and a small dog. It is a figure which shows the recognition result of the operation direction which concerns on the same embodiment. It is a figure for demonstrating the operation example which concerns on the same embodiment. It is a figure for demonstrating the operation example in the case of recognizing the human according to the embodiment. It is a figure for demonstrating the operation example in the case of recognizing the motor vehicle based on the embodiment.

Explanation of symbols

10, 10A, 10B, 10C, 10D, 10E, 10F, 10G, 10H Traffic signal control device 11 Analysis unit 12 Operation pattern model generation unit 13 Operation pattern model storage unit 14 Detected object recognition unit 15 Traffic signal control unit 20, 20A Traffic light 21 Signal pole 30, 30A, 30B, 30C Area 40 Human 50, 50a, 50b Car 60 Animal 101 Pyroelectric sensor 102 Fresnel lens 103 A / D converter 104a Human behavior model 104b Animal behavior model 104d Environmental factor model 104c, 104c1, 104c2 Car model 104 Storage device 105 Processor

Claims (11)

A sensor that outputs a signal according to the operation of the detected object existing in a predetermined area near the traffic signal;
An operation pattern model storage means for storing an operation pattern model of the detected object generated based on a signal output from the sensor;
Detected object recognition means for recognizing information on the detected object based on a signal output from the sensor and an action pattern model stored in the action pattern model storage means;
A traffic signal control device comprising: traffic signal control means for controlling the traffic signal based on a recognition result by the detected object recognition means. The traffic according to claim 1, wherein the information on the detected object recognized by the detected object recognition means includes at least one of an attribute of the detected object and an operation content of the detected object. Traffic light control device. 3. The traffic according to claim 1, wherein the traffic signal control unit controls the traffic signal when the detected object is recognized by the detected object recognition unit as a crossing person. 4. Traffic light control device. The traffic signal control means includes:
The traffic signal control device according to any one of claims 1 to 3, wherein at least one of control of switching timing of the traffic signal and output control of sound is performed. The traffic signal control means includes:
When the detected object recognition means recognizes that the attribute of the detected object is at least one of a child, a visually impaired person, and a person on a wheelchair,
The traffic signal control device according to any one of claims 1 to 4, wherein a timing at which the traffic signal is switched to a lighting state indicating a crossing prohibition is delayed. The traffic signal control means includes:
When the detected object recognition means recognizes that the height of the detected object from the ground is below a predetermined value,
The traffic signal control device according to any one of claims 1 to 5, wherein a timing at which the lighting state of the traffic signal switches to a state representing a crossing prohibition is delayed. The traffic signal control means includes:
When the detected object recognition unit recognizes that the attribute of the detected object is a visually impaired person,
The traffic signal control device according to any one of claims 1 to 6, wherein control is performed to output sound from the traffic signal at a predetermined timing. The detected object recognition means recognizes the moving direction of the detected object,
The traffic signal control device according to any one of claims 1 to 7, wherein the traffic signal control unit controls a traffic signal existing in a moving direction recognized by the detected object recognition unit. The motion pattern model storage means stores at least one motion pattern model among an adult, a child, a visually impaired person, a person in a wheelchair, an animal, a car, and an environmental factor. The traffic signal control device according to any one of claims 1 to 8, wherein An operation pattern model storage step for storing an operation pattern model obtained by modeling an operation pattern of a detected object existing in a predetermined area near a traffic signal;
Detected for recognizing information about the detected object based on a signal output from a sensor that detects the detected object existing in the predetermined area and the motion pattern model stored in the motion pattern model storing step Body recognition step,
And a traffic signal control step for controlling the traffic signal based on a recognition result in the detected object recognition step. On the computer,
An operation pattern model storage step for storing an operation pattern model obtained by modeling an operation pattern of a detected object existing in a predetermined area near a traffic signal;
Detected for recognizing information about the detected object based on a signal output from a sensor that detects the detected object existing in the predetermined area and the motion pattern model stored in the motion pattern model storing step Body recognition step,
A traffic signal control step for controlling the traffic signal based on a recognition result in the detected object recognition step.

Images