JP5513036B2 - Detection device - Google Patents

Description

  The present invention relates to a detection device that detects a detected object that enters a predetermined area based on fluctuations in the output of a thermal radiation sensor such as infrared rays, and in particular, determines whether or not the detected object is a person. It is related with the detection apparatus which can be.

  A passive infrared sensor is a sensor that detects energy of heat rays (far infrared rays, hereinafter simply referred to as infrared rays) using an infrared detection element. The infrared detection element uses a pyroelectric element or the like using a condensing optical system, and condenses infrared rays radiated from an object in an area to be detected (detection area). Then, the fluctuation amount is detected, and when the fluctuation amount exceeds a predetermined threshold, it is determined that a person or an animal has entered the detection area. Such passive infrared sensors are currently used to open and close automatic doors and to detect intruders in security systems.

However, since the passive infrared sensor detects infrared rays, it cannot determine whether an infrared ray is emitted from an animal or a person. In addition, misinformation may occur due to disturbance such as sunlight.
In order to solve the above-described problems, various passive infrared sensors capable of detecting only a human body have been proposed. One of the passive infrared rays that detects only the human body is equipped with two sets of units composed of a combination of an optical system and a light-receiving element. Two sets of units are placed up and down in the detection aerial to detect infrared fluctuations. There is one that determines whether or not it has occurred at a scale larger than a predetermined range in the area.

  That is, the passive infrared sensor having two sets of units is a human body detection signal indicating that a human body is detected only when a change in infrared rays exceeding a threshold is detected by both of the two sets of upper and lower units. Is output. For this reason, it is possible to prevent a heat source, direct sunlight, and a small animal that has entered the detection area farther than the detection area from being erroneously determined as a human body, and to improve the detection accuracy of the infrared detection system. In addition, as a prior art of such a passive infrared sensor, there exists what was described in patent document 1, for example.

Patent No. 3086406

FIG. 6 is a diagram for explaining a passive infrared detection device (hereinafter simply referred to as a detection device) described in Patent Document 1. In this description, the wording described in Patent Document 1 is used as it is.
The detection device shown in FIG. 6 is attached to a height position on the wall surface 20 corresponding to an adult's waist. In the housing 21 attached to the wall surface 20, a first sensor unit 5 composed of the light receiving element 1 and the optical system 3 and a second sensor unit 6 composed of the light receiving element 2 and the optical system 4 are stored vertically. Has been.

The sensor units 5 and 6 individually collect infrared energy radiated from different regions that are vertically separated from each other in a predetermined zone Z to be warned of illegal intruders indicated by two-dot chain lines in the figure. It is arranged in the direction. That is, the upper first sensor unit 5 is arranged so that the light receiving direction is substantially horizontal toward the upper half of the human body H to be detected, and the first sensor unit 5 does not reach the ground G in the upper space in the alert zone Z. Detection area A1 is set. In addition, the lower second sensor unit 6 is disposed with the light receiving direction directed to the ground G that is separated from the wall surface 20 by a predetermined detection distance L, and the second detection area A2 is located in the lower space in the alert zone Z. Is set.
Here, the detection distance L is a distance from the wall surface 20 to a position where the center line C of the second detection area A2 set by the second sensor unit 6 and the ground G intersect. From the detection distance L, the height of the first detection area A1, and the widths of the first and second detection areas A1 and A2 (dimensions in the direction perpendicular to the drawing sheet), a warning zone Z is determined.

  The electric signals output from the light receiving elements 1 and 2 are amplified by the amplifier circuits 8 and 9, respectively, and the signal intensity of the amplified signals, that is, the fluctuation amount of the infrared light beam, is constantly monitored by the level detection circuits 10 and 11 including a comparison circuit. Has been. The level detection circuits 10 and 11 constantly compare the signal level of the input electric signal with a predetermined detection level set in the detection level setting units 12 and 13. When the level detection circuits 10 and 11 detect the fluctuation of the electric signal exceeding the detection level, a high level detection signal is output. That is, the level detection circuits 10 and 11 output a binary signal composed of a high level detection signal and a low level non-detection signal. The human body detection circuit 14 composed of an AND circuit outputs a human body detection signal a from the human body detection circuit 14 when a high level detection signal is output from both level detection circuits 10 and 11. The alarm signal generation circuit 17 is activated by the input of the human body detection signal a and generates an alarm signal b for operating an alarm generation device (not shown) such as an illumination lamp, a buzzer or a siren.

  The conventional detection device outputs the human body detection signal a only when the signal levels of the electrical signals output from both the light receiving elements 1 and 2 simultaneously exceed the detection level. For example, there is a high temperature generation source such as a boiler far from the first detection area A1, and an object passes in front of the high temperature generation source, or a car or a truck travels on a road far from the first detection area A1. When passing, the infrared energy radiated from these enters the light receiving element 1 through the lens 3 of the upper first sensor unit 5. The light receiving element 1 outputs an electric signal having a signal level corresponding to a large variation in the amount of incident infrared energy, and a detection signal is output from the level detection circuit 10 that detects this.

  However, since the light receiving direction of the lens 4 of the lower second sensor unit 6 is directed downward, the above-described infrared energy hardly enters. Therefore, the signal level of the electrical signal from the lower light receiving element 2 hardly changes, and the output of the level detection circuit 11 is held at a low level. As a result, the human body detection circuit 14 does not erroneously output the human body detection signal a.

In addition, when sunlight is directly irradiated in the daytime, the signal level of the electric signal output from the light receiving element 1 of the upper first sensor unit 5 is changed so as to be high as described above. a is not output.
On the other hand, when a small animal M such as a dog or a cat enters the warning area Z, infrared energy emitted from the small animal M is incident on the light receiving element 2 via the lens 4 of the second sensor unit 6 below. Almost no light is incident on the upper first sensor unit 5 arranged with the light receiving direction substantially horizontal. Accordingly, since only the electric signal output from the lower light receiving element 2 exceeds a predetermined level, the human body detection signal a is not erroneously output in this case.

  On the other hand, when the human body H enters the warning zone Z, both the detection areas A1 and A2 are shielded, so that the infrared energy radiated from the human body H is surely connected to the lenses 3 and 4 of the upper and lower sensor units 5 and 6. Through the light receiving elements 1 and 2. For this reason, the electrical signals of the light receiving elements 1 and 2 both exceed a predetermined level, and both level detection circuits 10 and 11 output detection signals. As a result, it is possible to reliably detect the human body H that has entered the alert zone Z set in both the detection areas A1 and A2, and to output the human body detection signal a.

However, in the above-described prior art, as shown in FIG. 7, even when the human body is short, the infrared light incident on the upper sensor unit 5 is not sufficient when the height is low. The signal level output by the signal becomes lower than a predetermined level. In such a case, although there is a human body in the detection area, an alarm signal cannot be output.
In addition, as shown in FIG. 8, when a small animal or the like passes near the sensor, it may straddle the upper and lower detection areas at the same time, and the amount of infrared light incident on both sensor units becomes a sufficient amount. Although it is not, it may be detected and cause false alarms.
The present invention has been made in view of such a point, and accurately detects only a human body, and does not generate false alarms due to disturbances or other objects that emit infrared rays (detected body). An object is to provide an apparatus.

In order to solve the above problems, the detection device according to claim 1 of the present invention detects infrared rays in different detection areas, and outputs an electrical signal corresponding to the detected amount of variation in each infrared ray. (Infrared sensors 101a and 101b: FIG. 1) Thermal radiation detection means (twin mirror type infrared sensor 101: FIG. 1) provided with at least two, and information concerning the frequency of the electrical signal according to the amount of infrared fluctuation Model storage means (model storage unit 106: FIG. 1) for storing a plurality of probability models created based on the feature quantity, and at least two of the thermal radiation detection means (twin mirror type infrared sensor 101: FIG. 1). First information on the frequency of the first electrical signal output by the first infrared sensor of the two infrared sensors, and the at least two Feature extraction means for extracting a second feature quantity and the second information relating to the frequency were ligated calculated independently of the electrical signal output by the second infrared sensor of the infrared sensor (feature amount extraction Unit 103), the feature quantity extracted by the feature quantity extraction unit (feature quantity extraction unit 103: FIG. 1), and the probability model stored by the model storage unit (model storage unit 106: FIG. 1) Probability model detection means (recognition processing unit 104: FIG. 1) for detecting a probability model corresponding to a feature quantity from the plurality of probability models based on likelihood, and the probability model detection means (recognition processing unit 104: A detected object recognition means (recognition processing unit 104: FIG. 1) for recognizing the detected object that has entered the detection area based on the probability model detected by FIG. That.

  The detection device according to claim 2 of the present invention is the detection device according to claim 1, wherein the thermal radiation detection means is an infrared detection element (for example, the pyroelectric element 301a shown in FIG. 301b, 301c, 301d) are provided in a single housing (for example, the case 300 shown in FIG. 3), and a plurality of the electrical signals (for example, sensor output 1 and sensor output 2 shown in FIG. 3) are simultaneously provided from the housing. Is output.

According to a third aspect of the present invention, there is provided the detection device according to the first or second aspect, wherein the feature amount extraction means (feature amount extraction unit 103: FIG. 1) includes the first electric signal and the The second electrical signal is divided at predetermined time intervals to generate a plurality of frames, and a first feature amount representing a component ratio of each frequency with respect to a plurality of frequencies included in each frame is set as the first information and It is calculated as the second information .
According to a fourth aspect of the present invention, in the invention according to the third aspect, the component ratio is represented by a logarithm in the first feature amount.

According to a fifth aspect of the present invention, there is provided the detection device according to any one of the first to fourth aspects, wherein the feature amount extraction unit (feature amount extraction unit 103: FIG. 1) includes the first electrical signal and The second electrical signal is divided at predetermined time intervals to generate a plurality of frames, and a first frame that is one of the plurality of frames and a time immediately before the first frame or A second feature amount representing a component ratio of each frequency with the second frame located immediately after is calculated as the first information and the second information , respectively.

Sensing apparatus according to claim 6 is the invention according to claim 5, Oite before Symbol second feature quantity, the component ratio is characterized by being represented by the logarithm.
According to a seventh aspect of the present invention, in the invention according to any one of the first to sixth aspects, the feature amount extraction means (feature amount extraction unit 103: FIG. 1) includes the first electrical signal and The second electrical signal is divided at predetermined time intervals to generate a plurality of frames, and a third feature amount representing a correlation between the first electrical signal and the second electrical signal for each frame. It is characterized by extracting.

According to the detection apparatus of Claim 1, the to-be-detected body which entered into each different area can be detected with two infrared sensors. Moreover, since the information concerning the frequency of the electric signal output by the infrared sensor is extracted as the feature amount, the difference concerning the frequency of the electric signal can be detected. In addition, a probability model corresponding to the feature amount can be stored in advance, and the stored feature amount can be compared with the probability model. Furthermore, a probability model having the highest probability that this feature amount is extracted is detected from among the probability models, and a detected object that matches the detected probability model is recognized as a detected object that has entered the detection area. Can do.

According to the detection device of the second aspect, an infrared sensor having a plurality of pyroelectric elements can be used as the thermal radiation detection means of the present invention.
According to the detection device of the third aspect, the probability model can be stored in the detection device. For this reason, it is possible to create a probabilistic model according to the environment, position, etc. in which the detection device is placed in advance, and to detect the detected object using the created probabilistic model.

According to the detection device of the fourth aspect of the present invention, the electricity indicating the change in the infrared rays in the detection area based on the first feature amount representing the component ratio of each frequency to the plurality of frequencies included in each of the plurality of frames. A change in frequency of the signal can be detected. For this reason, it is possible to detect a change in the infrared rays in the detection area regardless of the magnitude of the change in the amplitude of the infrared rays.
Therefore, even if the amount of change in infrared rays by the detected object is small, it can be detected that the detected object has entered the detection area.

  According to the detection device of claim 5, each frequency of a first frame that is one frame among a plurality of frames and a second frame that is positioned immediately before or immediately after the first frame in time. A second feature amount representing the component ratio can be extracted. For this reason, since the frequency change of the electrical signal which shows the time change of the infrared rays in a detection area can be detected, the to-be-detected body in a detection area can be detected irrespective of the magnitude | size of the change of the amplitude of an infrared ray. it can. Therefore, even if the amount of change in infrared rays by the detected object is small, the movement of the detected object that has entered the detection area can be detected.

According to the detection device of the sixth aspect, since the component ratio is represented by logarithm in at least one of the first feature value and the second feature value, the feature value is always obtained as an absolute value and used for the subsequent control. Can be easy to do.
According to the detection device of the seventh aspect, it is possible to extract the third feature amount that represents the correlation between the two electric signals respectively detected by the at least two infrared sensors for each frame. For this reason, it can be judged whether a to-be-detected body has a magnitude | size detected simultaneously by both among two infrared sensors. Therefore, it is possible to determine whether or not the detected object is a person based on the approximate size of the detected object.

It is a figure for demonstrating the detection apparatus of one Embodiment of this invention. It is a figure for demonstrating extraction of the feature-value performed by the feature-value extraction part shown in FIG. It is a figure for demonstrating the pyroelectric sensor of the modification of one Embodiment of this invention. It is a figure for demonstrating the result of Experimental example 1 of the detection apparatus of this invention. It is a figure for demonstrating the result of Experimental example 2 of the detection apparatus of this invention. It is a figure for demonstrating the passive infrared detection apparatus described in patent document 1. FIG. It is another figure for demonstrating the passive type infrared detection apparatus described in patent document 1. FIG. It is another figure for demonstrating the passive type infrared detection apparatus described in patent document 1. FIG.

Hereinafter, a detection apparatus according to an embodiment of the present invention will be described with reference to the drawings.
Overall Configuration FIG. 1 is a diagram for explaining a detection device according to an embodiment of the present invention. The detection device according to the present embodiment includes a twin mirror infrared sensor 101, two A / D converters 102a and 102b, a feature amount extraction unit 103, a model storage unit 106, and a recognition processing unit 104. Yes. Note that the model generation unit 105 in the figure is configured to store and generate a probability model stored in the model storage unit 106. In the present embodiment, it is assumed that the model generation unit 105 is outside the detection device, creates a probability model in advance, and the created probability model is stored in the model storage unit 106.

  The twin mirror type infrared sensor 101 includes infrared sensors 101a and 101b that convert the amount of change in infrared rays emitted from the detected object into an electrical signal and output it as an analog signal. That is, the infrared sensors 101a and 101b have the same function as the configuration in which the sensor unit 5 and the amplifier circuit 8 shown in FIG. 6 are combined, or the configuration in which the sensor unit 6 and the amplifier circuit 9 are combined. The twin mirror infrared sensor 101 outputs a signal similar to the signal output from the amplifier circuits 8 and 9 in FIG.

  The A / D converters 102a and 102b convert the two analog output signals of the twin mirror infrared sensor 101 into digital signals, respectively. In this embodiment, the sensor output output from the A / D converter 102a is referred to as sensor output A, and the sensor output output from the A / D converter 102b is referred to as sensor output B. The feature amount extraction unit 103 calculates feature amount data from the sensor output A and the sensor output B.

The feature amount data is information regarding the frequency in a relatively short time for each of the sensor output A and the sensor output B. The feature amount data and this extraction method will be described later.
The model storage unit 106 is a storage device that stores a model generated based on the feature amount data extracted by the feature amount extraction unit 103. Based on the model contents stored in the model storage unit 106 and the feature amount data extracted by the feature amount extraction unit 103, the recognition processing unit 104 can detect the object to be detected by the twin mirror infrared sensor 101. It is the structure which recognizes the attribute information of the to-be-detected body which exists in a range (detection area).

Note that the model stored in the model storage unit 106 is created by modeling the feature data into an HMM (Hidden Markov Model). HMM is one of probabilistic models, and estimates an unknown parameter from observable information on the assumption that a phenomenon proceeds by a Markov process including an unknown parameter. The estimated parameters include a state transition probability and a symbol output probability.
In the present embodiment, the feature amount extracted by the feature amount extraction unit 103 is accumulated, and the model created in advance by the model generation unit 105 is stored in the model storage unit 106.

Hereafter, each above-mentioned structure is demonstrated in detail.
Feature Quantity Extraction Unit FIG. 2 is a diagram for explaining feature quantity extraction performed by the feature quantity extraction unit 103. The sensor output A shown in FIG. 2 is the sensor output output from the A / D converter 102a shown in FIG. 1 as described above. The sensor output B is the sensor output output from the A / D converter 102b shown in FIG. The sensor output A and the sensor output B are each divided into short sections (sensor outputs output from the A / D converter within a predetermined time), and each section is hereinafter referred to as a frame.

  Each frame described above is arranged in time series. Each frame may overlap in time with each other. In FIG. 2, a plurality of frames sA-1, sA-2, sA-3, and sA-4 generated by dividing the sensor output A are overlapped with each other in the frame sections adjacent to each other. ing. Also, in the plurality of frames sB-1, sB-2, sB-3, and sB-4 generated by dividing the sensor output B, half of the frame sections overlap each other in the adjacent frame sections. .

The data indicated by the sensor output divided into frames is converted into feature data in units of frames. In the present embodiment, three feature quantities, ie, a first feature quantity (first feature quantity data), a second feature quantity (second feature quantity data), and a third feature quantity (third feature quantity data) are used as feature quantities. Extract.
The first feature value data is obtained by converting each frame signal into a spectrum that is a frequency domain signal by FFT (Fast Fourier Transform), and further converting the converted spectrum into a ratio of each frequency component. The logarithm is taken to ensure the dynamic range.

The first feature amount data is calculated independently for the sensor output A and the sensor output B. The first feature quantity data fA1-1, fA1-2, fA1-3, and fA1-4 calculated from the sensor output A are the first feature quantity data fB1-1, fB1-2, and fB1 calculated from the sensor output B. -3 and fB1-4 are connected to the same sensor output timing. The first feature amount data after connection is denoted as f1-1, f1-2, f1-3, and f1-4 in the drawing.
The second feature amount data is obtained by converting a signal of each frame into a spectrum that is a frequency domain signal by FFT, and for each frequency component, a certain frame (first frame) and immediately before or immediately after this frame (this In the embodiment, the ratio of the spectrum with the frame (second frame) output in the immediately preceding frame is taken and the logarithm of this ratio is taken.

  According to the first feature value data and the second feature value data described above, the presence or absence of the detected object is detected by detecting the change in the detection area based on the content (change in frequency) of the sensor output, not the change in the amplitude of the sensor output. Can be detected. Therefore, since the first feature value data and the second feature value data are parameters having a low dependency on the signal amplitude of the sensor output, even when the amplitude of the sensor output is small, the signal amplitude is large. Similarly, it can be used as a parameter for detecting an object to be detected.

The amplitude of the sensor output is, for example, when the environmental temperature (background temperature) is high and the difference between the detected object and the background temperature is small, or when the detected object passes through the detection area in a short time Get smaller. However, according to the first feature value data and the second feature value data, it is possible to detect the detected object even in a situation that cannot be detected by the prior art.
The second feature amount data is calculated independently for sensor output A and sensor output B. The second feature value data fA2-1, fA2-2, fA2-3, and fA2-4 calculated from the sensor output A are the second feature value data fB2-1, fB2-2, and fB2 calculated from the sensor output B. -3 and fB2-4 are connected to the same output timing. The second feature data after concatenation is shown in the figure as f2-1, f2-2, f2-3, f2-4
.

Further, the third feature amount data is a correlation coefficient of frames output simultaneously between the frame of sensor output A and the frame of sensor output B. In the figure, the third feature amount data calculated for each frame is denoted as f3-1, f3-2, f3-3, and f3-4.
The absolute value of the third feature amount data increases when the detected object enters the detection areas of both of the two sensors of the twin mirror type infrared sensor 101 at the same time. On the other hand, it becomes smaller when the detected object enters only the detection area of one of the two sensors. Therefore, according to the third feature amount data, it is possible to determine whether or not the detected object has a size that can simultaneously enter the detection areas of the two sensors. The determination result serves as a guideline for determining whether the detected object is a human or a small animal.

Further, since the third feature value data is calculated in units of frames, it is possible to observe the temporal variation of the third feature value data. In this embodiment, the presence / absence of the detected object can be determined more accurately than the prior art that determines the presence / absence of the detected object based on the instantaneous values of the sensor outputs of the two sensor units.
Further, the third feature amount data as the correlation coefficient can represent the difference between the two in a range of ± 1 regardless of the amplitudes of the sensor output A and the sensor output B. By using such third feature amount data, and the first feature amount data and the second feature amount data that do not depend on the amplitude of the sensor output, the present embodiment has a higher amplitude (intensity) of the sensor output than in the past. It is possible to detect the detection target even when the change is small.
The third feature values f3-1, f3-2, f3-3, and f3-4 are, as illustrated, the first feature value data f1-1, f1-2, f1-3, f1-4, and the second feature. Along with the quantity data f2-1, f2-2, f2-3, and f2-4, they are linked in time series.

Model Generation Next, the model generation unit 105 will be described. In this embodiment, it is assumed that the model generation unit 105 is provided outside the detection device. The model generation unit 105 creates a probability model in advance based on data obtained by storing feature amounts acquired by other detection devices, feature amounts acquired by the device itself, and the like. According to the present embodiment as described above, the probability model can be stored in the model storage unit 106 in advance at the time of shipment of the detection device. For this reason, the detection apparatus can determine the detection target using a probability model created based on a large number of feature values immediately after shipment.

  The model generation unit 105 generates an HMM that is a probability model based on the feature amount data extracted by the feature amount extraction unit 103. Since the generation of the HMM is performed as preparation for recognition processing executed in the recognition processing unit 104, the HMM generated once and stored in the model storage unit 106 does not operate thereafter. Since the HMM and the method for generating the HMM are well known, further description is omitted in this specification.

  At this time, in the present embodiment, it is assumed that each HMM is associated with the cause of infrared fluctuation based on the parameters of the created HMM. In this way, each HMM can be associated with the cause of infrared fluctuations. In addition, as a cause of the fluctuation | variation of infrared rays, you may make it make a detected body match | combine two types other than a human body and a detected body other than a human body. Further, the detected body may be a human body, the detected body is other than a human body, and two or more types such as disturbances may be associated.

Recognition processing unit The recognition processing unit 104 includes the first feature value data, the second feature value data, and the third feature value data generated by the feature value extraction unit 103, and the HMM stored in the model storage unit 106. Based on the attribute of the object to be detected.
In the present embodiment, the recognition processing unit 104 is calculated by the feature amount extraction unit 103 from the HMM stored in the model storage unit 106 using a known maximum likelihood sequence estimation (MLSE). The HMM that is most likely to generate the first feature value data, the second feature value data, and the third feature value data is detected.

(Modification)
The embodiment described above uses the twin mirror type infrared sensor 101 as the thermal radiation detection means. However, the embodiment of the present invention is not limited to the configuration using the twin mirror type infrared sensor 101 as the thermal radiation detection means. That is, this embodiment can be used as thermal radiation detection means as long as it has at least two infrared sensors that can detect infrared rays in different detection areas. In addition, the infrared sensor may be any sensor that can output an electrical signal corresponding to the amount of variation in infrared rays.

  As another example of such thermal radiation detection means, for example, a configuration in which a plurality of pyroelectric elements are provided in one housing can be cited. The thermal radiation according to the present embodiment has a configuration in which one electrical signal is output from each of the pyroelectric elements or a part of the plurality of pyroelectric sensors to extract a plurality of electrical signals in total from one housing. It can be used as a detection means. Such thermal radiation detection means is referred to as a pyroelectric sensor in this embodiment.

  In this specification, infrared rays are detected by utilizing the phenomenon that the charge on the pyroelectric plate surface of ceramics (such as lead zirconate titanate (PZT)) with spontaneous polarization increases according to changes in ambient temperature. The element that performs this is called a pyroelectric element. A resistance element and a field effect transistor are added to the pyroelectric plate, and an increase in charge is output as an electric signal. The pyroelectric sensor referred to in the present specification refers to an infrared sensor configured by enclosing a plurality of such pyroelectric elements in one package.

FIGS. 3A and 3B are diagrams for explaining a pyroelectric sensor according to a modification of the present embodiment. FIG. 3A is a diagram for explaining a pyroelectric element included in the pyroelectric sensor. FIG. 3B is an equivalent circuit for explaining the electrical connection state of the pyroelectric element shown in FIG.
As shown in FIG. 3A, the infrared sensor according to the modification includes four pyroelectric elements 301 a, 301 b, 301 c, and 301 d in one case 300. Of the four pyroelectric elements, the sensor output 1 that is an electrical signal is output from the pyroelectric elements 301a and 301b, and the sensor output 2 is output from the pyroelectric elements 301b and 301c.
For this purpose, as shown in FIG. 3B, the pyroelectric element 301a and the pyroelectric element 301b are connected in series, and the pyroelectric element 301c and the pyroelectric element 301d are connected in series. Then, the pyroelectric element 301b and the pyroelectric element 301c are connected to the ground. Then, the sensor outputs may be extracted from the pyroelectric elements 301a and 301d, respectively.

Moreover, if a mirror or a lens is provided in front of such a pyroelectric sensor, the infrared detection range of the pyroelectric element can be expanded. In addition, the pyroelectric sensor of the modification of this embodiment is not limited to the structure provided with four pyroelectric elements, and may be provided with two or more pyroelectric elements.
Furthermore, a pyroelectric sensor including a plurality of pyroelectric elements has an existing configuration. However, in the existing pyroelectric sensor, a plurality of pyroelectric elements are all connected in series to obtain one sensor output. In the present embodiment, the electrical connection of pyroelectric elements in an existing pyroelectric sensor can be changed and used as thermal radiation detection means.

(Experiment for confirming the effect of the embodiment)
Next, Experimental Example 1 and Experimental Example 2 for confirming the effects of the present embodiment described above will be described.
[Experimental Example 1]
Experimental Example 1 uses a detection device in which the present invention is applied to a commercially available twin mirror type intrusion detection sensor (hereinafter referred to as the intrusion detection sensor of the present invention), and determines whether or not the detected object is a person. It is.
FIG. 4 is a diagram for explaining the results of Experimental Example 1. The horizontal axis of the graph shown in FIG. 4 is the height range (height category) of the subject who is the detected object, and the vertical axis is the identification rate corresponding to each height category. The identification rate refers to the probability of the number of times that the detected object is correctly determined to be a human body among the number of times that the detected object enters the detection area. The prior art shown in the figure is a commercially available twin mirror type intrusion detection sensor (hereinafter referred to as a commercially available intrusion detection sensor) in a state where the present invention is not applied. The present invention is the above-described intrusion detection sensor of the present invention. In the intrusion detection sensor and the commercially available intrusion detection sensor of the present invention, an alarm is issued when it is determined that the detected object is a person. For this reason, the above-described identification rate refers to the probability of the number of times that an alarm is generated out of the number of times that the detected object enters the detection area.

In Experimental Example 1, a subject was passed through a detection area as a detected body, and it was determined whether or not the detected body was a person using the intrusion detection sensor of the present invention and a commercially available intrusion detection sensor. Each height category and number of subjects is as follows.
Height: Less than 100cm 7: Height: 100cm to less than 115cm 7: Height: 115cm to less than 130cm 7: Height: 130cm to less than 145cm 7: Height: 145cm to less than 160cm 7: Height: 160cm to less than 175cm 7: Height: 175cm to 7: Total 49

In Experimental Example 1, subjects of any height are walking at the normal walking speed from end to end of the detection area. The number of walks is 100 per person, and the walked course is a vertical, horizontal, and diagonal straight line with respect to the detection area, covering the detection area evenly.
In Experimental Example 1, data is acquired by moving a warm detected object other than a person on the same course as the subject. The number of acquired data is 1800 samples.
Note that the intrusion detection sensor of the present invention is installed at a height of 2.5 m from the floor regardless of whether the body to be detected is a human body or a body other than the human body.

  In Experimental Example 1, the sensor output of the intrusion detection sensor of the present invention was acquired under the above conditions. The sensor output is converted into first feature value data and second feature value data. Also, using the converted feature data as learning data, nine probability models of humans and seven probability models of detected objects other than humans were created. Then, all the converted feature data is used as evaluation data, and whether or not the detected object is a person is identified based on the probability model.

  As a result, as shown in FIG. 4, the discrimination rate of the intrusion detection sensor of the present invention is higher in all subject categories than the discrimination rate of the commercially available intrusion detection sensor. In particular, it can be seen that the discrimination rate for subjects with a short height is greatly improved. Further, when the discrimination rate of the detected object other than the human body was measured by the same experiment, the discrimination rate of 100% was obtained for the intrusion detection sensor of the present invention, and 99.7% was obtained for the commercially available intrusion detection sensor. From these results, it can be said that the human body recognition apparatus of the present invention can accurately detect only the human body with higher accuracy than a commercially available intrusion detection sensor.

[Experiment 2]
FIG. 5 is a diagram for explaining Experimental Example 2. Experimental Example 2 is an experiment on the identification rate of a detected object other than a human body. In Experimental Example 2, the installation height of the intrusion detection sensor is changed. The horizontal axis in FIG. 5 indicates the installation height of the intrusion detection sensor, and the vertical axis indicates the identification rate corresponding to each installation height. The identification rate refers to the probability of correctly determining that the detected object is not a human body out of the number of times the detected object has entered the detection area. The intrusion detection sensor and the commercially available intrusion sensor of the present invention issue an alarm only when it is determined that the detected object is a human body. For this reason, the identification rate shown on the vertical axis in FIG. 5 refers to the probability of the number of times that an alarm has not occurred among the number of times the detected object has entered the detection area.

As the installation height of the intrusion detection sensor, three types of 0.8 m, 0.5 m, and 0.2 m are set. A body to be detected other than a human body moves in the same detection area as in Experimental Example 1 along the same course as in Experimental Example 1. In Experimental Example 2, 160 samples of data are acquired for one installation height of the intrusion detection sensor.
The sensor output output under the above conditions is converted into first feature value data, second feature value data, and third feature value data. In Experimental Example 2, nine probability models in which the detected object generated in Experimental Example 1 is a human body and seven probability models of detected objects other than the human body are stored. That is, the probability model used in Experimental Example 2 is only created based on the feature amount obtained when the installation height of the intrusion detection sensor is 2.5 m.

According to FIG. 5, it can be seen that the intrusion sensor of the present invention can obtain a higher identification rate than the commercially available intrusion sensor regardless of the installation height. Note that, using the probability model used in Experimental Example 2, an identification rate for identifying the detected object as a person was calculated. As a result, the same result as that obtained in the first experimental example was obtained.
From the above results, it can be said that the present invention can more accurately determine whether or not the detected object is a human body than a commercially available intrusion sensor that is a conventional detection apparatus, and can provide a detection apparatus.

DESCRIPTION OF SYMBOLS 101 Twin mirror type infrared sensor 101a, 101b Infrared sensor 102a, 102b Converter 103 Feature-value extraction part 104 Recognition processing part 105 Model production | generation part 106 Model storage part 300 Case 301a, 301b, 301c, 301d Pyroelectric element

Claims (7)

Thermal radiation detection means comprising at least two infrared sensors that detect infrared rays in different detection areas and output an electrical signal corresponding to the detected amount of variation in the infrared rays;
Model storage means for storing a plurality of probability models created on the basis of feature quantities indicating information relating to the frequency of the electrical signal according to the amount of variation in infrared rays;
A first information according to the frequency of the first electrical signal output by the first infrared sensor of said which is the heat radiation detecting means comprises at least two infrared sensors, of said at least two infrared sensors Feature amount extraction means for extracting feature amounts extracted by connecting independently the second information relating to the frequency of the second electrical signal output by the second infrared sensor , and
A probability model corresponding to the feature quantity is detected from the plurality of probability models based on the likelihood of the feature quantity extracted by the feature quantity extraction means and the probability model stored by the model storage means. A probability model detection means;
Based on the probability model detected by the probability model detection means, detected object recognition means for recognizing the detected object that has entered the detection area;
A detection device comprising:   2. The detection device according to claim 1, wherein the thermal radiation detection unit includes a plurality of infrared detection elements as one of the infrared sensors in one casing, and outputs a plurality of the electrical signals simultaneously from the casing. . The feature amount extraction means includes:
A first feature amount representing a component ratio of each frequency with respect to a plurality of frequencies included in each frame by dividing the first electric signal and the second electric signal at predetermined time intervals to generate a plurality of frames. the sensing device according to claim 1 or 2, characterized in that respectively calculated as the first information and the second information. The detection apparatus according to claim 3, wherein the component ratio is represented by a logarithm in the first feature amount. The feature amount extraction means includes:
The first electrical signal and the second electrical signal are divided at predetermined time intervals to generate a plurality of frames, and the first frame, which is one of the plurality of frames, and in time, The second feature amount representing a component ratio of each frequency with a second frame located immediately before or after the first frame is calculated as the first information and the second information , respectively. 5. The detection device according to any one of 1 to 4. Oite before Symbol second feature amount detection apparatus according to claim 5, wherein the component ratio is characterized by being represented by the logarithm. The feature amount extraction means includes:
The first electric signal and the second electric signal are divided at predetermined time intervals to generate a plurality of frames, and the correlation between the first electric signal and the second electric signal is calculated for each frame. The detection device according to any one of claims 1 to 6, wherein a third feature amount represented by (2) is extracted.

Images