JP5065744B2 - Individual detector - Google Patents

Description

  The present invention generates a distance image having a pixel value as a distance value to an object existing in the detection region by imaging the detection region, and detects an object in the detection region using the distance image It is about.

  As this type of individual detector 1, for example, as shown in FIG. 8, a distance image sensor 2 installed above the detection area A1 and imaging the detection area A1 to generate a distance image P1 (see FIG. 9), There is provided an apparatus including object detection means 3 for detecting an object (for example, human body H1, H2) in a detection area A1 using a distance image P1 generated by the distance image sensor 2 (for example, Patent Documents). 1).

  As an example of the distance image sensor 2, for example, the distance value is determined by calculating the distance to each location in the detection area A1 based on the time from irradiating the detection area A1 to receiving the reflected light. Some generate a distance image P1 as a value. The distance image sensor 2 is fixed at a fixed position with respect to a background object behind the detection area A1 (floor F1, door 8, etc. in the example of FIG. 8).

  For example, as shown in FIG. 9, the object detection unit 3 holds a distance image having a distance value from the distance image sensor 2 to the background object as a pixel value as a background distance image P2, and obtains it from the distance image sensor 2. A difference distance image P3 that is a difference between the distance image P1 and the background distance image P2 to be generated is generated, and a foreground distance image P4 is generated by extracting an area where the absolute value of the distance value is greater than or equal to a threshold th1 in the difference distance image P3. Then, the region extracted as the foreground distance image P4 is detected as an object. The background distance image P2 is generated, for example, by averaging distance values for each pixel for a plurality of distance images P1 acquired from the distance image sensor 2 in a state where no object is present in the detection area A1.

  The operation of the individual detector 1 will be described with reference to FIG. 2 showing the first embodiment of the present invention.

  The object detection unit 3 generates the difference distance image P3 by subtracting the distance image P1 acquired from the distance image sensor 2 from the background distance image P2 (that is, subtracting the distance value for each pixel). Therefore, when only the human bodies H1 and H2 exist on the floor F1 of the detection area A1 as shown in FIG. 2A, ideally, the differential distance image P3 is in the area corresponding to the human body H1 and the floor F1 of the human body H1. The height (height) from the floor F1 of the human body H2 in the area corresponding to the human body H2 as the distance value. Then, the object detection unit 3 compares the distance value for each pixel of the difference distance image P3 with the threshold value th1, and the area where the distance value is equal to or greater than the threshold value th1 is shown in FIG. 2B, foreground distance image P4. Extract as In other words, an area of the difference distance image P3 whose distance value is lower than the threshold th1 is regarded as a background object (here, the floor F1), and is present at a height equal to or higher than the threshold th1 from the floor F1 as shown in FIG. The target object (here, human bodies H1 and H2) is extracted as the foreground distance image P4. From this foreground distance image P4, an object existing in the detection area A1 is detected.

By the way, the distance value obtained by the distance image sensor 2 may cause an error due to the influence of an external light component such as ambient light. Here, even if an error occurs in the distance value of the background object, the distance value of the difference distance image P3 of the background object is not less than the threshold th1, so that the background object is not mistaken for the object. The threshold th1 is set in consideration of the error of the distance value. That is, assuming that the error of the distance value for the same pixel of the background object (here, the floor F1) varies according to the normal distribution, the standard deviation σ1 representing the degree of variation of the error distribution is taken into account, for example, FIG. ), A threshold value th1 having a value (σ1 + M) larger than σ1 by a predetermined margin M is set. Thereby, the reliability which can consider the area | region where distance value is less than threshold value th1 among the difference distance images P3 as a background object is maintained constant. This threshold th1 is obtained by connecting distance data (distance value and distance data) of a plurality of distance images P1 generated by the distance image sensor 2 by connecting an external device (for example, a personal computer) to the individual detector 1 when the individual detector 1 is installed. It is manually set to an optimal value while monitoring the error) with an external device.
Japanese Patent Laying-Open No. 2006-64695 (FIG. 1)

  However, the error probability of the distance value of the background object may fluctuate due to the environmental change in the detection area A1. For example, the light received when the distance image sensor 2 measures the distance value of the background object includes signal components (light output from the distance image sensor 2 and reflected by the background object) and noise components (external light such as ambient light). Component), the ratio of the signal component in the light received by the distance image sensor 2 decreases when the light reflectance of the floor F1 decreases due to, for example, the replacement of the floor F1 as a background object. The SN ratio of the distance value of the background object decreases, and the probability of occurrence of an error in the distance value of the background object increases. When the error occurrence probability increases, the degree of variation also increases with respect to the distance value of the floor F1 actually measured by the distance image sensor 2. Thus, when the degree of variation in the distance value of the background object fluctuates and becomes large, the standard deviation σ2 of the error distribution after the fluctuation becomes larger than the standard deviation σ1 of the error distribution before the fluctuation as shown in FIG. Therefore, if the threshold value th1 is a value set in consideration of the standard deviation σ1 before the change, the reliability that the region in the difference distance image P3 whose distance value is lower than the threshold value th1 can be regarded as a background object is lowered. Accordingly, the distance value of the difference image P3 of the background object is equal to or greater than the threshold th1, and a problem that the background object and the object cannot be distinguished from each other in the foreground distance image P4 as shown in FIG. The reliability of detection is reduced.

  Further, if the threshold value th1 is manually reset according to the environmental change in the detection area A1, a decrease in the reliability of detection of the target can be avoided, but there is a problem that it takes time to reset the threshold value th1. is there. In particular, when a part of the individual detector 1 is embedded in the ceiling, it is very troublesome to connect an external device to the individual detector 1 and reset the threshold th1.

  The present invention has been made in view of the above-described reasons, and can prevent a decrease in the reliability of detection of an object due to an environmental change in a detection area, and manually set a threshold according to an environmental change in the detection area. An object of the present invention is to provide an individual detector that does not require time and effort to set again.

  According to the first aspect of the present invention, a distance image sensor that generates a distance image having a pixel value as a distance value to an object existing in the detection area by imaging the detection area while being fixed in position relative to a background object behind the detection area. And a distance image having a distance value from the distance image sensor to the background object as a pixel value as a background distance image, and generating a difference distance image that is a difference between the distance image obtained from the distance image sensor and the background distance image, A target detection unit that detects, as a target, a region in which the absolute value of the distance value is equal to or greater than a predetermined threshold in the distance image; and a threshold setting unit that sets the threshold used in the target detection unit, A plurality of distance images are acquired from the distance image sensors at different times, and an index value indicating the degree of variation in distance value is calculated for the same pixel of the background object among the plurality of distance images. Characterized in that it has a calculation unit for, and a setting unit that sets automatically statistically the threshold value based on the index value calculated by the calculation unit.

  According to this configuration, the threshold value setting means acquires a plurality of distance images from the distance image sensor at different times, and indicates the degree of variation in distance value for the same pixel excluding the object among the plurality of distance images. Since it has a calculation unit that calculates a value and a setting unit that automatically sets the threshold statistically based on the index value calculated by the calculation unit, the distance of the background object due to the environmental change of the detection region Even if the degree of variation in value may vary, it is possible to maintain a certain degree of reliability so that a region where the absolute value of the distance value is below the threshold in the difference distance image can be regarded as a background object. Therefore, it is possible to avoid the occurrence of a problem that the background object and the target object cannot be distinguished because the absolute value of the distance value of the differential distance image of the background object is equal to or greater than the threshold, and the reliability of the target object detection due to the environmental change in the detection region Can be avoided. Further, since the threshold value is automatically set by the threshold value setting means, there is an advantage that it does not take time and effort to manually reset the threshold value according to the environmental change of the detection region.

  According to a second aspect of the present invention, in the first aspect of the invention, the object is a human body, and a human body detection sensor for detecting the presence or absence of the human body in the detection region is provided, and the calculation unit includes a human body in the detection region. When the image does not exist, the distance image is acquired from the distance image sensor, and the index value is calculated.

  According to this configuration, since the index value calculated by the calculation unit is calculated from the distance image acquired from the distance image sensor when no human body exists in the detection region, it is affected by the presence or absence of the human body in the detection region. It is possible to reliably avoid the index value from becoming larger than necessary. Therefore, it is possible to avoid setting an unnecessarily large threshold value in the setting unit, and as a result, it is possible to improve the detection sensitivity of the object.

  According to a third aspect of the present invention, in the first or second aspect of the present invention, the calculating unit calculates the index value for each pixel of the distance image, and the setting unit is configured for each pixel of the distance image. The threshold value is set.

  According to this configuration, since a threshold is set for each pixel of the distance image, an optimum threshold is set for each pixel when the degree of variation in the distance value is different for each pixel in the detection region. It is possible to improve the detection sensitivity of the object.

  According to a fourth aspect of the present invention, in the first or second aspect of the present invention, the calculation unit divides the distance image into a plurality of small regions each composed of a plurality of adjacent pixels, and the same pixel for each small region. And the setting unit sets the threshold value for each small region of the distance image.

  According to this configuration, the threshold is set for each small area of the distance image. Therefore, when the degree of variation in the distance value is different for each small area in the detection area, the optimum value is set for each small area. A threshold can be set, and the detection sensitivity of the object can be improved. Further, as compared with the case where a threshold is set for each pixel of the distance image, there is an advantage that the amount of data processed by the threshold setting means can be reduced and the processing speed can be improved.

  According to the present invention, the threshold setting unit includes a calculation unit that calculates an index value indicating the degree of variation in the distance value, and a setting unit that automatically sets the threshold statistically based on the index value calculated by the calculation unit. Therefore, even if the degree of variation in the distance value of the background object may fluctuate due to the environmental change in the detection area, the absolute value of the distance value of the difference image of the background object is equal to or greater than the threshold value and the object It is possible to avoid the occurrence of a problem that makes it impossible to distinguish an object from each other, and it is possible to avoid a decrease in reliability of detection of an object due to an environmental change in a detection region. Further, since the threshold value is automatically set by the threshold value setting means, there is an effect that it does not take time and effort to manually reset the threshold value according to the environmental change of the detection region.

  In each of the following embodiments, as shown in FIG. 8, in the room entrance management system including the door 8 with the electric lock 7 and the authentication device 9 for unlocking, it is recognized as valid by the authentication by the authentication device 9. An example in which the individual detector 1 of the present invention is used for a co-entry detection apparatus 10 that detects so-called co-entry that another human body H2 enters at the same time when the human body H1 enters the room is shown. In this example, when the human body H1 enters the room, the individual detector 1 detects the human bodies H1 and H2 in the detection area (space near the door 8) A1 in response to a signal from the control device 11, and a plurality of human bodies H1 and H2 are detected. When it is detected, the co-incident detection means 12 determines that the co-incidence is present and causes the alarm means 13 to issue a notification. However, the use of the individual detector 1 of the present invention is not limited to the co-entry detection device 10.

(Embodiment 1)
The individual detector 1 according to the present embodiment generates a distance image P1 having a pixel value as a distance value to an object existing in the detection area A1 by imaging the detection area A1 (see FIG. 2) as shown in FIG. A distance image sensor 2 for detecting the object in the detection area A1 using the distance image P1 generated by the distance image sensor 2. Here, the distance image sensor 2 is attached to a ceiling (not shown), and is fixed at a fixed position with respect to the floor F1 as a background object behind the detection area A1, as shown in FIG. The actual distance to the floor F1 is unchanged. The detection area A1 is formed in a pyramid shape with the distance image sensor 2 as a vertex between the distance image sensor 2 and the floor F1.

  In this embodiment, the distance image sensor 2 includes, for example, a light source (not shown) that irradiates the detection area A1 with infrared rays, and a plurality of two-dimensionally arranged photosensitive units (not shown), and an optical system (not shown). And a light detecting element (not shown) for receiving the reflected infrared light from the detection area A1 through each photosensitive portion, and the time from when the infrared light is irradiated until the light is received (hereinafter referred to as round trip time). Based on the above, the distance value to the object existing in the detection area A1 is obtained for each pixel. However, since the round-trip time is very short, the light output from the light source and the light detection element are received using intensity-modulated light that is modulated so that the intensity of the light applied to the detection area A1 periodically changes at a constant period. The round trip time is obtained from the phase difference Φ with the light to be measured.

Here, as shown in FIG. 3A, when the light intensity (projection intensity) output from the light source is modulated by a sine wave having a predetermined modulation frequency (for example, 20 MHz), the light intensity is received by the light projection intensity and the light detection element. The phase difference Φ with respect to the light intensity (light reception intensity) can be calculated using the amount of light received by the light detection element obtained at a plurality of timings. For example, the received light amounts in the respective sections of 0 to 90 degrees, 90 to 180 degrees, 180 to 270 degrees, and 270 to 360 degrees of the phase of the intensity modulated light from the light source are q0, q1, q2, and q3 (FIG. 3 ( b) corresponds to the area of the shaded area), and the phase difference Φ does not change (that is, the distance to the object does not change) while obtaining the received light quantity q0, q1, q2, q3, and the object If there is no change in the reflectance, the phase difference Φ [rad] can be expressed by Φ = tan ⁻¹ {(q2−q0) / (q1−q3)}.

  Using the phase difference Φ [rad] obtained in this way and the modulation frequency f [Hz] of the intensity modulated light, the round trip time Δt [s] can be expressed as Δt = Φ × (1 / 2πf) [s]. it can. Then, by using this round-trip time Δt [s] and the speed of light c [m / s], the distance value from the distance image sensor 2 to the object is represented by Δt × c / 2 [m].

  As shown in FIG. 1, the object detection unit 3 includes a detection processing unit 31 and an identification processing unit 32, and the detection processing unit 31 stores in advance a distance image P <b> 1 obtained from the distance image sensor 2. A difference distance image P3 that is a difference with the background distance image P2 (distance image having the distance value from the distance image sensor 2 to the background object as a pixel value) P2 is generated, and the absolute value of the distance value in the difference distance image P3 A foreground distance image P4 is generated by extracting an area that exceeds a certain threshold, and the area extracted as the foreground distance image P4 is detected as an object. The background distance image P2 is obtained by, for example, averaging distance values for each pixel with respect to a plurality of distance images P1 acquired from the distance image sensor 2 in the state where the object does not exist in the detection area A1 and only the background object exists. Generated. The identification processing unit 32 subsequent to the detection processing unit 31 counts the number of people in the detection area A1 from the foreground distance image P4 generated by the detection processing unit 31 in order to identify co-entrance.

  The operation of detecting the object in the detection area A1 in the individual detector 1 having the above-described configuration is the same as the operation described with reference to FIG.

  By the way, even if the distance from the distance image sensor 2 to the object is not changed, if a plurality of distance images P1 are actually generated at different timings, the distance values of the same pixel usually vary among the plurality of distance images P1. Arise. Therefore, in order to prevent the distance value of the difference distance image P3 of the background object from exceeding the threshold th1 even if the distance value of the background object varies, the background object is prevented from being mistaken for the object. It is necessary to set the threshold th1 in consideration of the degree of variation in the distance value. However, the degree of variation in the distance value of the background object may fluctuate due to the environmental change of the detection area A1, for example, the reflectivity of the floor F1 (here, the reflection with respect to infrared rays) due to the replacement of the floor F1, which is the background object. When the ratio is decreased, the SN ratio of the distance value of the background object in the distance image P1 may decrease, and the variation of the distance value of the background object may increase. In the case of the present embodiment, the received light quantity q0, q1, q2, q3 used when the distance value is obtained by the distance image sensor 2 includes the received light quantity of ambient light in addition to the light output from the light source. If the reflectance of the object decreases and the S / N ratio of the received light quantity q0, q1, q2, q3 decreases, the error occurrence probability increases with respect to the distance value of the background object, and the degree of variation in the distance value of the background object increases. Become.

  Therefore, the individual detector 1 of the present embodiment includes a threshold setting unit 5 that sets thresholds th1 and th2 used in the target detection unit 3 as shown in FIG. 1 in addition to the distance image sensor 2 and the target detection unit 3 described above. It has. The threshold value setting means 5 automatically calculates the threshold values th1 and th2 based on the index value calculated by the calculation unit 51, and a calculation unit 51 that calculates an index value indicating the degree of variation in the distance value of the background object. And a setting unit 52 for setting.

  The calculation unit 51 acquires a number of distance images P1 from the distance image sensor 2 at different timings, and calculates an index value for the same pixel of the background object among the number of distance images P1. Here, one specific pixel in the distance image P1 is used as a reference pixel, distance values of reference pixels extracted from a number of distance images P1 are accumulated, and an index value is calculated from the accumulated distance values. . In order to calculate the index value in the calculation unit 51, only the distance value of the reference pixel of the background object is used, and the distance value is accumulated only for the reference pixel whose distance value is equal to or greater than a predetermined value. That is, for example, when a human body is present at a position corresponding to the reference pixel in the detection area A1, the distance value of the reference pixel is equal to or less than a predetermined value, and the distance value of the reference pixel is discarded without being accumulated in the calculation unit 51. . In the present embodiment, it is assumed that the error regarding the distance value of the same pixel of the background object is distributed according to the normal distribution, and the calculation unit 51 stores a number of distance values in which standard deviations indicating the degree of variation of the normal distribution are accumulated. And is used as an index value. As is well known, the standard deviation is the square root of the variance and indicates the degree of variation from the average value (median value) of the normal distribution.

  The calculation of the index value by the calculation unit 51 described above is automatically performed at a constant cycle. As an example, since it takes an acquisition time of about several seconds for the calculation unit 51 to acquire a large number of distance images for calculating the index value from the distance image sensor 2, the calculation unit 51 uses the index every time the acquisition time elapses. Designed to calculate a value. However, the cycle in which the calculation unit 51 calculates the index value is not limited to the acquisition time. For example, if the index value is calculated at a cycle of once every several hours, the power consumption can be kept low.

  Each time the calculation unit 51 calculates an index value, the setting unit 52 automatically sets the thresholds th1 and th2 statistically based on the index value. Here, even if the degree of variation in the distance value of the background object may fluctuate due to the environmental change in the detection area A1, the area where the absolute value of the distance value is below the thresholds th1 and th2 in the difference distance image P3. The thresholds th1 and th2 are set based on the index value so that the reliability that can be regarded as a background object can be kept constant. Specifically, by adding a predetermined margin to the index value, the thresholds th1 and th2 are set to values that are larger than the index value by the margin. Here, the margin is a constant value set in advance, but a function of the index value may be used as the margin.

  Hereinafter, the operation of the threshold setting means 5 will be described with reference to FIGS. In the following description, the actual distance from the distance image sensor 2 to the background object (floors F1, F2) is x, and the floor F1 is replaced with the floor F2 during the period when the individual detector 1 is used. The example which the reflectance (reflectance with respect to infrared rays) of a background object falls is shown. FIG. 5 shows an error distribution when the distance value of the reference pixel of the background object is measured a plurality of times, with the horizontal axis representing the distance value and the vertical axis representing the frequency.

  It is assumed that the reference pixel distance value accumulated in the calculation unit 51 is distributed according to a normal distribution with a standard deviation σ1 as indicated by a solid line in FIG. 5 before the floor F1 is replaced. In this case, the calculation unit 51 calculates the standard deviation σ1 as the index value, and the setting unit 52 sets the threshold th1 to a value (σ1 + M) obtained by adding the predetermined margin M to the index value σ1. Therefore, a region in the difference distance image P3 whose distance value is lower than the threshold th1 (= σ1 + M) is regarded as a background object, and as shown in FIG. 2A, an object existing at a height equal to or higher than the threshold th1 from the floor F1. (Here, the human bodies H1, H2) are extracted as the foreground distance image P4. Therefore, even if the distance value of the background object obtained by the distance image sensor 2 varies so that the standard deviation is σ1, the difference distance image P3 of the background object is regarded as the background object because the distance value is substantially below the threshold value. Thus, it is possible to avoid the occurrence of a problem that the distance value of the difference distance image P3 of the background object becomes equal to or greater than the threshold th1 and the background object cannot be distinguished from the object.

  On the other hand, after the replacement from the floor F1 to the floor F2, the distance values of the reference pixels accumulated in the calculation unit 51 are distributed according to a normal distribution with a standard deviation σ2 as indicated by a broken line in FIG. To do. Here, since the dispersion of the distance value of the background object increases due to the decrease in the reflectance of the background object (floor F2), σ1 and σ2 have a relationship of σ1 <σ2. The calculation unit 51 calculates the standard deviation σ2 as the index value, and the setting unit 52 sets the threshold th2 to a value (σ2 + M) obtained by adding a predetermined margin M to the index value σ2. Therefore, an area in the difference distance image P3 whose distance value is lower than the threshold value th2 (= σ2 + M) is regarded as a background object, and as shown in FIG. 4A, an object existing at a height equal to or higher than the threshold value th2 from the floor F2. (Here, the human bodies H1, H2) are extracted as the foreground distance image P4. Therefore, even if the distance value of the background object obtained by the distance image sensor 2 varies so that the standard deviation becomes σ2, the difference image P3 of the background object is regarded as the background object because the distance value is substantially below the threshold value. Thus, it is possible to avoid the occurrence of the problem that the absolute value of the distance value of the difference distance image P3 of the background object is not less than the threshold value and the background object and the object cannot be distinguished.

  As described above, the threshold setting unit 5 acquires a plurality of distance images P1 from the distance image sensor 2 at different timings, and indicates the degree of variation in distance value for the same pixel of the background object among the plurality of distance images P1. Since the calculation unit 51 that calculates the index value and the setting unit 52 that automatically sets the thresholds th1 and th2 statistically based on the index value calculated by the calculation unit 51, the environmental change of the detection region A1 is affected. Even if the degree of variation of the distance value of the background object may vary due to this, the reliability that the area where the absolute value of the distance value is below the thresholds th1 and th2 in the difference distance image P3 can be regarded as the background object is constant. Can keep. Therefore, the absolute value of the distance value of the differential distance image P3 of the background object is greater than or equal to the threshold values th1 and th2, and it is possible to avoid the occurrence of a problem that the background object and the object cannot be distinguished, and the object due to the environmental change in the detection region A1. It is possible to avoid a decrease in the reliability of detection. Further, since the thresholds th1 and th2 are automatically set by the threshold setting means 5, there is an advantage that it does not take time to manually reset the thresholds th1 and th2 according to the environmental change of the detection area A1.

  In this embodiment, the threshold setting unit 5 sets two types of thresholds th1 and th2. However, the thresholds that can be set by the threshold setting unit 5 are not limited to the two types of thresholds th1 and th2. An arbitrary threshold value is set according to the degree of variation in the distance value of the background object. In addition, the distance image sensor 2 is exemplified by the configuration for obtaining the distance value to the object based on the round trip time of the infrared rays. However, the present invention is not limited to this configuration. For example, the configuration for obtaining the distance value to the object by the principle of the triangulation method. May be adopted.

(Embodiment 2)
The individual detector 1 of the present embodiment includes a human body detection sensor 6 that detects the presence or absence of a human body in the detection area A1, as shown in FIG. 6, and the distance image sensor when the calculation unit 51 does not exist in the detection area A1. 2 is different from the individual detector 1 of the first embodiment in that the distance image P1 is acquired from 2 and the index value is calculated.

  The human body detection sensor 6 is a so-called heat ray sensor that detects the movement of a person by detecting infrared rays emitted from the human body. Between the distance image sensor 2 and the threshold setting means 5, a switch element SW that is turned on / off according to the output of the human body detection sensor 6 is interposed. The switch element SW is turned on during a period when it is determined by the human body detection sensor 6 that no human body exists in the detection area A1, and is turned off during a period when the human body detection sensor 6 determines that a human body exists in the detection area A1.

  According to the configuration of the present embodiment, the index value calculated by the calculation unit 51 is always calculated from the distance image P1 acquired from the distance image sensor 2 in a state where no human body exists in the detection area A1. Thus, it is possible to reliably avoid the index value from becoming larger than necessary due to the influence of the presence or absence of the human body in the detection region A1. That is, as illustrated in the first embodiment, when the error of the distance value of the background object is distributed according to the normal distribution of the standard deviation σ2, the calculation unit 51 acquires a plurality of values acquired from the distance image sensor 2 to calculate the index value. If the distance image P1 acquired in a state where the human body exists in the detection area A1 is included in the distance image P1, the index value calculated by including the distance value of the human body in a part of the reference pixels is In this embodiment, the index value is calculated from the distance image P1 acquired in a state where no human body exists in the detection area A1, and therefore the index value is larger than necessary than σ2. There is no. Therefore, it is possible to avoid setting the threshold value th2 larger than necessary in the setting unit 52, and as a result, it is possible to improve the detection sensitivity of the object.

  Other configurations and functions are the same as those of the first embodiment.

(Embodiment 3)
In the individual detector 1 of the present embodiment, the threshold value setting means 5 does not set a single threshold value for the entire detection area A1, but the detection area A1 is divided into a plurality of sections and each section is individually separated. Is different from the individual detector of the first embodiment. Specifically, the calculation unit 51 calculates an index value for each pixel of the distance image P1, and the setting unit 52 sets a threshold value for each pixel of the distance image P1. In this case, as shown in FIG. 7B, in the detection area A1, a threshold is set for each section a1 corresponding to one pixel of the distance image P1.

  That is, according to the configuration of the present embodiment, for example, as shown in FIGS. 7A and 7B, the right half and the left half of the detection area A1 have different reflectivities of the floors F10 and F20 as background objects. For the detection area A1 on the floor F10 having the lower reflectance (that is, the greater the degree of variation in the distance value), the standard deviation σ10 of the variation in distance value is calculated as the index value, and the threshold th10 is the index value σ10. Is set to a value obtained by adding a predetermined margin M (σ10 + M), and the detection area A1 on the floor F20 having a higher reflectance (that is, a smaller degree of variation in distance value) has a distance value as an index value. A standard deviation σ20 of variation (where σ20 <σ10) is calculated, and the threshold th20 is set to a value (σ20 + M) obtained by adding a predetermined margin M to the index value σ20 (that is, th10> th20). . Therefore, as shown in FIGS. 7A and 7B, a human body H1 having a height greater than or equal to the threshold th10 exists on the floor F10 having a lower reflectivity, and less than the threshold th10 on the floor F2 having a higher reflectivity. Even when the human body H2 has a height, if the height of the human body H2 is equal to or greater than the threshold th20, the distance value of the difference distance image P3 for each of the human body H1 and the human body H2 is equal to or greater than the thresholds th10 and th20 and can be detected as an object. Become.

  On the other hand, in the configuration in which the only threshold th10 is set for the entire detection area A1, when the reflectance of the floors F10 and F20 as background objects is different between the right half and the left half of the detection area A1, the distance By calculating the index value (σ10) using the reference pixel of the floor F10 having the lower reflectivity (that is, the greater degree of variation of the distance value) in the image, a threshold value is obtained as shown in FIG. th10 is set to a value (σ10 + M) obtained by adding a predetermined margin M to the index value σ10 in the entire detection area A1. Accordingly, when the human body H1 having a height equal to or higher than the threshold th10 exists on the floor F10 having a lower reflectance and the human body H2 having a height lower than the threshold th10 exists on the floor F20 having a higher reflectance, the human body H1. Although the human body H2 can be detected, the distance value of the differential distance image P3 is less than the threshold th10 and is mistaken for a background object.

  As described above, in the configuration of the present embodiment, by setting a threshold value for each pixel of the distance image P1, the degree of variation in the distance value of the background object is different for each section a1 in the detection area A1. In this case, there is an advantage that an optimum threshold value can be set for each section a1, and the detection sensitivity of the object can be improved.

  As another example of the above embodiment, the calculation unit 51 divides the distance image P1 into a plurality of small regions each including a plurality of adjacent pixels, and the index value for the same pixel that serves as a reference pixel for each small region. And the setting unit 52 may set a threshold value for each small region of the distance image P1. In this configuration, the detection area A1 is set with a threshold value for each section a1 corresponding to a small area composed of a plurality of adjacent pixels in the distance image P1, so that the background object is similar to the example of the third embodiment. When the degree of variation of the distance value is different for each section a1 in the detection area A1, an optimum threshold can be set for each section a1, and the detection sensitivity of the object can be improved. There is an advantage. In addition, compared to the case where a threshold value is set for each pixel of the distance image P1, there is an advantage that the amount of data processed by the threshold value setting means 5 can be reduced and the processing speed can be improved. Note that the small area is desirably set to a smaller number of pixels than the number of pixels occupied by the object in the distance image P1.

  By the way, in each of the above-described embodiments, the change in the reflectance of the floor F1 due to the replacement of the floor F1 or the like is exemplified as the environmental change in the detection region A1 that causes the variation in the distance value of the background object. The present invention is not limited to this example. For example, the degree of variation in the distance value of the background object may fluctuate due to an environmental change such as a change in the intensity of the ambient light incident on the detection area A1, so that the individual detection of the present invention If the device 1 is used, an optimum threshold value can be set according to these various environmental changes.

  In the above embodiments, the threshold setting unit 5 sets the threshold using the standard deviation as an index value. However, the present invention is not limited to this example. For example, the threshold setting unit 5 sets the threshold using a constant multiple of the standard deviation as an index value. Alternatively, the threshold value may be set using the variance as an index value.

It is a block diagram which shows the structure of Embodiment 1 of this invention. The operation example is shown, (a) is a schematic side view, (b) is a schematic diagram of a foreground distance image. It is operation | movement explanatory drawing of a distance image sensor same as the above. The operation example is shown, (a) is a schematic side view, (b) is a schematic diagram of a foreground distance image. It is explanatory drawing which shows the dispersion | variation in the distance value of a background object same as the above. It is a block diagram which shows the structure of Embodiment 2 of this invention. (A) is a schematic side view which shows operation | movement of Embodiment 3 of this invention, (b) is a schematic top view of (a), (c) is a schematic side view which shows operation | movement of a comparative example. It is a schematic block diagram which shows a prior art example. It is a block diagram of the principal part which shows a structure same as the above. The operation example is shown, (a) is a schematic side view, (b) is a schematic diagram of a foreground distance image.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Individual detector 2 Distance image sensor 3 Object detection means 5 Threshold setting means 6 Human body detection sensor 51 Calculation part 52 Setting part A1 Detection area F1, F2, F10, F20 Floor (background object)
H1, H2 Human body (object)
P1 Distance image P2 Background distance image P3 Difference distance image th1, th2, th10, th20 Threshold value σ1, σ2, σ10, σ20 Index value

Claims (4)

  A distance image sensor that generates a distance image with a pixel value as a distance value to an object existing in the detection area by imaging the detection area with a fixed position relative to a background object behind the detection area, and a background from the distance image sensor Using a distance image with a distance value to an object as a pixel value as a background distance image, a difference distance image that is a difference between the distance image obtained from the distance image sensor and the background distance image is generated. A target detection unit that detects a region having a value equal to or greater than a predetermined threshold as a target; and a threshold setting unit that sets the threshold used by the target detection unit. The threshold setting unit has different times from the distance image sensor. A calculation unit that acquires a plurality of distance images and calculates an index value indicating a degree of variation in distance value for the same pixel of the background object among the plurality of distance images; Individual detector; and a setting unit for setting automatically the in calculated the index value statistically the threshold value based on the.   The object is a human body, and is provided with a human body detection sensor that detects the presence or absence of a human body in the detection region, and the calculation unit acquires the distance image from the distance image sensor when there is no human body in the detection region. The individual detector according to claim 1, wherein the index value is calculated.   3. The calculation unit according to claim 1, wherein the calculation unit calculates the index value for each pixel of the distance image, and the setting unit sets the threshold value for each pixel of the distance image. The individual detector described. The calculation unit divides the distance image into a plurality of small regions each including a plurality of adjacent pixels, calculates the index value for the same pixel for each small region, and the setting unit calculates each small image of the distance image. The individual detector according to claim 1, wherein the threshold is set for each region.

Images