JP2011163921A - Detection device and method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a device for accurately detecting a detecting object. <P>SOLUTION: The detection device includes a selector for selecting, based on the value of height, positional information for detection used in detection processing from positional information of each observation point measured by a sensor installed downward at a predetermined height, a memory for previously storing a feature value of the detecting object set based on the shape of the detecting object, and a detector for detecting the detecting object from the positional information for detection based on the selected positional information for detection and the feature value of the detecting object stored in the memory. <P>COPYRIGHT: (C)2011,JPO&INPIT

Description

  The present invention relates to a technique for detecting a specific object in a space.

  Conventionally, a technique for detecting the presence and position of a person from a space has been proposed. For example, the technique disclosed in Patent Document 1 is as follows. First, a downward image is captured by a plurality of cameras installed at a certain height. Each observation unit acquires the position information of the person according to the corresponding camera capturing the person underneath, and identifies the face and clothes according to the other camera capturing the side of the person. Information is acquired, and feature extraction processing is performed.

  The technique disclosed in Patent Document 2 is as follows. First, a target area that is a region of a moving body and a target distance that is a distance to the moving body are detected from the captured image. Next, in the target area, a head area corresponding to the target distance when the moving body is assumed to be a person is set. Then, by storing a sample image of a predetermined size corresponding to the head region of the person, converting the head region in the captured image to a predetermined size and comparing it with the sample image, whether the moving body is a person Determine whether or not.

JP 2002-218449 A JP 2007-156939 A

  However, the technique disclosed in Patent Literature 1 is easily influenced by patterns other than the detection target, such as an object and a floor surface, and the detection accuracy is not sufficient. Further, the technique disclosed in Patent Document 2 has a problem that it is difficult to detect a detection target in an environment where detection targets (in this case, people) are dense. That is, in an environment where the detection targets are densely packed, the detection targets overlap in the field of view of the camera (sensor). Therefore, it has been difficult to detect a detection target located in the back as viewed from the camera.

  In view of the above circumstances, an object of the present invention is to provide a technique that enables a detection target to be detected with high accuracy.

  One aspect of the present invention is a detection device (for example, the detection device 20 in the embodiment), which is measured by a sensor (for example, the sensor 10 in the embodiment) installed downward from a predetermined height. A selection unit (for example, the selection unit 201 in the embodiment) that selects detection position information to be used in the detection process from the position information of the observation point based on the height value, and is set based on the shape of the detection target. On the basis of the storage unit (for example, the storage unit 202 in the embodiment) that stores the detection target feature amount in advance, the selected position information for detection, and the detection target feature amount stored in the storage unit, A detection unit (for example, the detection unit 203 in the embodiment) that detects the detection target from the detection position information.

  One embodiment of the present invention is the above-described detection device, wherein the shape of the detection target is a humanoid shape.

  One aspect of the present invention is the above-described detection device, wherein the detection unit classifies the selected position information for detection into one or a plurality of sets based on the position information of each observation point. The detection target is detected by determining whether or not each set is the detection target.

  One aspect of the present invention is the above-described detection device, wherein the selection unit includes a first height (for example, a first height H1 in the embodiment) and a second height from which the height from the floor surface is set in advance. Position information with respect to observation points located between heights (for example, the second height H2 in the embodiment) is selected as the position information for detection.

  One aspect of the present invention is the above-described detection device, wherein the first height and the second height are a height from a substantially waist position to a substantially shoulder position in the humanoid shape to be detected. It is characterized in that it is preset as a value between.

  One aspect of the present invention is the above-described detection device, wherein the selection unit has a height from a floor surface higher than a preset third height (for example, the third height H3 in the embodiment). Position information with respect to the observation point is selected as the position information for detection.

  One aspect of the present invention is the above-described detection device, wherein the first height is preset as a value between a height to a substantially shoulder position in the humanoid shape to be detected. To do.

  One aspect of the present invention is the above-described detection device, in which the selection unit detects peak observation points where the height from the floor surface is a peak value with respect to the surroundings based on the measured position information. A predetermined offset value is subtracted from the point detection unit (for example, the peak observation point detection unit 2011 in the embodiment) and the height from the floor at the peak observation point detected by the peak observation point detection unit. The calculation unit (for example, the calculation unit 2012 in the embodiment) that calculates the subtraction value by calculating the relative first height and the relative second height for each peak observation point based on the peak value and the subtraction value. And, for each predetermined range around the peak observation point, position information for the observation point located between the relative first height and the relative second height corresponding to the peak observation point, the detection position information Determination unit (for example, determination unit 2013 in the embodiment) for selecting as a, characterized in that it comprises a.

  One aspect of the present invention is the above-described detection device, in which the selection unit detects peak observation points where the height from the floor surface is a peak value with respect to the surroundings based on the measured position information. A subtraction value is calculated by subtracting a predetermined offset value from a point detection unit and a height from the floor surface at the peak observation point detected by the peak observation point detection unit, and the peak value and the subtraction are calculated. A calculation unit that calculates a relative third height for each peak observation point based on a value, and is higher than the relative third height corresponding to the peak observation point for each predetermined range around the peak observation point And a determining unit that selects position information for the observation point as the position information for detection.

  One aspect of the present invention is a detection method, in which a detection device including a storage unit that stores in advance a detection target feature amount set based on a shape of a detection target is installed downward from a predetermined height A selection step of selecting, based on a position value, position information for detection to be used in detection processing from position information measured by the sensor, and the position information for detection selected by the detection device, And a detection step of detecting the detection target from the detection position information based on the detection target feature quantity stored in the storage unit.

  In the present invention, a detection target is detected based on position information measured by a sensor installed downward from a predetermined height. For this reason, even in an environment where detection targets are densely packed, the detection targets are unlikely to overlap each other within the field of view of the sensor, so that the detection targets can be detected with high accuracy. Further, in the present invention, since the detection target is detected based on the position information, it is possible to suppress the influence of the pattern such as the floor surface and detect the detection target with high accuracy.

  Further, when the selection unit is configured to include a peak observation point detection unit, a calculation unit, and a determination unit, the relative first height and the relative second height for each detection target according to the height of the detection target. A height or a relative third height is determined. Therefore, it becomes possible to detect the detection target with high accuracy regardless of the height and state of the detection target.

It is a figure showing the system configuration of a detection system. It is a figure showing the 1st specific example of the installation example of a sensor, and the process of a selection part. It is a figure showing the installation example of a sensor, and the 2nd example of a process of a selection part. It is a schematic block diagram showing the functional structure of the selection part corresponding to the 3rd specific example of the process of a selection part. It is a figure showing the outline of the 3rd specific example of a process of a selection part. It is a figure showing the 3rd example of the installation example of a sensor, and the process of a selection part. It is a figure showing the 4th specific example of the example of installation of a sensor, and the process of a selection part. It is a figure showing the 1st specific example of a detection target feature-value. It is a figure showing the specific example of the pattern matching process by a detection part. It is a figure showing the 2nd specific example of a detection target feature-value. It is a flowchart showing the flow of a process of a detection apparatus. It is a figure showing the specific example of the pattern matching process by the detection part of a modification.

  FIG. 1 is a diagram illustrating a system configuration of the detection system 1. The detection system 1 includes two or more sensors 10 and a detection device 20. In FIG. 1, a plurality of sensors 10 are illustrated, but one sensor 10 may be provided. In FIG. 1, a plurality of sensors 10 are connected to one detection device 20, but one sensor may be connected to each detection device 20.

  The sensor 10 is a sensor that detects a three-dimensional position coordinate for each of a plurality of observation points on the surface of an object existing in a predetermined region. The sensor 10 is configured using, for example, a distance sensor that measures a distance for each observation point, and an arithmetic device that calculates position coordinates for each observation point based on the distance measured by the distance sensor and the installation position of the distance sensor. May be. By installing the distance sensor at a predetermined position and performing so-called calibration in advance, the position coordinates for each observation point can be calculated as described above.

  The sensor 10 generates information representing the position coordinates for each of a plurality of observation points (hereinafter referred to as “position information”) and outputs the information to the detection device 20. In the following description, the position coordinates are constituted by values in three axes of the X axis, the Y axis, and the Z axis. The X axis and the Y axis are axes extending along a horizontal plane, and the Z axis is an axis extending in the vertical direction.

  The sensor 10 is installed downward from a predetermined height, and detects position coordinates of an object located in a measurement region below the sensor 10. As will be described later, the detection device 20 in this embodiment uses a person as one of the detection targets. Therefore, the sensor 10 is installed at a position higher than at least a person having a general height. The sensor 10 is installed in, for example, an indoor ceiling, a column tip, or a high position on a wall.

  The detection device 20 receives position information from the sensor 10 and detects a detection target located in the measurement region of the sensor 10 based on the position information. Hereinafter, a case where the detection target has a humanoid shape will be described. The detection target having a humanoid shape is, for example, a person or a humanoid robot.

  The detection device 20 includes a CPU (Central Processing Unit) connected via a bus, a memory, an auxiliary storage device, and the like, and functions as a device including a selection unit 201, a storage unit 202, and a detection unit 203 by executing a detection program. To do. All or some of the functions of the detection device 20 may be realized using hardware such as an application specific integrated circuit (ASIC), a programmable logic device (PLD), or a field programmable gate array (FPGA).

The selection unit 201 selects the position information for detection based on the height value of each observation point included in the position information. The position information for detection is a part of position information output by the sensor 10 and is position information used in the detection process in the detection unit 203.
The storage unit 202 is configured using a storage device such as a magnetic hard disk device or a semiconductor storage device, and stores a detection target feature amount in advance. The detection target feature amount is a feature amount set in advance based on the shape of the detection target. In the present embodiment, the detection target feature amount is set based on a humanoid shape. The detection target feature amount may be set based on a general human model according to the detection process of the detection unit 203. In addition, the detection target feature amount may be generated by supervised learning processing (for example, support vector machine: SVM) using actually measured values of a plurality of persons as teacher data in accordance with the detection processing of the detection unit 203.

The detection unit 203 detects a detection target from the detection position information based on the detection position information selected by the selection unit 201 and the detection target feature amount.
FIG. 2 is a diagram illustrating an installation example of the sensor 10 and a first specific example of processing of the selection unit 201. In FIG. 2, two sensors 10 are installed on the ceiling surface 32 in a space having a floor surface 31 and a ceiling surface 32. In addition, two people 30 stand upright on the floor 31. In the first specific example, the selection unit 201 obtains position information for the observation point where the height from the floor surface 31 is located between the preset first height H1 and second height H2 as the detection position. Select as information. The selection unit 201 stores the values of the first height H1 and the second height H2 in advance. Note that when the height value (Z value) of the position coordinates at each observation point is different from the height value from the floor surface 31, the selection unit 201 preliminarily calculates the height of the position coordinates and the height from the floor surface 31. Remember the difference. Then, the selection unit 201 selects the detection position information based on the difference value stored in advance.

  In the case of FIG. 2, the first height H1 is set to the height of a general person's shoulder. The second height H2 is set to the height of a general human waist. The first height H1 and the second height H2 do not need to be limited to the shoulder height and the waist height, respectively, and any height between approximately the shoulder height and approximately the waist height. May be set. A general human model is appropriately set according to a person to be detected. For example, when the detection system 1 is installed in an elementary school, the detection target is an elementary school student, and thus a general human model is represented using statistical values such as the average value of the physique of elementary school students.

FIG. 3 is a diagram illustrating a second specific example of the installation example of the sensor 10 and the processing of the selection unit 201. The installation location of the sensor 10 in FIG. 3 is the same as in FIG. In the second specific example, the selection unit 201 selects position information for an observation point whose height from the floor 31 is higher than a predetermined third height H3 as detection position information. The selection unit 201 stores the value of the third height H3 in advance. Note that when the height value (Z value) of the position coordinates at each observation point is different from the height value from the floor surface 31, the selection unit 201 preliminarily calculates the height of the position coordinates and the height from the floor surface 31. Remember the difference. Then, the selection unit 201 selects the detection position information based on the difference value stored in advance.
In the case of FIG. 3, the third height H3 is set to the height of a general person's shoulder. The third height H3 need not be limited to the height of the shoulder, and may be set to any height as long as it is approximately the height of the shoulder.

  With this configuration, position coordinates at an observation point above the shoulder can be selected as detection position information. The shape above the human shoulder has a characteristic shape compared to other objects. In addition, the shape above the person's shoulder is less likely to change because it is less likely to be covered with clothes compared to the body. Therefore, it is possible to improve the accuracy of detecting a humanoid shape.

  FIG. 4 is a schematic block diagram showing the functional configuration of the selection unit 201 corresponding to the third specific example of the processing of the selection unit 201. In the case of FIG. 4, the selection unit 201 includes a peak observation point detection unit 2011, a calculation unit 2012, and a determination unit 2013. As described above, all or some of the functions of the selection unit 201 may be realized using hardware such as ASIC, PLD, or FPGA.

FIG. 5 is a diagram illustrating an outline of a third specific example of the process of the selection unit 201. Hereinafter, a third specific example of the processing of the selection unit 201 will be described together with the configuration of each functional unit illustrated in FIG.
The peak observation point detection unit 2011 detects an observation point whose height from the floor 31 is a peak value with respect to the surroundings (hereinafter referred to as “peak observation point”). Specifically, the peak observation point is a point where the height value (Z value) becomes a peak (maximum value) relative to the observation points located on the periphery in the horizontal plane. In FIG. 5, the axis extending in the horizontal direction is an axis representing coordinates on the horizontal plane represented by the X axis and the Y axis, and the axis extending in the vertical direction is an axis representing the height represented by the Z axis. In FIG. 5, three points 60 are peak observation points.

  The calculation unit 2012 stores an offset value (Offset) in advance. The calculation unit 2012 calculates a subtraction value V (V1 to V3) for each peak observation point 60 by subtracting the offset value from the peak value of each peak observation point 60. Then, the calculation unit 2012 calculates the relative first height and the relative second height for each peak observation point 60 based on the peak value and the subtraction value V. For example, the relative first height may coincide with the peak value, and the relative second height may coincide with the subtraction value V. Further, the relative first height may be calculated by adding a value (for example, 90%) of a predetermined ratio of the width between the peak value and the subtraction value V to the subtraction value V. Similarly, the relative second height may be calculated by adding a value (for example, 10%) of a predetermined ratio of the width between the peak value and the subtraction value V to the subtraction value V. The specific calculation method of the relative first height and the relative second height may be another method.

  The determination unit 2013 sets a predetermined range for each peak observation point 60, and the position information of the observation point where the height value is located between the relative first height and the relative second height within each predetermined range, Select as detection position information. The predetermined range is a range on the horizontal plane set based on the position of the peak observation point 60 on the horizontal plane. For example, the predetermined range is set as a range within a circle having a radius R1 with the peak observation point 60 as the center. In FIG. 5, ranges A <b> 1 to A <b> 3 set on the horizontal plane are ranges set as ranges within a circle having a radius R <b> 1 with each peak observation point 60 as the center.

  FIG. 6 is a diagram illustrating a third specific example of the installation example of the sensor 10 and the processing of the selection unit 201. The installation location of the sensor 10 in FIG. 6 is the same as in FIG. In the third specific example, as described with reference to FIG. 5, the selection unit 201 selects the position information for detection according to the peak observation point 60.

  Hereinafter, the third specific example of the process of the selection unit 201 will be described in more detail with reference to FIG. In FIG. 6, on the floor surface 31, there are an upright adult 30-1, a bent adult 30-2, an upright child 30-3, an upright humanoid robot 30-4, and a pole 40-1 extending in the vertical direction. is there. First, the peak observation point detection unit 2011 detects the peak observation points 60-1 to 60-5. Next, the calculation part 2012 subtracts an offset value (Offset) for every peak observation point 60-1 to 60-5, and calculates a subtraction value. And the calculation part 2012 calculates a relative 1st height and a relative 2nd height for every peak observation point 60-1 to 60-5 based on a subtraction value and a peak value. In the example of FIG. 6, the relative first height is the same value as the peak value, and the relative second height is the same value as the subtraction value. And the determination part 2013 selects the positional information for a detection based on relative 1st height and relative 2nd height for every peak observation point 60-1 to 60-5.

  FIG. 7 is a diagram illustrating a fourth specific example of the installation example of the sensor 10 and the processing of the selection unit 201. The installation location of the sensor 10 in FIG. 7 is the same as that in FIG. In the fourth specific example, the selection unit 201 includes a peak observation point detection unit 2011, a calculation unit 2012, and a determination unit 2013 as shown in FIG. However, in the fourth specific example, the calculation unit 2012 calculates the relative third heights H4-1 to H4-5 based on the peak value and the subtraction value. The specific calculation method of the relative third height may be the same as the relative first height or the relative second height, and other methods may be adopted. In the example of FIG. 7, the relative third height is the same value as the subtraction value. In the fourth specific example, the determination unit 2013 selects position information for an observation point higher than the relative third height as detection position information.

  FIG. 8 is a diagram illustrating a first specific example of the detection target feature amount. FIG. 8 shows a cut surface at a predetermined height between the waist and shoulders of the human body and a plurality of points located on the contour of the cut surface. FIG. 8A represents the outline of the cut surface and a plurality of points, and FIG. 8B represents only the plurality of points. The cut surface at a predetermined height between the waist and shoulders of the human body is almost the same at any height, and consists of both arms and torso parts. For example, the coordinates of a plurality of points set as shown in FIG. 8B are stored in advance in the storage unit 202 as detection target feature amounts. Such a detection target feature amount is generated by, for example, generating a general human model in advance, generating a cut surface from the generated model, and setting a plurality of points positioned on the contour line of the cut surface May be. Further, detection is performed by setting a plurality of points positioned on the outline of the cut surface for each measured value of the shape of a plurality of people, and using these as teacher data for supervised learning processing according to the processing of the detection unit 203 A target feature amount may be generated.

  In this case, the detection unit 203 projects the position information for detection selected by the selection unit 201 onto the XY plane, and performs pattern matching with the detection target feature amount. FIG. 9 is a diagram illustrating a specific example of pattern matching processing by the detection unit 203. In FIG. 9, a rectangle 200 represents a binary XY plane image generated by projecting the detection position information onto the XY plane. In the XY plane image, “1” is given to the point where the projected observation point exists, and “0” is given to the point where the observation point does not exist. In FIG. 9, a pattern 202A represents a detection target feature amount. In the XY plane image, the detection unit 203 shifts the detection target feature amount from the upper left in the arrow 50-1 direction by a predetermined width. Each time the detection target feature amount is shifted, the detection unit 203 calculates a difference amount from the XY plane image to determine whether or not the detection target is located at that position. Then, the detection unit 203 shifts the pattern 202A from the left end in the direction of the arrow 50-2 in the row shifted in the Y-axis direction with a predetermined width after the pattern 202A touches the right end by shifting in the direction of the arrow 50-1. Similar processing is performed. The detection unit 203 repeatedly executes the above processing until the pattern 202A touches the lower right end of the Y axis.

  Next, a process performed each time the detection unit 203 shifts the pattern 202A will be described. The detection unit 203 associates the closest point among the points included in the detection target feature amount for each coordinate having “1” in the XY plane image. Next, the detection unit 203 calculates a statistic such as a total value or an average value of the distances between the associated points as a difference amount. Then, when the difference amount is smaller than a preset threshold value, the detection unit 203 determines that the detection target is located at the current position of the pattern 202A. On the other hand, when the difference amount is equal to or larger than the threshold, the detection unit 203 determines that the detection target is not located at the position of the pattern 202A.

  FIG. 10 is a diagram illustrating a second specific example of the detection target feature amount. FIG. 10 shows a cut surface by a predetermined vertical plane in a shape from the vicinity of the shoulder of the human body to the tip of the head, and a plurality of points positioned on the outline of the cut surface. FIG. 10A shows the outline of the cut surface and a plurality of points, and FIG. 10B shows only the plurality of points. For example, the coordinates of a plurality of points set as shown in FIG. 10B are stored in advance in the storage unit 202 as detection target feature amounts. Such a detection target feature amount is generated by, for example, generating a general human model in advance, generating a cut surface from the generated model, and setting a plurality of points positioned on the contour line of the cut surface May be. Further, detection is performed by setting a plurality of points positioned on the outline of the cut surface for each measured value of the shape of a plurality of people, and using these as teacher data for supervised learning processing according to the processing of the detection unit 203 A target feature amount may be generated.

In this case, the detection unit 203 projects the detection position information selected by the selection unit 201 onto a predetermined vertical plane, and performs pattern matching with the detection target feature amount. Specifically, the detection unit 203 extracts the coordinates of each point located on the contour line generated by projection on the coordinate axis set on the vertical plane, and performs pattern matching with the detection target feature amount. Pattern matching between the projected detection position information and the detection target feature amount is performed in the same manner as the processing shown in FIG.
In addition, the detection unit 203 may be configured to store a plurality of vertical plane parameters in advance and perform the above-described processing on each vertical plane. With this configuration, the detection target can be detected with high accuracy without depending on the direction of the detection target.

  FIG. 11 is a flowchart showing a process flow of the detection apparatus 20. First, the selection unit 201 receives position information from the sensor 10 (step S101). Next, the selection unit 201 selects detection position information from the received position information (step S102). Next, the detection unit 203 reads the detection target feature amount from the storage unit 202 (step S103). Then, the detection unit 203 starts matching between the detection position information and the detection target feature amount (step S104).

  The detection unit 203 calculates the difference amount and determines whether it is less than a preset threshold value (step S105). When the calculated difference amount is less than the threshold (step S105—YES), the detection unit 203 determines that the detection target is located at the current pattern position (step S106). On the other hand, when the calculated difference amount is equal to or greater than the threshold (step S105—NO), the detection unit 203 determines that the detection target is not located at the current pattern position (step S107).

  Next, the detection unit 203 determines whether or not the determination in step S105 has been performed at all positions (step S108). In the case of the example in FIG. 8, this process is determined based on whether or not the pattern 202A is shifted from the upper left end to the lower right end in the XY plane image. Further, in the case of the example in FIG. 10, the determination is made based on whether or not the pattern 202A is shifted from the upper left end to the lower right end in all vertical planes. If not determined (step S108—NO), the detection unit 203 shifts the position of the pattern 202A by a predetermined width (step S109), and returns to the process of S105. On the other hand, when the determination has been made at all positions (step S108—YES), the detection unit 203 outputs the determination results of step S106 and step S107 that have been repeatedly executed up to that point as detection results (step S110). .

In the detection device 20 configured in this way, a detection target is detected based on position information measured by the sensor 10 installed downward from a predetermined height. For this reason, even in an environment where the detection targets are densely packed, the detection targets do not overlap each other within the field of view of the sensor 10, so that the detection targets can be detected with high accuracy.
In addition, since the detection device 20 detects the detection target based on the position information, it is possible to suppress the influence of the pattern such as the floor surface and detect the detection target with high accuracy.

  Further, when the selection unit 201 performs the selection by the process of the third specific example or the fourth specific example, it is possible to detect the detection target with high accuracy regardless of the height and state of the detection target. Specifically, the selection range of the position information for detection is set for each object according to the peak observation point 60 of each object. Therefore, for example, in the case of FIG. 6, the humanoid robot 30-4 may be an upright adult 30-1, a bent adult 30-2, or a short child 30-3. Even so, the detection position information is appropriately selected.

<Modification>
The sensor 10 may be configured as the distance sensor alone. In this case, the arithmetic device that calculates the position information for each observation point based on the distance measured by the distance sensor and the installation position of the distance sensor may be installed as a separate device from the sensor 10. Further, the detection device 20 may be configured to further include a calculation unit that performs processing corresponding to the calculation device.

  In FIG. 8 and FIG. 10, the detection target feature amount is a point set in the contour of a predetermined cut surface, and thus is represented by two-dimensional coordinates. However, the detection target feature amount need not be limited to that represented by two-dimensional coordinates, and may be configured to be represented by three-dimensional coordinates. In this case, the detection unit 203 performs pattern matching with the detection target feature amount based on the three-dimensional values of the X coordinate, the Y coordinate, and the Z coordinate of the detection position information.

  FIG. 12 is a diagram illustrating a specific example of pattern matching processing performed by the detection unit 203 according to the modification. The detection unit 203 may be configured to classify the detection position information into one or a plurality of sets based on the position coordinates of each observation point. Such classification processing is realized by applying an existing clustering technique. Then, the detection unit 203 performs pattern matching with the detection target feature amount for each set. In FIG. 12, a set 200-1 and a set 200-2 are generated as a result of the classification process. In this case, the detection unit 203 does not shift the pattern 202A by a predetermined width from the upper left end to the lower right end as shown in FIG. 9, but the place where the classified sets (200-1 and 200-2) are located. Pattern matching only in With this configuration, the time required for pattern matching can be reduced.

  The embodiment of the present invention has been described in detail with reference to the drawings. However, the specific configuration is not limited to this embodiment, and includes designs and the like that do not depart from the gist of the present invention.

DESCRIPTION OF SYMBOLS 10 ... Sensor, 20 ... Detection apparatus, 201 ... Selection part, 202 ... Memory | storage part, 203 ... Detection part, 2011 ... Peak observation point detection part, 2012 ... Calculation part, 2013 ... Determination part

Claims (10)

From the position information of each observation point measured by a sensor installed downward from a predetermined height, a selection unit that selects detection position information used in the detection process based on the height value;
A storage unit that stores in advance a detection target feature amount set based on the shape of the detection target;
A detection unit that detects the detection target from the detection position information based on the selected detection position information and the detection target feature quantity stored in the storage unit;
A detection device comprising:   The detection apparatus according to claim 1, wherein the shape of the detection target is a human shape. The detection unit classifies the selected position information for detection into one or a plurality of sets based on the position information of each observation point, and determines whether each classified set is the detection target. The detection apparatus according to claim 2, wherein the detection target is detected by the detection.   The selection unit selects, as the detection position information, position information with respect to an observation point located between a first height and a second height whose height from the floor is set in advance. The detection device according to claim 2 or 3.   The first height and the second height are preset as values between a height from a substantially waist position to a substantially shoulder position in the humanoid shape to be detected. Item 5. The detection device according to Item 4.   The said selection part selects the positional information with respect to the observation point whose height from a floor surface is higher than preset 3rd height as said positional information for a detection. Detection device.   The detection device according to claim 6, wherein the first height is set in advance as a value between a height to a substantially shoulder position in the humanoid shape to be detected. The selection unit includes:
From the measured position information, a peak observation point detection unit for detecting an observation point whose height from the floor is a peak value with respect to the surroundings;
By calculating a subtraction value by subtracting a predetermined offset value to the height from the floor at the peak observation point detected by the peak observation point detection unit, based on the peak value and the subtraction value, A calculation unit for calculating a relative first height and a relative second height for each peak observation point;
For each predetermined range around the peak observation point, position information for the observation point located between the relative first height and the relative second height corresponding to the peak observation point is used as the detection position information. A decision unit to select;
The detection apparatus according to claim 2, further comprising: The selection unit includes:
From the measured position information, a peak observation point detection unit for detecting an observation point whose height from the floor is a peak value with respect to the surroundings;
By calculating a subtraction value by subtracting a predetermined offset value to the height from the floor at the peak observation point detected by the peak observation point detection unit, based on the peak value and the subtraction value, A calculation unit for calculating a relative third height for each peak observation point;
For each predetermined range around the peak observation point, a determination unit that selects position information for an observation point higher than the relative third height corresponding to the peak observation point as the detection position information;
The detection apparatus according to claim 2, further comprising: A detection device including a storage unit that stores in advance a detection target feature amount set based on the shape of the detection target is detected from position information measured by a sensor installed downward from a predetermined height. A selection step for selecting the position information for detection used in the step based on the position value;
A detection step in which the detection device detects the detection target from the detection position information based on the selected detection position information and the detection target feature quantity stored in the storage unit;
A detection method comprising:

Images