METHOD AND SYSTEM FOR DETERMINING THE POSITION OF AN OBJECT

Abstract

The invention relates to a method for determining a position of an object (24) in at least one plane of a space (20), in particular in a shop, comprising the following steps: at least two cameras (10) are used to capture respective images of the space (20), each image being an azimuthal projection of a hemisphere on the image plane; determining an azimuth of the object (24) in a respective image; assigning the position to the object (24) using the determined azimuth of the object (24) and a position of at least two previously determined reference points (22).

Description

[0001] Method and system for determining the position of an object

[0002] The subject matter concerns a system and a method for determining the position of an object in a space, in particular a plane of a space. The method and the system are used especially for the detection of people in retail stores.

[0003] Position determination, especially indoor position determination, is well known. It is known to use indoor position determination methods to determine the position of an object, for example a mobile device, in a room.

[0004] Furthermore, indoor surveillance systems based on cameras are known. The cameras capture images of a room. Objects can be detected in these images, for example, using suitable imaging techniques. It is also known to detect faces or similar features in images, such as moving images, and to track the movement of objects via a series of images. However, with known methods for monitoring a room using cameras, only the object's position relative to the camera can be determined. This is sufficient for security applications, but can be disadvantageous if absolute positional information is required, for example, to enrich augmented reality applications with location data.

[0005] For this reason, the subject matter was based on the task of precisely determining the position of an object in a space. This task is solved by a method according to claim 1 and a system according to claim 19.

[0006] It has been recognized that using information about

[0007] Using reference points and two cameras, the position of an object in a plane can be determined. One of the cameras is, in particular, a moving image camera, for example, a CCD camera or a CMOS camera.

[0008] The camera is, in particular, a high-resolution camera. It features a wide-angle lens with a very wide field of view, up to 180°. The camera preferably has a 4K resolution.

[0009] The lens is in particular a wide-angle lens or a fisheye lens or the like.

[0010] To determine position, two such cameras first capture an image of a room. The cameras are preferably time-synchronized, meaning they share at least the same time standard. This means that the timestamps of simultaneously captured images from the two cameras are compatible and, in particular, identical. This allows two images captured essentially simultaneously by the two cameras to be combined.

[0011] Each image is an azimuthal projection of the hemisphere around the camera onto the camera's image plane. Using azimuthal projection, it is possible to convert the object's azimuth coordinate into an image azimuth coordinate. Every azimuth angle of the object can be derived from an azimuth angle in the image area. Specifically, the objects are captured by the camera according to their spherical coordinates on the hemisphere and converted into Cartesian coordinates in the image plane. Thus, the azimuth of the object relative to the respective camera can be determined from each individual image.

[0012] Using azimuthal projection, it is also possible to determine the object coordinate.

[0013] to convert polar angles into image coordinates. Each polar angle of the

[0014] The object's range can be derived from its position (e.g., radius or distance from the image center) within the image area. Specifically, the objects are captured by the camera according to their spherical coordinates on the hemisphere and converted into Cartesian coordinates in the image plane.

[0015] Thus, the polar angle of the object to the respective camera can be determined from each individual image.

[0016] Now, for each camera or each image from a camera, an azimuth of the object and / or a polar angle within the image area is available. The object's position can then be determined by using the determined azimuths and / or polar angles of the object together with the position of at least two previously determined reference points.

[0017] The reference points could be, for example, two previously measured objects. The actual azimuths and / or positions of these objects could then be recorded.

[0018] Polar angles to the image axis of the cameras can be measured. It is also possible to measure the absolute distance of the reference points to the image axis of the respective cameras.

[0019] Thus, the azimuth and / or polar angle in the image of a reference point can be assigned an absolute position in space. Through trigonometric derivation, the absolute position of the object in space can then be calculated, at least from the object's azimuth in the two image planes and the positions of the reference points in the images, especially their azimuths.

[0020] To determine the two-dimensional position of an object in space, two cameras are required. When the camera's hemisphere is projected onto the image plane, an object moving along a straight line between the object and the camera will always have the same azimuth on the image plane. Only the elevation (polar angle) will change. However, if only the azimuth is evaluated, the image from a single camera is insufficient to determine the object's position along this line. By using the second image, the two lines along the two azimuths on the image plane can be intersected, allowing a position in the object's plane to be determined. This position, along with information about the reference points, can then be used to determine the object's absolute position in the plane of space.

[0021] As already mentioned, each object detected on a meridian of the hemisphere around the camera can be assigned exactly one azimuth.

[0022] The reference points can be used as standards for calibrating the camera. This makes it possible to calibrate each individual camera using the

[0023] To calibrate reference points. By absolutely dimensioning the position of the reference point, the position of the object can also be absolutely dimensioned by using the absolute dimension of the reference point, its azimuth, the azimuth of the object in the image, and / or its polar angle, as well as the polar angle of the object in the image.

[0024] The azimuthal projection is, in particular, a central projection. The projection can be gnomonic, conformal, equidistant, axis-preserving, or orthographic. Preferably, each azimuth of the image plane can be assigned an azimuth of the object plane. Preferably, each polar angle of the image plane can be assigned a polar angle of the object plane.

[0025] Assign object level.

[0026] When the object is mapped onto the image plane, an azimuth of the

[0027] Object coordinates are projected onto an azimuth corresponding to the image coordinates. Thus, the azimuth of the object coordinates can be deduced from the image coordinates, particularly the azimuth of the image coordinates. The object coordinates are preferably elevation and azimuth in the object plane. The image coordinates are also preferably azimuth and elevation in the image plane. The mapping is preferably such that the azimuth of the object plane is equal to the azimuth of the image plane.

[0028] According to one embodiment, it is proposed that the optical axis of the camera is essentially parallel to the surface normal of the plane in which the position of the object is determined. Then the image plane is preferably in

[0029] Essentially parallel to the plane in which the object moves and in which the object's position is determined. In particular, it may be necessary for the optical axes of at least the two cameras to be parallel to each other.

[0030] The cameras are preferably mounted on the ceiling of the room. The cameras preferably capture a spatial angle of more than 90°, in particular up to 180°.

[0031] It is preferred if a straight line between the cameras is assumed to be 0° azimuth for both cameras. It is also preferred if the azimuth counting direction is the same for both cameras.

[0032] The plane in which the object is located and in which its position is to be determined is preferably spanned by two axes perpendicular to the optical axis of the cameras. In particular, the plane can be described in an orthogonal coordinate system. The axes used to dimension the position of the reference point can be essentially orthogonal to each other and to the optical axis of the camera. Specifically, the actual distance of the reference point to the optical axis of the respective camera can be determined along these two axes. Using this dimension, the actual position of the object in the plane can then be calculated from the known azimuths of the reference point and the azimuths of the object.

[0033] As previously explained, the azimuth of an object can be determined in the image plane and thus also in the object plane. A vector spanning between the optical axis or the focal point of the camera and the object itself has the corresponding azimuth. If the object moves along the direction of the vector, its azimuth does not change. Therefore, the object's position in the plane cannot be determined with just one image. For this reason, it is proposed that the object's azimuth be determined in images from at least two cameras. However, the use of more than two cameras, for example, four or more, is preferred. The more information available, the more accurately the position can be determined.

[0034] The camera's resolution determines the angular resolution.

[0035] Using two images each from two different cameras, one position of the object in space can be determined, although this position is inaccurate depending on the angular resolution. Using the determined positions from at least two camera pairs, the position can be determined more precisely. For example, the calculated positions can be averaged. This allows for the determination of a geometric or quadratic mean of the calculated positions. Furthermore, this varying angular resolution can depend on the polar angles. Therefore, under

[0036] Using polar angles, the determination of the position can be adjusted by weighting the measured azimuths of the images using the determined polar angles.

[0037] According to one embodiment, it is proposed that an object center and its position be determined. For example, a human skeleton model can be used to determine the object center. So-called "head detection" algorithms are known, for instance, which can recognize a person's head in an image, and a point on the recognized head can be defined as the object center. It is also possible to use other methods.

[0038] For example, a background subtraction method could be used. In both models, the center point of the captured object can be determined computationally. This center point is used to determine the object's position.

[0039] As explained, it has been recognized that the accuracy of object detection can depend on the polar angle of the object, i.e., the object's distance from the camera. An object is detected particularly via the aforementioned polar angles.

[0040] Object recognition methods were identified, in particular a

[0041] The object's center point is determined. The further the object is from the camera, the less accurate the object detection. The object's distance depends on the polar angle at which it is detected.

[0042] This insight can be addressed as follows. From the position of an object (object center) in an image, both its azimuth and its polar angle can be derived. This is simply a matter of the imaging function. For example, an azimuth in the image represents the azimuth in space, and a distance of the object from an image center represents the polar angle.

[0043] As already explained, the camera, especially starting from the

[0044] From the camera's center point, a vector is determined in the direction of the object's center point. If the azimuth and elevation (polar angle) of an object are known, a dimensionless spatial vector can be determined that extends from the camera's center point in the direction of the azimuth and polar angles. Each vector has a root whose position is known from the camera's position. The distances and angles between the cameras are known, so the path of the vectors in space can be determined. For each of at least two cameras, such a vector is determined for each detected object. Whether an object has actually been detected depends on whether this object was captured by two cameras simultaneously, i.e., whether the vectors intersect or come close to each other.Unlike two vectors that lie only in one plane, as is the case with vectors that lie in the plane of the camera and only in the direction of the given azimuth, two solid angles do not usually intersect.

[0045] To determine whether the vectors intersect or approach each other, a virtual cylinder is placed around each vector. The radius of this cylinder depends on the polar angle of the vector. This relationship can be reciprocal, meaning that large radii are assumed for small polar angles and vice versa. This assumption about the radii stems from the understanding that the distance to the object depends on the polar angle at which the object is detected. The resolution in the image plane is higher with a smaller polar angle, meaning that the accuracy of object detection is better in areas with small polar angles than in areas with large polar angles.

[0046] The next step is to determine whether the vectors and / or their cylinders intersect in space. For this, the distance between the two vectors along their direction of propagation can be determined. In particular, the minimum distance between the vectors can be calculated. This minimum distance can then be compared to the radii of the cylinders. For example, the sum of the two radii can be determined and compared to the minimum distance between the vectors. If the cylinders intersect, the minimum distance between the vectors will be smaller than the sum of the radii. Therefore, it is checked whether the minimum distance between the vectors is smaller than the sum of the radii. If this is the case, the cylinders intersect in space.

[0047] The minimum distance between the vectors can be determined by a distance vector, which is preferably perpendicular to at least one of the vectors. The distance vector has a length that corresponds to the magnitude of the minimum distance between the vectors.

[0048] Since the vector with the largest polar angle has higher object detection accuracy, the distance vector, or its magnitude, is subdivided according to the polar angles of the two vectors. Preferably, the distance vector is subdivided proportionally to the polar angles, or rather, the radii of the cylinders. The quotient between the larger polar angle in the numerator and the sum of the polar angles in the denominator is a measure of the subdivision of the distance vector. Depending on this quotient, the distance vector can be divided into two sections separated by a point. This point on the distance vector shifts in the direction of the vector with the larger polar angle. This point is considered the object's center point.

[0049] Furthermore, it has been recognized that more than one object can be present in a single image. Generally, one would like to be able to assign the position of each individual object. For example, if an object is a person, that person might be carrying a mobile device. Using the mobile device, it may be possible to identify one object among a multitude of objects in the images.

[0050] For example, it is possible that an application is installed on a mobile device that communicates with a central computer and / or the cameras in

[0051] A communication link is established and information for position determination is exchanged. For example, it is possible for a mobile device to determine its location.

[0052] It sends out acceleration information. The central computer and / or the camera can receive acceleration information from the mobile device.

[0053] Acceleration information from the mobile device can be correlated with motion information from the images. If this information is correlated, particularly through cross-correlation, it is possible, for example, to detect which motion information from the mobile device correlates best with the motion information of which object in the image. Then, the object with the highest correlation regarding the

[0054] Motion information is combined with that of the mobile device to identify the object whose position is to be determined. In this context, it should be noted that acceleration information can be interpreted as motion information and vice versa.

[0055] First, it is possible to detect at least two objects in the respective camera images. Then, motion information about the objects can be extracted from the images. This motion information can be time-stamped. Furthermore,

[0056] If movement information is received from mobile devices that are also timestamped, a temporal correlation of the

[0057] Motion information from the images of the different objects is compared with motion information from the mobile devices of the different objects. By cross-correlating this motion information, each object in the image can be assigned to a specific object with a particular mobile device. This allows the identification of the object in the image from which the motion information was received.

[0058] As previously explained, a mobile device can be attached to the object. This mobile device could be, for example, a laptop, phone, smartphone, tablet, PDA, smart glasses, headset, or similar device. Motion information can be captured using the mobile device and / or any accelerometers and / or gyroscopes it may contain. The captured data

[0059] Motion information can be transmitted, for example, with timestamps. In particular, synchronizing the time between the mobile device or the application on the mobile device and the cameras and / or the central computer via the communication link is advantageous in order to establish a common time standard for the timestamps on the mobile device and the cameras and / or the central computer. This time standard can then be used to generate the timestamps.

[0060] This motion information is then received and analyzed centrally, or in a specific camera or in one or more of the cameras, to identify the object to which a particular mobile device is located. This allows for a correlation between a mobile device at an object and an image of that object.

[0061] To narrow down a search area within an image, it can be helpful to first define a spatial region using the mobile device. A spatial region can be an approximate position or an area on a plane. Using known indoor positioning methods, such as RFID beacons, the mobile device's approximate position within the room can be determined. A radius around this approximate position can be calculated, within which the mobile device could be located. This spatial region can be transmitted and received. The received spatial region is used to determine a more precise position within the image. Once the spatial region is known, an azimuth range can be determined for each image. Subsequently, objects can be searched for only within this azimuth range, particularly as described above.

[0062] Furthermore, anything that is independently inventive and compatible with all of this

[0063] The described features, which can be combined individually or in combination, allow for an association between an object in the image and an object in the plane using light information. A mobile device with an application installed on it can be positioned near the object. This application can emit a light signal at a defined time, determined, for example, by a central unit and / or at least one of the cameras. This light signal could, for example, be the activation and deactivation of the mobile device's flash or flash LED. The brightness of the mobile device's display can also be varied. This can involve modulation of the information, such as pulse-width modulation. At the specified time, the image from each camera can be searched for variations in the light intensity of objects.If the variation of light is modulated, different objects can be distinguished from one another in the image. If a specific pattern of light variation is detected in the image, the object located in that area can be identified as the object carrying the mobile device that is performing this modulation of the light.

[0064] For further use of the position information, it is proposed that the previously determined position of the object be sent to the mobile device located at the object. An augmented reality application could then run on the mobile device, enriching image information captured by the device's camera with additional information using the received position data.

[0065] Additional information may be location-dependent, so an exact

[0066] Positioning is important. The more accurately a position is determined, the more detailed the additional information can be. Particularly in the retail sector, very precise positional tracking with a resolution of less than 5 cm allows for the display of accurate product information for a currently viewed product.

[0067] Another aspect is a system for determining the position with two cameras, an evaluation unit and an assignment unit, which works according to the procedure described above.

[0068] In particular, the evaluation and / or assignment of the data can take place either in one of the cameras, in several cameras, or in a central computer. For example, the individual cameras can send their image data to a central computer, which then evaluates the received image data as described above. It is also possible for each camera to determine its position independently and simply send coordinates to a central computer. Alternatively, the cameras can transmit, for example, the azimuth and / or polar angle of an object, and the central computer can then calculate its position from this data, as described above.

[0069] It is also possible and independently inventive to first calibrate the system in such a way that an angular position is assigned to each object. For example, a mobile device can be held spatially in relation to an object. At this point, an object identifier is captured. Simultaneously, the at least two cameras, as described above, capture the azimuth and / or polar angle of the object or the mobile device. Thus, each camera has an azimuth and / or elevation for each object. These at least two azimuths and / or elevations for an object are linked with the

[0070] Object identifier. The object could be, for example, a commercial product.

[0071] The procedure described above can then be carried out, although the position of at least two previously determined

[0072] Reference points are not used. Rather, the assignment depends on the two azimuths and / or polar angles of a specific object. It is possible to evaluate which two azimuths and / or polar angles of the object captured by the cameras best match the azimuths and / or polar angles of which object, and to determine the corresponding object identifiers. Once each object has been assigned a specific position, its position can be precisely determined.

[0073] In the retail sector, a shop floor is usually precisely measured, and each product is identified with a product identifier and information about its absolute location.

[0074] The position within the store is stored in the level being monitored. If the object identifier is known, the absolute position can be determined. The object is explained in more detail below using a drawing showing an example. (The drawing shows:)

[0075] Fig. aa an exemplary hemisphere around a camera lens;

[0076] Fig. lb shows a mapping of the object plane onto an image plane;

[0077] Fig. 2 shows an arrangement of cameras relative to a plane in which a

[0078] Object moves;

[0079] Fig. 3 shows an arrangement for carrying out the procedure in question;

[0080] Fig. 4a-c shows a schematic representation of position determination based on the determined azimuths;

[0081] Fig. 5 an object in a room;

[0082] Fig. 6 shows communication between a mobile device and a central unit;

[0083] Fig. 7 shows the sequence of a process in question.

[0084] Fig. 1a shows a theoretical basis on which the description is built. A hemisphere 2 can be drawn around a camera. Great circles on hemisphere 2 form meridians 4. Each meridian 4 is assigned an azimuth f. The distance of an object from a camera can be given by r and the elevation by Q.

[0085] An image axis 6 of the camera runs perpendicular to an equatorial plane 8. Crucial for the following analysis is the mapping of the azimuth f onto an image plane. The image plane is preferably parallel to the equatorial plane 8. A mapping of the polar coordinates into image coordinates is shown in Fig. 1b. It can be seen that the azimuth f in the object plane can be represented by the azimuth F in the image plane. In the projection, the elevation O can be represented by the radius r.

[0086] Preferably, the object plane on hemisphere 4 is mapped onto an image plane by a central projection. In particular, the mapping is such that the azimuth f in the object plane corresponds to the azimuth F in the image plane. This can be understood as an azimuthal mapping. A slightly deviating mapping due to optics is still considered an azimuthal mapping. This can lead to distortions at the image edges.

[0087] Fig. 2 shows a camera 10 with a field of view 14. The field of view 14 is preferably between 90° and 180°. This field of view is preferably spanned at an angle between 180° and 360° around the optical axis 6. Thus, the camera 10 offers a panoramic view or a 180° view in the direction of the image plane.

[0088] Camera 10 is a wide-angle or fisheye camera and enables an azimuthal imaging of the object plane onto an image plane. Camera 10's optical axis 6 is parallel to a surface normal spanned by Cartesian axes 16 and 18. Axes 16 and 18 define the plane in which the object's position is determined.

[0089] Fig. 3 shows a room 20 with four cameras 10 arranged on its ceiling. The number of cameras can be determined by the size of the room 20.

[0090] In particular, the 10 cameras are arranged equidistant from each other, forming a camera grid. The distance between any two cameras can be, for example, 10m in the direction of the respective axes 16 and 18. Thus, with 16 cameras, an area of ​​approximately 2,500m² can be covered. 2The distance of an object in the plane spanned by axes 16 and 18, in a network of cameras spaced 10x10 meters apart, from an optical axis 6 is a maximum of approximately 7 meters. With a 4K resolution of camera 10, a spatial resolution with a blur of 2.1 cm can be achieved for each pixel.

[0091] Fig. 4 schematically shows how a position determination can be carried out. First, the cameras 10 are shown in room 20; the representation is a normal projection onto the plane spanned by axes 16 and 18 in room 20. This plane can be understood as the object plane.

[0092] The axis 18 is in particular a straight line between the cameras 10. Along the axis 18, an azimuth of zero degrees can be assumed for each camera 10.

[0093] First, reference points 22 can be measured by the cameras 18. Each reference point 22 can be assigned an azimuth a or β for each of the cameras 10. Furthermore, a dimension along axis 16 (yij) and a dimension along axis 18 (xi) can be assigned to each reference point 22, as shown in Fig. 4b.

[0094] This measurement can be the absolute distance of a reference point 22 from the optical axis 6 of each camera 10. The distance between the cameras can be known, so that the measurement along the axis 18 divides the distance between the cameras 10 into two linear units. Fig. 4b shows, for example, the detected azimuths ai, ßi of each camera 10 for a reference point 22, as well as the distances xi, yi of the reference point 22 from one of the cameras 10.

[0095] If an object 24 is detected in the image of the cameras 10, as shown in Fig. 4c, then an azimuth a can be measured at each of the cameras 10. c , ßx This object 24 will be determined. A position can be determined from these known values.

[0096] Azimuths a c , ß x the measure for x x and y x , which can be measured along axes 18 and 16 respectively, can be determined. For this purpose, a measure for y can be obtained by means of trigonometric transformations, as shown in formulas (2), (4) and (6). x and x x be determined. y x = x x ima : (2)

[0097]

[0098] where x x dependent on ci, ai and ßi and can be e.g. by a

[0099] Taylor series expansions can be used for estimation. Other series expansions are also possible. Starting from a specific x x can also be y x If the distance between the cameras is known, then the distance of object 24 from the second camera can also be determined.

[0100] As shown, the azimuth can be used to determine a c , ß x as well as the azimuth txi, ßi of the

[0101] Reference point 22 as well as the dimensions xi, yi of reference point 22 the dimensions for x x and y x Calculate the position of object 24. This makes it possible to determine the position of an object in a room using images from two cameras.

[0102] An object 24 in a room is schematically represented in Fig. 5. A center point 24a can be determined in an image of this object 24. For example, a human skeleton model can be used for this purpose. Based on the image of object 24, a skeleton model can be calculated, and a center point 24a can be assigned to this skeleton model.

[0103] Object 24 can also be separated from the image using a so-called background subtraction method. Subsequently, the center point of the area of ​​the separated object 24 can be calculated, which is then determined as the center point 24a of object 24. Other methods for determining the center point 24a of object 24 are also possible. In particular, the position of the

[0104] The center point 24a of object 24 is determined.

[0105] For this purpose, a mobile device 26, as shown in Fig. 6, can be carried on the object 24. The mobile device 26 can be, for example, a smartphone, a tablet computer, smart glasses, or the like. The mobile device 26 is in radio communication 28 with a central unit 30. The radio connection can be, for example, WLAN, ZigBee, LON, Lemonbeat, UMTS, LTE, 5G, or the like. The mobile device 26 can have a built-in accelerometer and / or tilt sensor, and the motion information recorded by it can be transmitted via the

[0106] Radio link 28 is transmitted to the central station 30.

[0107] An application can run on the mobile device 26, which can be triggered or controlled, for example, by the central unit 30 via the mobile network connection 28. Position information can also be transmitted from the central unit 30 via the

[0108] Radio connection 28 is transmitted to the mobile device 26.

[0109] Especially when an application program supporting augmented reality is running on the mobile device 26, an exact [value] is required for this application.

[0110] Position determination is helpful. Such precise position determination can be carried out according to a procedure shown in Fig. 7.

[0111] For example, if a person enters a store and an application program is started on their mobile device 26 which supports augmented reality, this application can in a step 32 transmit information to the central office 30 that position information is requested.

[0112] Subsequently, in step 34, the central unit 30 activates the position determination. For this purpose, in step 36, the application on the mobile device 26 is instructed via the radio connection 28 to transmit movement information and / or

[0113] to transmit light information. In particular, identification information can be modulated, and light information can be transmitted from the mobile device 26 according to this modulation. It is also possible for motion information to be transmitted from the mobile device 26 to the central unit 30.

[0114] Simultaneously, image acquisition can be performed in step 38. In this step, cameras 10 are controlled so that they detect objects in their images. Movements and / or variations in brightness information can be determined from the detected objects.

[0115] In step 40, the motion information from the mobile device 26 can be correlated with the motion information of various objects in the images from the cameras 10. This is achieved through a cross-correlation of these

[0116] Motion information can be used to determine which object in an image from camera 10 the mobile device 26 is positioned next to. Since the motion information in the image must correlate at least temporally with the motion information of the mobile device 26, the object around which the mobile device 26 is being moved can be determined.

[0117] Alternatively or additionally, brightness information can be evaluated. If an identifier is modulated, for example, it is possible to detect in the image which object emitted the corresponding modulated information. In particular, it is possible to establish a temporal correlation of the variation of the

[0118] to generate light information in the image and the variation of the light information on the mobile device 26 and to determine which of the objects recognizable in the image the mobile device 26 is carrying.

[0119] After the object carrying the mobile device 26 has been detected in step 40, a position determination is carried out in step 42 according to the procedures shown in Figures 4a-c. In this step, the azimuths of the identified object are determined by the cameras 10 and the absolute positions x are determined. x , y x of object 24.

[0120] Then, in step 44, this position information can be used via the

[0121] Radio connection 28 is transmitted to the mobile device 26 and made usable there for the augmented reality application.

[0122] Reference symbol list

[0123] 2 hemisphere

[0124] 4 Meridian

[0125] 6 Optical axis

[0126] 8 Equatorial plane

[0127] 10 Camera

[0128] 14 perspectives

[0129] 16, 18 axis

[0130] 20 rooms

[0131] 22 Reference point

[0132] 24 objects

[0133] 24a Object center

[0134] 26 Mobile device

[0135] 28 radio connection

[0136] 30 Central Computer

[0137] 32 Activation

[0138] 34 Object Recognition

[0139] 36. Capturing motion information or sending of

[0140] Light information

[0141] 38 Evaluating image information

[0142] 40 Correlating image information and

[0143] Motion information / light information

[0144] 42 Determining an azimuth and assigning a position to an object 44 Transmitting the position information

Claims

Patent claims 1. Method for determining the position of an object in at least one plane of a space, in particular in a shop, comprising: Capturing an image of the room by at least two cameras, where each image is an azimuthal projection of a hemisphere onto the image plane, Determining the azimuth of the object in a given image, Assigning the position to the object using the object's determined azimuth and the position of at least two previously determined reference points.

2. Method according to claim 1, characterized by that each camera is calibrated by capturing the position of at least two reference points in the plane, wherein the position of each reference point is dimensioned relative to one of at least two cameras.

3. Method according to claim 1 or 2, characterized by that the azimuthal projection is a central projection, in particular a gnomonic projection.

4. Method according to any of the preceding claims, characterized by that an azimuth of the object coordinates is projected into an azimuth of the image coordinates.

5. Method according to any of the preceding claims, characterized by that an elevation of the object coordinates is projected onto a radius of the image coordinates.

6. Method according to any of the preceding claims, characterized by an elevation of the object in a given image is determined, that the position to the object is determined using the specified elevation of the is assigned to the object.

7. Method according to any of the preceding claims, characterized by Using the azimuth and elevation of the object, a spatial vector is determined starting from each of the cameras.

8. Method according to any of the preceding claims, characterized by a minimum distance between two space vectors is determined, and in particular, the distance is put into a relationship to the respective elevations of the space vectors.

9. Method according to any of the preceding claims, characterized by that the optical axis of the camera is essentially parallel to the The surface normal of the plane in which the position of the object is determined.

10. Method according to any of the preceding claims, characterized by that the optical axes of the cameras are essentially aligned parallel to each other.

11. Method according to any of the preceding claims, characterized by that the position of the reference point is determined as the distance of the object from the optical axis of a respective camera, in particular the distance along two axes that are essentially orthogonal to the optical axis of the camera.

12. Method according to any of the preceding claims, characterized by that the azimuth and / or elevation of the object is determined in images from at least two cameras.

13. Method according to any of the preceding claims, characterized by that an object center is determined and that the position of the The object's center point is determined.

14. Method according to any of the preceding claims, characterized by that at least two objects are initially determined in the image, that subsequently, based on motion information from one of the objects received by the object and based on motion information from the objects detected in the image, the object from which the motion information was received is determined.

15. Procedure according to one of the preceding claims, characterized by that a mobile device is positioned on the object, that the mobile device captures movement information and sends the captured movement information.

16. Procedure according to one of the preceding claims, characterized by that a location area is first determined using the mobile device, that the location area is transmitted by the mobile device, and that only the location area is evaluated in the image for position determination.

17. Method according to any of the preceding claims, characterized by that light information is emitted from the mobile device, that the light information detected in the image is evaluated, and that the object on which the light information was detected is determined.

18. Method according to any of the preceding claims, characterized by The object's position is sent to the mobile device.

19. System for determining the position of an object in at least one plane of a space, in particular in a shop, comprising: at least two cameras are set up to each capture an image of the room, each image being an azimuthal projection of one hemisphere onto the image plane, an evaluation device set up to determine the azimuth of the object in a given image, An assignment device is set up to assign the position to the object using the specified azimuth of the object and a position of at least two previously determined reference points.

20. System according to claim 19, characterized by that the evaluation unit and / or the allocation unit are located in at least one of the cameras or in a central computer.