Search

EP-4736119-A1 - METHOD FOR LOCALIZATION OF AT LEAST ONE OBJECT

EP4736119A1EP 4736119 A1EP4736119 A1EP 4736119A1EP-4736119-A1

Abstract

There is provided a method for localization of at least one object in a main reference system. The method comprises receiving first and second image data obtained by a first camera and by a second camera, respectively. The first and second cameras have respective point of views of portions of the main reference system, wherein the respective point of views are at angle with regards to each other. The method comprises detecting a first object based on the received first image data, and calculating a first pyramid which envelops the detected first object. The method further comprises detecting the first object based on the received second image data, and calculating a second pyramid which envelops the detected first object. The method further comprises calculating an intersection of the first pyramid and the second pyramid, and determining a location of the first object based on the calculated intersection.

Inventors

  • PULERI, MARZIO
  • PEPE, Teresa

Assignees

  • Telefonaktiebolaget LM Ericsson (publ)

Dates

Publication Date
20260506
Application Date
20230630

Claims (17)

  1. 1. A method (1000) for localization of at least one object (t1 , t2) in a main reference system (RS0), the method comprising: receiving (111) first image data obtained by a first camera (11) located at a first position in the main reference system, wherein the first camera has a first point of view of at least a portion of the main reference system, and wherein the first point of view has an origin (11’) at the first position; receiving (121) second image data obtained by a second camera (12) located at a second position in the main reference system, wherein the second camera has a second point of view of at least another portion of the main reference system, and wherein the second point of view has an origin at the second position and is at angle with regards to the first point of view; detecting (112) a first object (t1) of the at least one object based on the received first image data; calculating (113) a first pyramid which originates from the origin of the first camera and which envelops the detected first object; detecting the (122) first object based on the received second image data; calculating (123) a second pyramid which originates from the origin of the second camera and which envelops the detected first object; calculating (140) an intersection of the first pyramid and the second pyramid; determining (150) a location of the first object based on the calculated intersection.
  2. 2. The method according to claim 1 , wherein the first point of view corresponds to a first reference system (RS1) and the second point of view corresponds to a second reference system (RS3); and wherein positions on the main reference system, the first reference system and the second reference system can be mapped to each other by using affine transformations.
  3. 3. The method according to claim 2, wherein using affine transformations comprises using Denavit-Hartenberg rules convention.
  4. 4. The method according to any one of claims 1 to 3, wherein the determining a location of the object further comprises: calculating a Chebyshev center of a polyhedron defined by the intersection of the first pyramid and the second pyramid.
  5. 5. The method according to any one of claims 1 to 4, wherein detecting the object based on the received first and/or second image data comprises using a temporal foreground integration method.
  6. 6. The method according to any one of the preceding claims, wherein the at least one object comprises at least two objects, and wherein the method further comprises: detecting (212) a second object (t2) of the at least two objects based on the received first image data; calculating (213) another first pyramid which originates from the origin of the first camera and which envelops the detected second object; detecting (222) the second object based on the received second image data; calculating (223) another second pyramid which originates from the origin of the second camera and which envelops the detected second object; calculating (240) an intersection of the other first pyramid and the other second pyramid; determining (250) a location of the second object based on the calculated intersection.
  7. 7. The method according to claim 6, wherein the method further comprises: determining (260) that an intersection of a pyramid which envelops the first object and a pyramid which envelops the second object does not relate to the any one the at least two objects.
  8. 8. The method according to claim 7, wherein determining that an intersection of a pyramid which envelops the first object and a pyramid which envelops the second object does not relate to the any one the at least two objects comprise determining that a location of said intersection is outside of a predefined portion of the main reference system.
  9. 9. The method according to any one of the preceding claims, wherein the method further comprises: receiving (131) third image data obtained by a third camera (13) located at a third position in the main reference system, wherein the third camera has a third point of view of at least a third portion of the main reference system, and wherein the third point of view has an/its origin at the third position and is at angle with regards to at least one of the first point of view and the second point of view; detecting (132) the object based on the received third image data; calculating (133) a third pyramid which originates from the origin of the third camera and which envelops the detected object; and calculating (143) a second intersection of the third pyramid and the first pyramid or the second pyramid.
  10. 10. The method according to any one of the preceding claims, wherein the method further comprises: detecting (310) a predetermined object based on the received first image data, wherein a position of the predetermined object in the main reference system is predetermined and wherein a size and shape of the predetermined object is predetermined; calculating (320) an angle and a distance between the first camera and the predetermined object; updating (330) the first position in the main reference system, at which the first camera is located, based on the calculated angle and distance between the first camera and the predetermined object.
  11. 11. The method according to claim 10, wherein the predetermined object comprises a scannable code or a reference image.
  12. 12. The method according to any one of the preceding claims, wherein the main reference system corresponds to an area of interest.
  13. 13. A computer (400) for localization of at least one object in a main reference system, the computer comprising processing circuitry (402) configured to cause the computer to: receive first image data obtained by a first camera located at a first position in the main reference system, wherein the first camera has a first point of view of at least a portion of the main reference system, and wherein the first point of view has an/its origin at the first position; receive second image data obtained by a second camera located at a second position in the main reference system, wherein the second camera has a second point of view of at least another portion of the main reference system, and wherein the second point of view has an origin at the second position and is at angle with regards to the first point of view; detect a first object of the at least one object based on the received first image data; calculate a first pyramid which originates from the origin of the first camera and which envelops the detected first object; detect the first object based on the received second image data; calculate a second pyramid which originates from the origin of the second camera and which envelops the detected first object; calculate an intersection of the first pyramid and the second pyramid; determine a location of the first object based on the calculated intersection.
  14. 14. The computer according to claim 13, wherein the processing circuitry is further configured to cause the computer to perform a method according to any one of claims 2 to 12.
  15. 15. A system (500) for localization of at least one object in a main reference system (RSO), wherein the system comprises: a first camera (11) located at a first position in the main reference system; a second camera (12) located at a second position in the main reference system; and a computer (400) according to claim 13 or 14, wherein the computer is communicatively connected to the first camera and the second camera.
  16. 16. A computer program (401), comprising instructions which when executed on a processor (402)of a computer (400) causes the computer (400) to perform a method according to any of claims 1 to 12.
  17. 17. A computer program product (404) which comprises a computer readable storage medium (403) on which a computer program (401) according to claim 16 is stored.

Description

METHOD FOR LOCALIZATION OF AT LEAST ONE OBJECT TECHNICAL FIELD The disclosure herein relates to a computer, and a system, and methods thereof. In particular, the embodiments relate to enabling localization of at least one object. Computer programs and a computer program product are also disclosed. BACKGROUND Precise positioning of objects has been an issue for automated logistics. Several different technologies are used for this purpose each with their own advantages and disadvantages compared to each other. For example, in some cases, small radio terminals can be placed on a freight, thereby relying on Global Position System (GPS) positioning when outdoors or Wi-Fi, Long Term Evolution (LTE), or 5th Generation (5G) triangulation, which is available both indoors and outdoors. Bluetooth Low Energy (BLE) beacons with terminals on objects and vehicles are also used, for example these are used in certain mines and mining operations. Another way of detecting the position of a freight can be the use of stereovision. In this case the object position is determined using epipolar geometry or triangulation. The problem in these cases is the amount of devices needed, their overall cost, maintenance and as well as the accuracy that can be achieved utilizing these methods. Small GPS devices can have an accuracy of several meters and to increase the accuracy more expensive GPS devices or the integration with other radio reference signals may be needed. There are several techniques that can be used to detect the position location of an object. A first method is a Signal strength based, this utilizes a Received Signal Strength Indication (RSSI) localization technique based on measuring signal strength from a client device to several access points; it typically has an accuracy of about 2 to 4 meters. A second method is Fingerprinting based, again utilizing RSSI, and the second method relies on the creation of a map of signal strengths received by access points. Wherein, the position is inferred taking into consideration the measured vectors. The accuracy is between 0.6 and 1.3 m. Its main disadvantage is that any change to the environment, for example, adding or removing object within the environment, can modify the fingerprint that corresponds to each location, therefore requiring an update to the fingerprint database. A third method is angle of arrival based, this technique is generally more accurate than the other two, but requires special hardware, such as an array of six to eight antennas. A camera may be used to determine the position of an object, normally using epipolar geometry. Epipolar geometry may use stereovision utilizing two cameras that are able capture an image of the object from the same side. The two cameras may be placed at a specified distance apart so that they view the object with different angles, yet to apply epipolar geometry the two images must overlap. This means that the two cameras must be placed within a certain range from each other. The closer they are the lower is the position resolution as far as the distance from the object increases. In general, the error at 50 m is a few meters. If any one of the two cameras observing the object is masked by some obstacle, then the stereovision may not work. The article Abandoned Object Detection via Temporal Consistency Modeling and Back-Tracing Verification for Visual Surveillance. IEEE TRANSACTIONS ON INFORMATION FORENSICS AND SECURITY, VOL. 10, NO. 7, JULY 2015, by Kevin Lin, S.-C. C.-S.-T.-P. (2015), discloses methods for automatic detection of objects. SUMMARY An object of the invention is to provide an improved localization of objects using a plurality of cameras. According to a first aspect of the invention there is provided a method for localization of at least one object in a main reference system. The method comprises receiving first image data obtained by a first camera located at a first position in the main reference system. The first camera has a first point of view of at least a portion of the main reference system. The first point of view has an origin (i.e. its origin) at the first position. The method further comprises receiving second image data obtained by a second camera located at a second position in the main reference system. The second camera has a second point of view of at least another portion of the main reference system. The second point of view has an origin (i.e. its origin) at the second position and is at angle with regards to the first point of view. The method further comprises detecting a first object of the at least one object based on the received first image data, calculating a first pyramid which originates from the origin of the first camera and which envelops the detected first object. The method further comprises detecting the first object based on the received second image data, and calculating a second pyramid which originates from the origin of the second camera and which envelops the detected first ob