WO2022230738A1 - 物体追跡装置及び物体追跡方法 - Google Patents
物体追跡装置及び物体追跡方法 Download PDFInfo
- Publication number
- WO2022230738A1 WO2022230738A1 PCT/JP2022/018330 JP2022018330W WO2022230738A1 WO 2022230738 A1 WO2022230738 A1 WO 2022230738A1 JP 2022018330 W JP2022018330 W JP 2022018330W WO 2022230738 A1 WO2022230738 A1 WO 2022230738A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- processor
- detection target
- kalman filter
- kalman
- object tracking
- Prior art date
Links
- 238000000034 method Methods 0.000 title claims abstract description 91
- 238000001514 detection method Methods 0.000 claims abstract description 80
- 230000008569 process Effects 0.000 claims abstract description 67
- 230000007717 exclusion Effects 0.000 claims abstract description 27
- 238000003384 imaging method Methods 0.000 description 50
- 238000012545 processing Methods 0.000 description 25
- 238000010586 diagram Methods 0.000 description 14
- 230000003287 optical effect Effects 0.000 description 13
- 230000006870 function Effects 0.000 description 5
- 238000004891 communication Methods 0.000 description 4
- 230000033001 locomotion Effects 0.000 description 4
- 239000004065 semiconductor Substances 0.000 description 4
- 238000005401 electroluminescence Methods 0.000 description 3
- 238000012986 modification Methods 0.000 description 3
- 230000004048 modification Effects 0.000 description 3
- 230000008859 change Effects 0.000 description 2
- 238000010276 construction Methods 0.000 description 2
- 238000012937 correction Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- 238000013138 pruning Methods 0.000 description 2
- 238000003491 array Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 239000011230 binding agent Substances 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 230000000295 complement effect Effects 0.000 description 1
- 230000001186 cumulative effect Effects 0.000 description 1
- 230000001934 delay Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 239000004973 liquid crystal related substance Substances 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 230000002093 peripheral effect Effects 0.000 description 1
- 238000003908 quality control method Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/20—Analysis of motion
- G06T7/277—Analysis of motion involving stochastic approaches, e.g. using Kalman filters
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30248—Vehicle exterior or interior
- G06T2207/30252—Vehicle exterior; Vicinity of vehicle
Definitions
- the present disclosure relates to an object tracking device and an object tracking method.
- Patent Document 1 processes a video signal output from an in-vehicle camera that captures images around the vehicle, detects the presence or absence of approaching vehicles and pedestrians, and adds square frame marks to the approaching vehicles and pedestrians.
- An object tracking device includes: an input interface for obtaining sensor data; a processor that detects a detection target from the sensor data and tracks the detection target using a Kalman filter associated with each of the detection target and the observed value; an output interface for outputting a detection result of the detection target,
- the processor A first process of selecting, as an exclusion candidate Kalman filter that can be excluded from the association, one of the plurality of Kalman filters associated with the same detection target or observation value that has a low probability; and a second process of excluding the exclusion candidate Kalman filter that satisfies the initialization condition from the association and performing initialization.
- An object tracking method includes: obtaining sensor data; Detecting a detection target from the sensor data, and tracking the detection target using a Kalman filter associated with each of the detection target and the observed value; and outputting a detection result of the detection target, Tracking the detection target includes: A first process of selecting, as an exclusion candidate Kalman filter that can be excluded from the association, one of the plurality of Kalman filters associated with the same detection target or observation value that has a low probability; and a second process of excluding the exclusion candidate Kalman filter that satisfies the initialization condition from the association and performing initialization.
- FIG. 1 is a block diagram showing a schematic configuration of an object tracking system including an object tracking device according to one embodiment.
- FIG. 2 is a diagram showing a vehicle equipped with the object tracking system of FIG. 1 and a detection target.
- FIG. 3 is a flowchart showing an example of processing for tracking an image of an object on a moving image.
- FIG. 4 is a diagram showing an example of an image of an object on a moving image.
- FIG. 5 is a diagram for explaining the relationship between an object in real space, an image of the object in a moving image, and mass points in virtual space.
- FIG. 6 is a diagram showing an example of movement of mass points in virtual space.
- FIG. 7 is a diagram for explaining data association.
- FIG. 8 is a diagram illustrating a hierarchical structure of tracked object ID management.
- FIG. 9 is a diagram showing error ellipses of overlapping Kalman filters associated with the same observation point.
- FIG. 10 is a diagram showing error ellipses of overlapping Kalman
- FIG. 1 is a block diagram showing a schematic configuration of an object tracking system 1.
- FIG. Object tracking device 20 according to an embodiment of the present disclosure is included in object tracking system 1 .
- object tracking system 1 includes imaging device 10 , object tracking device 20 , and display 30 .
- the object tracking system 1 is mounted on a vehicle 100, which is an example of a moving body, as illustrated in FIG.
- the object tracking device 20 acquires moving images from the imaging device 10 as sensor data. That is, in the present embodiment, the sensor used for detecting the detection target is the imaging device 12 that captures visible light and is provided in the imaging device 10 .
- the object tracking system 1 is not limited to the configuration shown in FIG.
- the object tracking system 1 can include a device different from the imaging device 10 as long as it detects a detection target.
- the object tracking system 1 may be configured to include a measuring device that measures the distance to the detection target from the reflected wave of the irradiated laser light instead of the imaging device 10 .
- the object tracking system 1 may be configured to include a detection device having a millimeter wave sensor instead of the imaging device 10 .
- the object tracking system 1 may be configured to include an imaging device 10 including an imaging device 12 that captures light outside the visible light range.
- the object tracking system 1 is mounted on a moving object, and detects an object 40 (see FIG. 2) around the moving moving object.
- the object tracking system 1 is not limited to being mounted on a mobile object.
- the object tracking system 1 may be used in a facility such as a factory to detect employees, transport robots, products, and the like.
- the object tracking system 1 may be used in a welfare facility for the elderly, etc., and may detect elderly people and staff members in the room.
- the object tracking system 1 not only tracks objects for the safety of traveling or behavior, but also tracks objects for efficiency of work, quality control, or improvement of productivity at agricultural and industrial sites, for example. may be performed.
- the object that is the detection target of the object tracking device 20 includes not only objects such as mobile objects but also people.
- the x-axis direction is the width direction of the vehicle 100 on which the imaging device 10 is installed.
- the positive direction of the y-axis is the direction in which the vehicle 100 moves backward.
- the x-axis direction and the y-axis direction are directions parallel to the road surface on which the vehicle 100 is located.
- the z-axis direction is a direction perpendicular to the road surface.
- the z-axis direction can be referred to as the vertical direction.
- the x-axis direction, y-axis direction, and z-axis direction are orthogonal to each other.
- the x-axis direction, y-axis direction, and z-axis direction are not limited to this.
- the x-axis direction, y-axis direction, and z-axis direction can be interchanged with each other.
- the imaging device 10 includes an imaging optical system 11 , an imaging device 12 and a processor 13 .
- the imaging device 10 can be installed at various positions on the vehicle 100 .
- the imaging device 10 includes, but is not limited to, a front camera, a left side camera, a right side camera, a rear camera, and the like.
- the front camera, the left side camera, the right side camera, and the rear camera are installed on the vehicle 100 so as to be able to image the front, left side, right side, and rear peripheral areas of the vehicle 100, respectively.
- the imaging device 10 is attached to the vehicle 100 with the optical axis directed downward from the horizontal direction so as to be able to image the rear of the vehicle 100 .
- the imaging optical system 11 may be configured including one or more lenses.
- the imaging device 12 may be configured including a CCD image sensor (charge-coupled device image sensor) or a CMOS image sensor (complementary MOS image sensor).
- the imaging device 12 converts an object image (subject image) formed on the imaging surface of the imaging device 12 by the imaging optical system 11 into an electric signal.
- the imaging device 12 can capture moving images at a predetermined frame rate.
- a frame is each still image that constitutes a moving image.
- the number of images that can be captured in one second is called a frame rate.
- the frame rate may be, for example, 60 fps (frames per second) or 30 fps.
- the processor 13 controls the imaging device 10 as a whole and performs various image processing on the moving image output from the imaging device 12 .
- Image processing performed by the processor 13 may include arbitrary processing such as distortion correction, brightness adjustment, contrast adjustment, and gamma correction.
- the processor 13 may consist of one or more processors.
- Processor 13 includes one or more circuits or units configured to perform one or more data computing procedures or processes, eg, by executing instructions stored in associated memory.
- Processor 13 includes one or more processors, microprocessors, microcontrollers, application specific integrated circuits (ASICs), digital signal processors (DSPs), programmable logic devices (PLDs). ), field-programmable gate arrays (FPGAs), any combination of these devices or configurations, or other known device or configuration combinations.
- ASICs application specific integrated circuits
- DSPs digital signal processors
- PLDs programmable logic devices
- FPGAs field-programmable gate arrays
- the object tracking device 20 includes an input interface 21, a storage unit 22, a processor 23 and an output interface 24.
- the input interface 21 is configured to be able to communicate with the imaging device 10 by wired or wireless communication means.
- the input interface 21 acquires moving images from the imaging device 10 as sensor data.
- the input interface 21 may correspond to the transmission method of the image signal transmitted by the imaging device 10 .
- the input interface 21 can be rephrased as an input section or an acquisition section.
- the imaging device 10 and the input interface 21 may be connected by an in-vehicle communication network such as a CAN (control area network).
- the storage unit 22 is a storage device that stores data and programs necessary for processing performed by the processor 23 .
- the storage unit 22 temporarily stores moving images acquired from the imaging device 10 .
- the storage unit 22 stores data generated by processing performed by the processor 23 .
- the storage unit 22 may be configured using, for example, one or more of a semiconductor memory, a magnetic memory, an optical memory, and the like.
- Semiconductor memory may include volatile memory and non-volatile memory.
- Magnetic memory may include, for example, hard disks and magnetic tapes.
- the optical memory may include, for example, a CD (compact disc), a DVD (digital versatile disc), and a BD (blu-ray (registered trademark) disc).
- the processor 23 controls the object tracking device 20 as a whole.
- the processor 23 recognizes an object image included in the moving image acquired via the input interface 21 .
- the processor 23 maps the coordinates of the image of the recognized object to the coordinates of the object 40 in the virtual space 46 (see FIG. 6), and calculates the position and velocity of the mass point 45 (see FIG. 5) representing the object 40 in the virtual space 46.
- Mass point 45 is a point that has mass and no size.
- the virtual space 46 is a two-dimensional space in which the value in the z-axis direction is a predetermined fixed value in a coordinate system consisting of three axes of x, y, and z in the real space.
- the processor 23 may map-transform the coordinates of the tracked mass point 45 on the virtual space 46 to the coordinates on the moving image.
- the processor 23 detects a detection target from the moving image and performs tracking using a Kalman filter.
- the processor 23 can detect a plurality of detection targets from the moving image, and tracks each of the plurality of detection targets using a Kalman filter.
- the processor 23 can avoid such problems by associating one or more Kalman filters with each of a plurality of detection targets.
- the processor 23 manages observed values, Kalman filters, and unique identification information of tracked objects (hereinafter “tracked object IDs”) in each layer.
- the processor 23 determines whether or not the tracked objects are the same object (the same detection target), and executes processing for associating the observed value, the Kalman filter, and the tracked object ID. This makes it possible to further improve the accuracy of tracking a plurality of detection targets.
- the processor 23 selects or initializes, among a plurality of Kalman filters associated with the same detection target or observed value, those with low likelihood as candidates for exclusion from the association. For example, when many Kalman filters are associated with the same detection target or observation value, the processor 23 can prevent an increase in computational load by excluding some. Details of the processing performed by the processor 23 will be described later.
- Processor 23, like processor 13 of imaging device 10, may include multiple processors.
- the processor 23 may be configured by combining a plurality of types of devices, like the processor 13 .
- the output interface 24 is configured to output an output signal from the object tracking device 20 .
- the output interface 24 can be called an output unit.
- the output interface 24 may output the detection result of the detection target such as the coordinates of the mass point 45, for example.
- the output interface 24 may be configured including a physical connector and a wireless communication device.
- the output interface 24 may be connected to a network of the vehicle 100 such as CAN, for example.
- the output interface 24 can be connected to the display 30, the control device of the vehicle 100, the alarm device, and the like via a communication network such as CAN.
- the information output from the output interface 24 may be appropriately used by each of the display 30, the control device, and the alarm device.
- the display 30 can display moving images output from the object tracking device 20 .
- the display 30 receives the coordinates of the mass point 45 representing the position of the image of the object from the object tracking device 20, the display 30 generates an image element according to the coordinates (for example, a warning to be displayed together with the approaching object) and superimposes it on the moving image.
- Display 30 may employ various types of devices.
- the display 30 may be a liquid crystal display (LCD), an organic EL (electro-luminescence) display, an inorganic EL display, a plasma display (PDP), a field emission display (FED), Electrophoretic displays, twist ball displays, etc. may be employed.
- Non-transitory computer-readable media include, but are not limited to, magnetic storage media, optical storage media, magneto-optical storage media, and semiconductor storage media.
- Magnetic storage media include magnetic disks, hard disks, and magnetic tapes.
- Optical storage media include optical discs such as CDs, DVDs and BDs.
- Semiconductor storage media include ROM (read only memory), EEPROM (electrically erasable programmable read-only memory), and flash memory.
- the flowchart in FIG. 3 shows the processing performed by the processor 23 by acquiring sequential frames of a moving image.
- the processor 23 of the object tracking device 20 tracks the position of the object image 42 (see FIG. 4) each time it acquires a moving image frame according to the flowchart of FIG.
- there may be multiple objects 40 to be detected including pedestrians 40A, automobiles 40B, and bicycles 40C.
- the objects 40 are not limited to moving objects and people, but may include various objects such as road obstacles.
- one of the plurality of objects 40 specifically, the pedestrian 40A included in the moving image of the imaging device 10 installed behind the vehicle 100 will be used.
- Other objects 40 for example, automobile 40B and bicycle 40C are also tracked by similar processing.
- FIG. 4 shows an example of one frame of a moving image.
- an image of an object 40 behind the vehicle 100 (object image 42) is displayed in a two-dimensional image space 41 made up of the uv coordinate system.
- the u coordinate is the horizontal coordinate of the image.
- the v coordinate is the vertical coordinate of the image.
- the origin of the uv coordinates is the upper left point of the image space 41 .
- the positive direction is the direction from left to right.
- the positive direction of the v-coordinate is the direction from top to bottom.
- the processor 23 recognizes the object image 42 from each frame of the moving image by image recognition (step S102).
- Methods for recognizing the object image 42 include various known methods.
- the method of recognizing the object image 42 includes a method of recognizing the shape of an object such as a car or a person, a method of template matching, a method of calculating a feature amount from an image and using it for matching, and the like.
- a function approximator capable of learning the relationship between input and output can be used to calculate the feature amount.
- a neural network for example, can be used as the function approximator capable of learning the relationship between input and output.
- the processor 23 maps the coordinates (u, v) of the object image 42 in the image space 41 to the object coordinates (x', y') in the virtual space 46 (see FIG. 6) (step S103).
- the coordinates (u, v) in the image space 41 which are two-dimensional coordinates, cannot be transformed into the coordinates (x, y, z) in the real space.
- the coordinates (u, v) in the image space 41 are changed to the coordinates (x, y, z 0 ) in the real space (z 0 is (fixed value) can be mapped to the coordinates (x', y') in the two-dimensional virtual space 46 corresponding to the fixed value).
- the virtual space 46 is two-dimensional in this embodiment, it may be three-dimensional depending on the input information (type of sensor).
- a representative point 43 located in the center of the bottom of the image 42 of the object is identified.
- the representative point 43 can be the lowest v-coordinate position and the center of the u-coordinate range of the area occupied by the object image 42 in the image space 41 .
- This representative point 43 is assumed to be the position where the object 40 corresponding to the object image 42 is in contact with the road surface or the ground.
- FIG. 5 shows the relationship between an object 40 located in a three-dimensional real space and an image 42 of the object on a two-dimensional image space 41.
- the center of the imaging optical system 11 of the imaging device 10 is directed to the corresponding coordinates (x, y, z) in the real space.
- Direction can be calculated.
- the internal parameters of the imaging device 10 include information such as the focal length and distortion of the imaging optical system 11 and the pixel size of the imaging device 12 .
- the reference plane 44 corresponds to the road surface or the ground on which the vehicle 100 is located.
- a particular point is a point corresponding to mass point 45 .
- the processor 23, as shown in FIG. , v y′ ) are traced (step S104). Since mass point 45 has position (x', y') and velocity (v x ', v y' ) information, processor 23 can determine the range of mass point 45's position (x', y') in successive frames. can be predicted. The processor 23 can recognize the mass point 45 located in the predicted range in the next frame as the mass point 45 corresponding to the image 42 of the object being tracked. The processor 23 sequentially updates the position (x', y') and velocity (v x ' , v y' ) of the mass point 45 each time a new frame is input.
- estimation using a Kalman filter based on a state space model can be adopted. Prediction/estimation using the Kalman filter improves robustness against undetectability and erroneous detection of the object 40 to be detected.
- the object tracking device 20 of the present disclosure by mapping the object image 42 to the mass point 45 in the real space, it becomes possible to apply a model that describes the motion in the real space. improves. Further, by treating the object 40 as a mass point 45 having no size, simple and easy tracking becomes possible.
- processor 23 may map coordinates of mass point 45 in virtual space 46 to coordinates (u,v) in image space 41 to indicate the estimated position.
- a mass point 45 located at coordinates (x', y') in virtual space 46 can be mapped into image space 41 as a point located at coordinates (x, y, 0) in real space. Coordinates (x, y, 0) in the real space can be mapped to coordinates (u, v) in the image space 41 of the imaging device 10 by a known method.
- the processor 23 mutually transforms the coordinates (u, v) on the image space 41, the coordinates (x', y') on the virtual space 46, and the coordinates (x, y, 0) on the real space. can be done.
- FIG. 7 is a diagram for explaining data association.
- Data association is a process of associating a Kalman filter with an observed value. Multiple Kalman filters may be associated with multiple observations in a data association.
- the observed value is the position of the detection target.
- the processor 23 assigns identifiers to the plurality of observed values and the plurality of Kalman filters to distinguish them.
- the processor 23 designates each of the plurality of observation values as observation value (1), observation value (2), observation value (3), . . . using serial numbers, for example.
- the processor 23 uses symbols and serial numbers, for example, to designate the plurality of Kalman filters as KF(1), KF(2), KF(3), and so on.
- the processor 23 performs data association between M observed values and N Kalman filters.
- M is an integer of 2 or more.
- N is an integer greater than or equal to M.
- processor 23 has made data associations between 3 observations and 5 Kalman filters.
- Observed value (1) is the position of pedestrian 40A detected in frame (k) of the moving image.
- Observation (2) is the position of vehicle 40B detected in frame (k) of the moving image.
- Observed value (3) is the position of bicycle 40C detected in frame (k) of the moving image.
- Frame (k-1) is the frame immediately before frame (k) in the moving image.
- Frame (k-2) is the frame two frames before frame (k) in the moving image. Assume that the current frame is frame (k).
- KF(2) has been used to track the pedestrian 40A until frame (k-1), but is initialized because it satisfies the initialization conditions described later and is used to track the position of the detection target.
- KF(5) is a Kalman filter newly prepared by recognizing a new bicycle 40C in frame (k-2). KF(5) has started tracking the detected object because the newly recognized bicycle 40C was also recognized in the current frame (k). The other Kalman filters continue to track their detection targets from frame (k-2).
- the processor 23 associates observed value (1) with KF (1).
- Processor 23 associates observation (2) with KF(3) and KF(4).
- the processor 23 associates KF(5) with observed value (3).
- the processor 23 allows overlap of detection results in the tracking process of multiple detection targets. That is, processor 23 uses KF(3) and KF(4) to predict the range of observation (2), ie, the position of vehicle 40B.
- local optimization can be achieved by allowing overlap in data associations.
- a method that associates multiple observed values with multiple Kalman filters on a one-to-one basis without allowing duplication may cause a chain of misassociations due to global optimization. In this embodiment, since duplication is allowed, problems such as misassociation chains do not occur.
- one or more Kalman filters are associated with one observed value, and tracking failure is less likely to occur for any observed value, so robustness can be improved.
- Tracking object ID management a plurality of Kalman filters can be associated with one observed value as described above, it is also possible that a plurality of observed values are associated with one object to be detected. For example, when the detection target is the automobile 40B, and it disappears from the moving image due to a lane change, etc., and then reappears in the moving image, a new observation value may be associated with another object.
- object tracking device 20 preferably identifies each tracked object and keeps track of its correspondence with observed values.
- the processor 23 performs tracking object ID management using a hierarchical structure as described below, groups a plurality of Kalman filters, and determines whether or not they correspond to the same object.
- FIG. 8 is a diagram showing the hierarchical structure of tracking object ID management (ID management) in this embodiment.
- Tracking object ID management is a process of associating a Kalman filter with a detection target.
- the processor 23 manages observed values, Kalman filters, and tracked object IDs in each layer.
- the processor 23 also enables accurate object tracking by associating observed values, Kalman filters, and tracked object IDs.
- the tracked object ID is the unique identification information of the tracked object as described above. If the tracked object ID associated with multiple observations or multiple Kalman filters is the same, then these observations or Kalman filters are associated with tracking the same object.
- the processor 23 performs grouping of a plurality of Kalman filters when frames of the moving image are acquired. The processor 23 then updates the association of observations, Kalman filters and tracked object IDs.
- processor 23 groups KF(1), KF(2) and KF(3) and assigns an identifier "tracked object ID(1)" to the object to be tracked using these Kalman filters. to control the tracking of this object.
- the processor 23 also groups KF(4) and KF(5), assigns an identifier “tracked object ID(2)” to the object to be tracked using these Kalman filters, and performs tracking control of this object. I do.
- the processor 23 associates the Kalman filters corresponding to the objects determined to be the same, and controls the tracking in a hierarchical structure that also associates the detection results of the detection targets corresponding to these Kalman filters. tracking becomes possible.
- the processor 23 can, for example, compare or select detection results using a plurality of linked Kalman filters to obtain a detection result with a high degree of certainty.
- multiple Kalman filters can be associated with one observation value, and multiple Kalman filters can be associated with one detection target (detection target having one tracked object ID).
- detection target detection target having one tracked object ID
- By associating a plurality of Kalman filters tracking failures are less likely to occur, and robustness can be improved.
- many Kalman filter correspondences increase the computational load and may cause delays in the control of the object tracker 20 by the processor 23 .
- the processor 23 performs overlapping Kalman filter management to exclude some of the overlapping Kalman filters (multiple Kalman filters associated with the same detected object or observed value) from the association, as described below.
- the processor 23 executes a first process and a second process as redundant Kalman filter management.
- the first process is a process in which the processor 23 selects overlapping Kalman filters with low "probability" as "excluded candidate Kalman filters” that can be excluded from the matching.
- the first process can be likened to pruning a Kalman filter of low importance, and can be called a pruning process.
- the second process is a process in which the processor 23 initializes by excluding the “excluded candidate Kalman filter” that has satisfied the initialization condition from the association.
- the second process initializes the Kalman filter to an empty state and can be referred to as the vacant process.
- the "certainty" of the Kalman filter in the first process is the accuracy of prediction/estimation of the position of the associated detection target or observed value, in other words, the degree of certainty.
- Processor 23 can determine the likelihood of the Kalman filter by, for example, the size of the error ellipse.
- the error ellipse indicates an estimated range based on the probability density distribution of the position, and indicates that the position is located inside the ellipse with a predetermined probability (99% as an example).
- the error ellipse is calculated using the standard deviation in the x' direction and the standard deviation in the y' direction of the two-dimensional virtual space 46 (see FIG. 6).
- the Kalman filter with the largest error ellipse may be selected as the 'excluded candidate Kalman filter' as having a low probability.
- the initialization condition in the second process is that the number of times the filter is selected as the exclusion candidate Kalman filter reaches the first value.
- the first value can be arbitrarily selected, but is "5" as an example.
- the Kalman filter selected as the exclusion candidate Kalman filter for the fifth time by the second processing is detected.
- the number of times that the exclusion candidate Kalman filter has been selected may be consecutive or cumulative. For example, when the number of consecutive times is used, the number returns to "0" if a certain Kalman filter is not selected as an exclusion candidate Kalman filter in the first processing on the way.
- the number of times an exclusion candidate Kalman filter has been selected may be counted for each Kalman filter by a counter provided in the processor 23 .
- FIG. 9 is a diagram showing error ellipses of overlapping Kalman filters associated with the same observation point in the data association.
- a plurality of Kalman filters are associated with the same observation point (one observation value)
- one object is recognized as two objects due to the influence of light reflection, and a new Kalman filter is applied to one of them.
- using multiple associated Kalman filters to control detection target tracking can be performed in parallel, but the computational effort can be increased. Therefore, it is preferable to perform duplicate Kalman filter management on data associations.
- Processor 23 computes an error ellipse for each of the three Kalman filters.
- Processor 23 performs a first process to select Kalman filters with relatively large error ellipses as candidate Kalman filters to be excluded.
- Processor 23 may select a plurality of candidate exclusion Kalman filters, but selects KF(q) with the largest error ellipse in the example of FIG. 9 as the candidate exclusion Kalman filter. After that, the processor 23 executes the second process.
- FIG. 9 the processor 23 executes the second process.
- the processor 23 determines that the number of times KF(q) has been selected as an exclusion candidate Kalman filter reaches a first value (eg, 5 times), i.e., if the initialization condition is satisfied, KF Unmap (q) and initialize KF(q).
- a first value e.g, 5 times
- the processor 23 may execute the first process and the second process when the number of Kalman filters associated with the same observed value exceeds the second value.
- the second value can be arbitrarily chosen, but is "2" in the example of FIG.
- the second value is the upper limit of the number of overlapping Kalman filters associated with the same observation point.
- the second value is preferably a small numerical value from the viewpoint of reducing the load of arithmetic processing, but is set to 2 or more in order to ensure robustness in object tracking processing.
- processor 23 initializes KF(q) by executing the first process and the second process because the number of overlapping Kalman filters exceeds two.
- KF(q) which has a relatively low degree of importance
- KF(r) KF(r)
- a plurality of exclusion candidate Kalman filters may be selected in the first process.
- a plurality of exclusion candidate Kalman filters may be initialized in the second process.
- the number of exclusion candidate Kalman filters may be determined based on the number of overlapping Kalman filters and the second value (the upper limit of the number of overlapping Kalman filters associated with the same observation point). For example, if the second value is '2' and the number of overlapping Kalman filters is '5', the processor 23 may select three excluded candidate Kalman filters that are their difference by a first process. . That is, the processor 23 may select three Kalman filters with relatively large error ellipses as candidates to be excluded from association with the same observation point by the first processing.
- FIG. 10 is a diagram showing error ellipses of overlapping Kalman filters associated with the same detection target in tracked object ID management.
- control of tracking of a detection target can be executed in parallel using a plurality of Kalman filters associated with the same detection target (same object having one tracked object ID), but computation processing increases. obtain. Therefore, it is preferable to perform duplicate Kalman filter management in tracking object ID management.
- the correspondence between the same detection target and the Kalman filter is performed by clustering such as DBSCAN (density-based spatial clustering of applications with noise).
- the processor 23 determines that the Kalman filters belong to one group when the centers of the error ellipses of the Kalman filters are included in a predetermined range.
- the predetermined range is indicated by a circle.
- the predetermined range may change according to the size of the tracked object. For example, if the tracked object is the automobile 40B, the predetermined range may be set larger than when the tracked object is the pedestrian 40A. Also, the predetermined range may be constant regardless of the type of tracked object.
- the clustering technique is not limited to DBSCAN. Clustering may be performed by other techniques, for example the k-means method.
- processor 23 performs a first process to select Kalman filters with relatively large error ellipses as candidate Kalman filters to be excluded. After that, the processor 23 executes the second process. In the example of FIG. 10, the processor 23 determines that the number of times KF(q) has been selected as an exclusion candidate Kalman filter reaches a first value (eg, 5 times), i.e., if the initialization condition is satisfied, KF Unmap (q) and initialize KF(q).
- a first value eg, 5 times
- the processor 23 performs the first processing and the second processing when the number of Kalman filters associated with the same detection target exceeds the second value. processing may be performed.
- the second value is "the upper limit of the number of overlapping Kalman filters associated with the same observation point" as described above, and is "2" as an example.
- processor 23 initializes KF(q) by executing the first process and the second process because the number of overlapping Kalman filters exceeds two. By excluding KF(q), which has a relatively low degree of importance, from the matching, an increase in the arithmetic processing of the processor 23 is prevented.
- KF(p) and KF(r) are still associated with the same detection target, ensuring robustness in tracking the position of this detection target.
- the processor 23 executes the first process and the second process when the number of Kalman filters associated with the same detection target exceeds a third value.
- the third value is "the upper limit of the number of overlapping Kalman filters associated with the same detection target" and is set regardless of the second value.
- the processor 23 executes the first process and the second process when the number of overlapping Kalman filters exceeds a second value ("2" as an example) in the data association, and in tracking object ID management, The first process and the second process may be performed when the number of overlapping Kalman filters exceeds a third value ("4" as an example).
- the processor 23 may select or combine the above execution timings and execution conditions to perform overlapping Kalman filter management.
- Processor 23 may perform a first process and a second process, eg, in data association and tracked object ID management.
- Processor 23 may perform the first process and the second process if the number of duplicated Kalman filters exceeds a second value, for example on data associations only.
- the processor 23 may perform the first process and the second process when the number of overlapping Kalman filters exceeds a third value, for example, only in tracked object ID management.
- the object tracking device 20 allows overlap of detection results in the process of tracking a plurality of detection targets with the above configuration. Therefore, the object tracking device 20 can track a plurality of objects with high accuracy without causing a chain of misassociations.
- the object tracking device 20 according to the present embodiment also executes redundant Kalman filter management in the process of tracking an object. Therefore, the object tracking device 20 can track the object with high accuracy without increasing the computational load.
- the object tracking system 1 includes the imaging device 10, the object tracking device 20, and the display 30, but at least two of them may be integrated.
- the functions of the object tracking device 20 can be installed in the imaging device 10 .
- the imaging apparatus 10 may include the storage unit 22 and the output interface 24 in addition to the imaging optical system 11 , the imaging element 12 and the processor 13 .
- the processor 13 may perform the processing performed by the processor 23 in the above embodiment on the moving image output by the imaging device 10 .
- Such a configuration may realize the imaging device 10 that tracks an object.
- Mobile objects in this disclosure include vehicles, ships, and aircraft.
- Vehicle in the present disclosure includes, but is not limited to, automobiles and industrial vehicles, and may include railroad and utility vehicles, and fixed-wing aircraft that travel on runways.
- Motor vehicles include, but are not limited to, cars, trucks, buses, motorcycles and trolleybuses, and may include other vehicles that travel on roads.
- Industrial vehicles include industrial vehicles for agriculture and construction.
- Industrial vehicles include, but are not limited to, forklifts and golf carts.
- Industrial vehicles for agriculture include, but are not limited to, tractors, cultivators, transplanters, binders, combines and mowers.
- Industrial vehicles for construction include, but are not limited to, bulldozers, scrapers, excavators, mobile cranes, tippers and road rollers.
- Vehicles include those driven by human power.
- classification of vehicles is not limited to the above.
- automobiles may include road-driving industrial vehicles, and the same vehicle may be included in multiple classes.
- Vessels in this disclosure include marine jets, boats, and tankers.
- Aircraft in this disclosure includes fixed-wing and rotary-wing aircraft.
- object tracking system 10 imaging device 11 imaging optical system 12 imaging device 13 processor 20 object tracking device 21 input interface 22 storage unit 23 processor 24 output interface 30 display 40 object 40A pedestrian 40B car 40C bicycle 41 image space 42 object image 43 Representative point 44 Reference plane 45 Mass point 46 Virtual space 100 Vehicle
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
Abstract
Description
センサデータを取得する入力インターフェイスと、
前記センサデータから検出対象を検出し、前記検出対象及び観測値のそれぞれに対応付けが行われたカルマンフィルタを用いて、前記検出対象の追跡を行うプロセッサと、
前記検出対象の検出結果を出力する出力インターフェイスと、を備え、
前記プロセッサは、
同一の検出対象又は観測値に対応付けが行われた複数の前記カルマンフィルタのうち確からしさが低いものを、前記対応付けから除外され得る除外候補カルマンフィルタとして選択する第1の処理と、
初期化条件を満たした前記除外候補カルマンフィルタを前記対応付けから除外して初期化する第2の処理と、を実行する。
センサデータを取得することと、
前記センサデータから検出対象を検出し、前記検出対象及び観測値のそれぞれに対応付けが行われたカルマンフィルタを用いて、前記検出対象の追跡を行うことと、
前記検出対象の検出結果を出力することと、を含み、
前記検出対象の追跡を行うことは、
同一の検出対象又は観測値に対応付けが行われた複数の前記カルマンフィルタのうち確からしさが低いものを、前記対応付けから除外され得る除外候補カルマンフィルタとして選択する第1の処理と、
初期化条件を満たした前記除外候補カルマンフィルタを前記対応付けから除外して初期化する第2の処理と、を含む。
図7は、データアソシエーションを説明するための図である。データアソシエーションは、カルマンフィルタを観測値に対応付ける処理である。データアソシエーションにおいて、複数のカルマンフィルタが、複数の観測値と対応付けられ得る。ここで、観測値は、検出対象の位置である。プロセッサ23は、複数の観測値及び複数のカルマンフィルタに識別子を付して区別する。本実施形態において、プロセッサ23は、例えば通し番号を用いて、複数の観測値のそれぞれを観測値(1)、観測値(2)、観測値(3)…とする。また、プロセッサ23は、例えば記号及び通し番号を用いて、複数のカルマンフィルタのそれぞれをKF(1)、KF(2)、KF(3)…とする。
ここで、上記のように1つの観測値に複数のカルマンフィルタが対応付けられ得るが、検出対象である1つの物体に複数の観測値が対応付けられることもあり得る。例えば、検出対象が自動車40Bであって、車線変更などによって動画像から一度消失した後に再び動画像に出現した場合などに、別物体として新たな観測値が対応付けられることがあり得る。正確な物体の追跡を行うために、物体追跡装置20は、それぞれの追跡物体を識別して、観測値との対応付けを把握することが好ましい。本実施形態において、プロセッサ23は、以下に説明するように階層構造を用いた追跡物体ID管理を実行し、複数のカルマンフィルタのグループ化を行って同一物体に対応するものか否かを判定する。
上記のように、1つの観測値に複数のカルマンフィルタが対応付けられ、1つの検出対象(1つの追跡物体IDを有する検出対象)に複数のカルマンフィルタが対応付けられ得る。複数のカルマンフィルタを対応付けることによって追跡の失敗が生じにくくなり、ロバスト性を向上させることができる。しかし、多くのカルマンフィルタの対応付けは演算の負荷を増大させて、プロセッサ23の物体追跡装置20の制御に遅延を生じさせ得る。プロセッサ23は、以下に説明するように、重複カルマンフィルタ(同一の検出対象又は観測値に対応付けが行われた複数のカルマンフィルタ)の一部を対応付けから除外する重複カルマンフィルタ管理を実行する。
10 撮像装置
11 撮像光学系
12 撮像素子
13 プロセッサ
20 物体追跡装置
21 入力インターフェイス
22 記憶部
23 プロセッサ
24 出力インターフェイス
30 ディスプレイ
40 物体
40A 歩行者
40B 自動車
40C 自転車
41 画像空間
42 物体の像
43 代表点
44 基準面
45 質点
46 仮想空間
100 車両
Claims (7)
- センサデータを取得する入力インターフェイスと、
前記センサデータから検出対象を検出し、前記検出対象及び観測値のそれぞれに対応付けが行われたカルマンフィルタを用いて、前記検出対象の追跡を行うプロセッサと、
前記検出対象の検出結果を出力する出力インターフェイスと、を備え、
前記プロセッサは、
同一の検出対象又は観測値に対応付けが行われた複数の前記カルマンフィルタのうち確からしさが低いものを、前記対応付けから除外され得る除外候補カルマンフィルタとして選択する第1の処理と、
初期化条件を満たした前記除外候補カルマンフィルタを前記対応付けから除外して初期化する第2の処理と、を実行する、物体追跡装置。 - 前記プロセッサは、前記カルマンフィルタの確からしさを誤差楕円の大きさにより判定する、請求項1に記載の物体追跡装置。
- 前記初期化条件は、前記除外候補カルマンフィルタとして選択された回数が第1の値に達することである、請求項1又は2に記載の物体追跡装置。
- 前記プロセッサは、同一の検出対象又は観測値に対応付けが行われた複数の前記カルマンフィルタの数が第2の値を超えた場合に、前記第1の処理及び前記第2の処理を実行する、請求項1から3のいずれか一項に記載の物体追跡装置。
- 前記プロセッサは、前記カルマンフィルタを前記観測値に対応付ける処理において、前記第1の処理及び前記第2の処理を実行する、請求項1から4のいずれか一項に記載の物体追跡装置。
- 前記プロセッサは、前記カルマンフィルタを前記検出対象に対応付ける処理において、前記第1の処理及び前記第2の処理を実行する、請求項1から5のいずれか一項に記載の物体追跡装置。
- センサデータを取得することと、
前記センサデータから検出対象を検出し、前記検出対象及び観測値のそれぞれに対応付けが行われたカルマンフィルタを用いて、前記検出対象の追跡を行うことと、
前記検出対象の検出結果を出力することと、を含み、
前記検出対象の追跡を行うことは、
同一の検出対象又は観測値に対応付けが行われた複数の前記カルマンフィルタのうち確からしさが低いものを、前記対応付けから除外され得る除外候補カルマンフィルタとして選択する第1の処理と、
初期化条件を満たした前記除外候補カルマンフィルタを前記対応付けから除外して初期化する第2の処理と、を含む、物体追跡方法。
Priority Applications (3)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
EP22795651.3A EP4332895A1 (en) | 2021-04-27 | 2022-04-20 | Object tracking device and object tracking method |
CN202280031202.5A CN117355863A (zh) | 2021-04-27 | 2022-04-20 | 物体跟踪装置及物体跟踪方法 |
US18/557,465 US20240212172A1 (en) | 2021-04-27 | 2022-04-20 | Object tracking device and object tracking method |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
JP2021-075289 | 2021-04-27 | ||
JP2021075289A JP2022169320A (ja) | 2021-04-27 | 2021-04-27 | 物体追跡装置及び物体追跡方法 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2022230738A1 true WO2022230738A1 (ja) | 2022-11-03 |
Family
ID=83847039
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/JP2022/018330 WO2022230738A1 (ja) | 2021-04-27 | 2022-04-20 | 物体追跡装置及び物体追跡方法 |
Country Status (5)
Country | Link |
---|---|
US (1) | US20240212172A1 (ja) |
EP (1) | EP4332895A1 (ja) |
JP (1) | JP2022169320A (ja) |
CN (1) | CN117355863A (ja) |
WO (1) | WO2022230738A1 (ja) |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPH11321494A (ja) | 1998-05-08 | 1999-11-24 | Yazaki Corp | 後側方監視装置 |
JP2019168953A (ja) * | 2018-03-23 | 2019-10-03 | 株式会社豊田中央研究所 | 状態推定装置及びプログラム |
-
2021
- 2021-04-27 JP JP2021075289A patent/JP2022169320A/ja active Pending
-
2022
- 2022-04-20 EP EP22795651.3A patent/EP4332895A1/en active Pending
- 2022-04-20 WO PCT/JP2022/018330 patent/WO2022230738A1/ja active Application Filing
- 2022-04-20 US US18/557,465 patent/US20240212172A1/en active Pending
- 2022-04-20 CN CN202280031202.5A patent/CN117355863A/zh active Pending
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPH11321494A (ja) | 1998-05-08 | 1999-11-24 | Yazaki Corp | 後側方監視装置 |
JP2019168953A (ja) * | 2018-03-23 | 2019-10-03 | 株式会社豊田中央研究所 | 状態推定装置及びプログラム |
Also Published As
Publication number | Publication date |
---|---|
JP2022169320A (ja) | 2022-11-09 |
EP4332895A1 (en) | 2024-03-06 |
US20240212172A1 (en) | 2024-06-27 |
CN117355863A (zh) | 2024-01-05 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
JP7140922B2 (ja) | マルチセンサデータ融合方法、及び装置 | |
US10937178B1 (en) | Image-based depth data and bounding boxes | |
US11748909B2 (en) | Image-based depth data and localization | |
US10984543B1 (en) | Image-based depth data and relative depth data | |
JP2022553491A (ja) | 車両ナビゲーション用のシステムおよび方法 | |
US11782158B2 (en) | Multi-stage object heading estimation | |
US11697412B2 (en) | Collision monitoring using statistic models | |
JP4956453B2 (ja) | 物体検出装置 | |
CN114072841A (zh) | 根据图像使深度精准化 | |
CN1963867A (zh) | 监视装置 | |
CN114902070A (zh) | 传感器退化检测和修复 | |
JP2020052695A (ja) | 物体検出装置、物体検出方法及び物体検出用コンピュータプログラム | |
GB2608201A (en) | Identifying objects using LiDAR | |
Verma et al. | Vehicle detection, tracking and behavior analysis in urban driving environments using road context | |
JP2020067896A (ja) | 進行可能方向検出装置及び進行可能方向検出方法 | |
JP7226368B2 (ja) | 物体状態識別装置 | |
WO2022091595A1 (ja) | 物体追跡装置及び物体追跡方法 | |
WO2022230738A1 (ja) | 物体追跡装置及び物体追跡方法 | |
Siddiqui et al. | Object/Obstacles detection system for self-driving cars | |
JP7425223B2 (ja) | 物体追跡装置及び物体追跡方法 | |
WO2022230739A1 (ja) | 物体追跡装置 | |
Gruyer et al. | Target-to-track collaborative association combining a laser scanner and a camera | |
US20230252638A1 (en) | Systems and methods for panoptic segmentation of images for autonomous driving | |
JP2022146384A (ja) | 物体検出装置 | |
JP2024037165A (ja) | オブジェクトをモデリングするための方法、機器、記憶媒体及び車両制御方法 |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 22795651 Country of ref document: EP Kind code of ref document: A1 |
|
WWE | Wipo information: entry into national phase |
Ref document number: 202280031202.5 Country of ref document: CN Ref document number: 18557465 Country of ref document: US |
|
WWE | Wipo information: entry into national phase |
Ref document number: 2022795651 Country of ref document: EP |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
ENP | Entry into the national phase |
Ref document number: 2022795651 Country of ref document: EP Effective date: 20231127 |