Object Tracking Device

Abstract

An object tracking device (20) according to one aspect of the present disclosure includes an acquisition unit (21), an object detection unit (22), a distance calculation unit (23), a speed calculation unit (24), a direction calculation unit (25), a candidate generation unit (26), a state estimation unit (29), and a candidate selection unit (30). The candidate generation unit (26) includes a reference speed calculation unit (27) and a turning speed calculation unit (28). The reference speed calculation unit (27) calculates the minimum turning speed that is equal to or greater than the speed lower limit value set according to the moving speed of the moving body as the reference speed. The turning speed calculation unit (28) calculates a plurality of turning speeds obtained by turning the calculated reference speed in the positive direction 0 to n (n is a natural number) times.

Description

Technical Field

[0001] The present disclosure relates to an object tracking device for tracking an object. Background Art

[0002] In the case of tracking an object based on observations by a radar device, the relative velocity of the observed object sometimes has ambiguity. For example, in the case of using a method of obtaining the relative velocity based on the phase rotation of the frequency components continuously detected for the same object, for the detected phase φ, the actual phase may be φ + 2π×m (m is an integer), and the relative velocity cannot be determined.

[0003] The following Non-Patent Document 1 proposes the following technique: By assuming a plurality of relative velocities and generating a plurality of targets, and tracking the generated plurality of targets, the true relative velocity is determined. Specifically, the likelihood is calculated for each of the plurality of assumptions, and the assumption with a higher likelihood than the others through tracking is determined as the true relative velocity.

[0004] Non-Patent Document 1: K. LI et al, ‘Multitarget Tracking with Doppler Ambiguity’, IEEE TRANSACTIONS ON AEROSPACE AND ELECTRONIC SYSTEMS VOL.49, NO.4 OCTOBER 2013

[0005] In the above technique for determining the true relative velocity, as the number of assumptions of the relative velocity increases, the estimation accuracy of the true relative velocity becomes higher, but as the number of assumptions increases, the processing load increases. However, the detailed research results of the inventors have found the following problems: If the number of assumptions is reduced, there may be a case where the relative velocities of the number of assumptions do not include the true relative velocity. Furthermore, the following problems have been found: If the number of assumptions is reduced, there may be a case where the tracking of the object cannot be performed, and the object may not be recognized. Summary of the Invention

[0006] One aspect of the present disclosure preferably can suppress the processing load and can suppress the non-recognition of an object.

[0007] One aspect of the present disclosure is an object tracking device, including an acquisition unit, an object detection unit, a distance calculation unit, a speed calculation unit, a direction calculation unit, a candidate generation unit, a state estimation unit, and a candidate selection unit. The acquisition unit is configured to acquire detection information from a sensor mounted on a moving body. The object detection unit is configured to detect an object existing around the moving body based on the detection information acquired by the acquisition unit. The distance calculation unit is configured to calculate the distance of the object detected by the object detection unit. The speed calculation unit is configured to calculate the relative speed of the object detected by the object detection unit. The direction calculation unit is configured to calculate the direction of the object detected by the object detection unit. The candidate generation unit is configured to, for an object first detected by the object detection unit, calculate a plurality of folded-back speeds assuming a U-turn based on the relative speed calculated by the speed calculation unit, and is configured to generate a plurality of target candidates corresponding to the plurality of folded-back speeds. The state estimation unit is configured to, for each target candidate included in the plurality of target candidates generated by the candidate generation unit, estimate the current state of the target candidate based on the past state of the target candidate and the observation information. The observation information includes the distance calculated by the distance calculation unit, the relative speed calculated by the speed calculation unit, and the direction calculated by the direction calculation unit. The candidate selection unit is configured to select, from the plurality of target candidates generated by the candidate generation unit for the same object, the target candidate that is presumed to be the actual target. The candidate generation unit includes a reference speed calculation unit and a folded-back speed calculation unit. The reference speed calculation unit is configured to calculate the smallest folded-back speed equal to or higher than a speed lower limit value as the reference speed, and the smallest folded-back speed equal to or higher than the speed lower limit value is the folded-back speed of the relative speed calculated by the speed calculation unit. The speed lower limit value is set according to the moving speed of the moving body. The folded-back speed calculation unit is configured to calculate a plurality of folded-back speeds obtained by folding back the reference speed calculated by the reference speed calculation unit in the positive direction 0 to n (n is a natural number) times.

[0008] The relative speed of the object observed by the sensor changes according to the speed of the object and the speed of the moving body. The objects that need to be noted for the moving body are the objects existing in the traveling direction of the moving body. The higher the speed of the moving body, the higher the possibility that the relative speed of the objects existing in the traveling direction of the moving body becomes larger in the negative side. That is, the relative speed to be assumed changes according to the speed of the moving body. In addition, it is necessary to preferentially identify the objects with a high risk of collision with the moving body. Therefore, when assuming the number of times of speed folding back, it is necessary to include the largest relative speed in the negative side among the relative speeds that may be observed.

[0009] Therefore, the object tracking device according to one aspect of the present disclosure calculates the reference speed as the minimum folding-back speed equal to or higher than the speed lower limit value set according to the speed of the moving body. Then, it calculates a plurality of folding-back speeds obtained by folding back the reference speed in the positive direction, and generates a plurality of target candidates corresponding to the plurality of folding-back speeds. As a result, it is possible to calculate a plurality of folding-back speeds within an appropriate range corresponding to the speed of the moving body. Therefore, it is possible to suppress the processing load and improve the estimation accuracy of the true relative speed. Furthermore, it is possible to suppress the processing load and suppress the non-identification of the object. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] Figure 1 is a block diagram showing the configuration of the object tracking device according to the present embodiment.

[0011] Figure 2 is a block diagram showing the functions of the object tracking device according to the present embodiment.

[0012] Figure 3 is a flowchart showing the object tracking process performed by the object tracking device according to the present embodiment.

[0013] Figure 4 is a flowchart showing the target candidate generation process performed by the object tracking device according to the present embodiment.

[0014] Figure 5 is a diagram showing the folding-back speed.

[0015] Figure 6 is a diagram showing the situation of generating a plurality of target candidates and tracking an object.

[0016] Figure 7 is a diagram showing the process of selecting a target from a plurality of target candidates.

[0017] Figure 8 is a diagram showing the speed lower limit value set for the vehicle speed and the plurality of folding-back speeds calculated within the range determined according to the speed lower limit value.

[0018] Figure 9 is a diagram showing an example of the speed lower limit value and the speed upper limit value set for the vehicle speed.

[0019] Figure 10 is a diagram showing an example of the speed lower limit value and the speed upper limit value set for the vehicle speed.

[0020] Figure 11 is a diagram showing an example of the speed lower limit value and the speed upper limit value set for the vehicle speed. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0021] Hereinafter, embodiments for implementing the present disclosure will be described with reference to the accompanying drawings.

[0022] <1. Structure>

[0023] First, with reference to Figure 1 the structure of the driving assistance system 100 according to the present embodiment will be described. The driving assistance system 100 includes a radar device 10, an object tracking device 20, and a driving assistance device 50.

[0024] The radar device 10 may be mounted at the center of the front of the vehicle 80 (for example, the center of the front bumper), and has the area at the center of the front of the vehicle 80 as the detection area. In addition, the radar device 10 may be respectively mounted on the left front side and the right front side of the vehicle 80 (for example, the left end and the right end of the front bumper), and has each of the areas of the left front and the right front of the vehicle 80 as the detection area. In addition, the radar device 10 may be respectively mounted on the left rear side and the right rear side of the vehicle 80 (for example, the left end and the right end of the rear bumper), and has each of the areas of the left rear and the right rear of the vehicle 80 as the detection area. It is not necessary to mount all five of these radar devices 10 on the vehicle 80. Only one of the five radar devices 10 may be mounted on the vehicle 80, or two or more of the five radar devices 10 may be mounted on the vehicle 80. In addition, the radar device 10 corresponds to the sensor in the present disclosure, and the vehicle 80 corresponds to the moving body in the present disclosure.

[0025] The radar device 10 is a millimeter-wave radar, and includes a transmitting array antenna including a plurality of antenna elements and a receiving array antenna including a plurality of antenna elements. The radar device 10 repeatedly transmits a transmission wave at a predetermined cycle, and receives a reflected wave generated by the reflection of the transmission wave by an object. Then, the radar device 10 mixes the transmission wave and the reflected wave to generate a beat signal, and outputs the sampled beat signal (that is, detection information) to the object tracking device 20. The radar device 10 is a radar of a modulation method that generates ambiguity in the observation speed, for example, a Fast Chirp Modulation (FCM) method radar.

[0026] The object tracking device 20 includes a microcomputer having a CPU and semiconductor memories such as a ROM and a RAM. The object tracking device 20 realizes various functions by the CPU executing various programs stored in the ROM. Specifically, as Figure 2As shown, the object tracking device 20 has functions of an acquisition unit 21, an object detection unit 22, a distance calculation unit 23, a speed calculation unit 24, a direction calculation unit 25, a candidate generation unit 26, a state estimation unit 29, and a target selection unit 30. The candidate generation unit 26 includes functions of a reference speed calculation unit 27 and a turning-back speed calculation unit 28. The object tracking device 20 performs object tracking processing and outputs target information generated by the execution of the object tracking processing to the driving assistance device 50. In addition, the detailed content of the object tracking processing will be described later.

[0027] The driving assistance device 50 uses the target information generated by the object tracking device 20, as well as the state information and operation information of the vehicle 80 obtained from various sensors mounted on the vehicle 80, to control the vehicle 80 and achieve driving assistance.

[0028] <2. Processing>

[0029] Next, with reference to Figure 3 the flowchart, the object tracking processing performed by the object tracking device 20 according to the first embodiment will be described. The object tracking device 20 repeatedly executes this processing at a predetermined cycle.

[0030] First, in S10, based on the detection information acquired by the acquisition unit 21, the object detection unit 22 detects at least one object existing around the vehicle 80. And for each detected object, the distance calculation unit 23 calculates the distance of the detected object, and the speed calculation unit 24 calculates the relative speed Vobs of the detected object. In addition, for each detected object, the direction calculation unit 25 calculates the direction of the detected object. The distance of the object is the distance from the vehicle 80 to the object, the relative speed Vobs of the object is the speed of the object relative to the vehicle 80. In addition, the direction of the object is the direction of the object relative to the vehicle 80. Furthermore, as Figure 5 shown, the relative speed Vobs calculated here has ambiguity based on the turning-back of the speed.

[0031] Next, in S20, the state estimation unit 29 determines whether there is unprocessed target information. Specifically, it determines whether there is a target candidate among the registered target candidates for which the subsequent processing of S30 to S60 has not been executed. In S20, if it is determined that there is an unprocessed target candidate, the process proceeds to the processing of S30.

[0032] In S30, the state estimation unit 29 estimates the current state of one of the unprocessed target candidates based on the past state of the target candidate and the observation values obtained in S10. Specifically, the state estimation unit 29 calculates a predicted value of the state quantity of the target candidate in the current processing cycle based on the estimated value of the state quantity of the target candidate calculated in the previous processing cycle (i.e., the past state). The state quantity of the target can have elements such as the distance, relative speed, and azimuth of the target candidate, similar to the observation values, or can have elements such as the X-axis coordinate value, Y-axis coordinate value, relative speed in the X direction, and relative speed in the Y direction. The X-axis is the axis along the width direction of the vehicle 80, and the Y-axis is orthogonal to the X-axis and along the long side direction of the vehicle 80. Additionally, the state quantity of the target can have other elements.

[0033] Furthermore, the state estimation unit 29 determines the observation value associated with the calculated predicted value from the observation values obtained in S10. Specifically, the state estimation unit 29 sets a specified range centered on the calculated predicted value and determines the observation value within the specified range and closest to the predicted value as the object associated with the predicted value. The specified range is the range of observation values presumed to be obtained from the same object as the predicted value.

[0034] Moreover, the state estimation unit 29 calculates the estimated value (i.e., the current state) in the current processing cycle using a Kalman filter or the like based on the calculated predicted value and the observation value determined to be associated. In the case where there is no observation value determined to be associated, for example, the calculated predicted value is set as the estimated value.

[0035] In addition, the state estimation unit 29 calculates the likelihood of the calculated estimated value. When calculating the estimated value based on the predicted value and the observation value determined to be associated with the predicted value, the likelihood of the estimated value is increased. At this time, it is also possible that the smaller the difference between the predicted value and the observation value, the higher the likelihood of the estimated value. Additionally, in the case where there is no observation value determined to be associated and the estimated value is calculated only based on the predicted value, the likelihood of the estimated value is decreased.

[0036] Next, in S40, the target selection unit 30 determines whether the processing of all target candidates generated from the same object has ended. That is, it determines whether the estimated values of the state quantities have been calculated for all target candidates generated from the same object. As Figure 5 shown, the observed value of the relative speed Vobs obtained in S10 contains ambiguity. Therefore, as Figure 6 shown, multiple target candidates assuming multiple folding-back speeds are generated from the same object. In this embodiment, three target candidates assuming three folding-back speeds are generated from one object P0.

[0037] As Figure 6 and 7 shown, when the observation value of the object P0 is first detected in the processing cycle 1, in the processing cycle 2 and later, three target candidates are generated for the object P0, and the three target candidates are respectively tracked. During the tracking, the likelihood of the estimated value of the target candidate corresponding to the predicted value associated with the observation value increases. In Figure 6 the example shown, the likelihood of the estimated value of the target candidate corresponding to the folding count of n = 0 increases. In addition, during the tracking, when the likelihood of the estimated value is below the elimination threshold, the target candidate is eliminated and the number of target candidates decreases.

[0038] Next, in S40, when it is determined that the processing of all target candidates generated from the same object has not ended, the process returns to the process of S20, and when it is determined that the processing of all target candidates has ended, the process proceeds to the process of S50.

[0039] In S50, the target selection unit 30 determines whether any of the target candidates generated from the same object satisfies the selection condition. The selection condition is a condition for determining that the target candidate is a real target, for example, a condition that the likelihood of the estimated value is above the selection threshold. When none of the target candidates satisfies the selection condition, the process returns to the process of S20, and when one of the target candidates satisfies the selection condition, the process proceeds to the process of S60.

[0040] In S60, as Figure 7 shown, the target selection unit 30 selects the target candidate that satisfies the selection condition as the real target and determines the target. When other target candidates generated from the same object remain, the other target candidates are eliminated, and in subsequent processing cycles, only the determined target is tracked. In addition, the target selection unit 30 sends the information of the determined target to the driving assistance device 50.

[0041] In addition, in S20, when it is determined that there are no unprocessed target candidates, the process proceeds to the process of S70.

[0042] In S70, the target candidate generation unit 26 determines whether there are any unused observation values among the observation values obtained in S10. That is, it is determined whether there are any observation values among the observation values detected in S10 that are not associated with any predicted value. The unused observation value corresponds to the observation value of the object first detected in the current processing cycle. In S70, when it is determined that there are no unused observation values, this process ends. On the other hand, in S70, when it is determined that there are unused observation values, the process proceeds to the process of S80.

[0043] Next, in S80, a target candidate is generated based on one of the unused observations. That is, for an object first detected in the current processing cycle, multiple target candidates assuming multiple turning times are generated. Specifically, the flowchart shown in Figure 4 is executed.

[0044] First, in S100, the reference speed calculation unit 27 calculates a reference speed for generating multiple target candidates. Specifically, as shown in Figure 8 , the reference speed calculation unit 27 uses the observed value of the relative speed Vobs to calculate the minimum turning speed that is equal to or greater than the speed lower limit value set according to the vehicle speed of the vehicle 80 as the reference speed.

[0045] Here, it is at least preferable to be able to identify an object approaching the vehicle 80. Therefore, as shown in Figures 9 - 11 , the speed lower limit value is set to a negative value.

[0046] Moreover, it is preferable to be able to identify a stationary object. The relative speed of a stationary object is a value obtained by making the vehicle speed of the vehicle 80 negative. Therefore, as shown in Figures 9 - 11 , the speed lower limit value is set to be less than or equal to the value obtained by making the vehicle speed of the vehicle 80 negative.

[0047] In addition, it is preferable to be able to identify an object existing in the vicinity of the vehicle 80, particularly an object that is likely to collide with the vehicle 80. Therefore, as shown in Figures 9 - 11 , the speed lower limit value is set to decrease as the vehicle speed of the vehicle 80 increases. Moreover, in the correspondence relationship between the vehicle speed of the vehicle 80 and the speed lower limit value, the speed lower limit value has a part that is inversely proportional to the vehicle speed of the vehicle 80.

[0048] For example, as shown in Figure 9 , the correspondence relationship may also have: a constant part where the speed lower limit value is set to be constant at vehicle speeds less than a first speed (for example, 80 km / h); an inverse part where the speed lower limit value is set to decrease as the vehicle speed increases at vehicle speeds greater than the first speed. In addition, as shown in Figure 10 , the correspondence relationship may also have an inverse part where the speed lower limit value is set to decrease as the vehicle speed increases over the entire range of vehicle speeds. In addition, as shown in Figure 11 , the correspondence relationship may also have: an inverse part where the speed lower limit value is set to decrease as the vehicle speed increases at vehicle speeds less than a second speed (for example, 100 km / h); a constant part where the speed lower limit value is set to be constant at vehicle speeds greater than the second speed.

[0049] In addition, multiple corresponding relationships between the vehicle speed of the vehicle 80 and the speed lower limit value may be prepared. In this case, the reference speed calculation unit 27 may also calculate the reference speed using an appropriate corresponding relationship according to the driving condition of the vehicle 80. The driving condition is, for example, a condition of driving in an urban area, a condition of driving on an expressway, or the like.

[0050] Next, in S110, the folding speed calculation unit 28 calculates (n + 1) folding speeds obtained by folding the reference speed calculated by the reference speed calculation unit 27 in the positive direction 0 to n (n is a natural number) times. In the present embodiment, as Figure 8 shown, the folding speed calculation unit 28 calculates the folding speed obtained by folding the reference speed 0 times in the positive direction (i.e., the reference speed), the folding speed obtained by folding the reference speed 1 time in the positive direction, and the folding speed obtained by folding the reference speed 2 times in the positive direction. Moreover, the folding speed calculation unit 28 generates and registers three object candidates corresponding to the three calculated folding speeds, respectively.

[0051] At this time, the folding speed calculation unit 28 excludes, from the three generated object candidates, the object candidates corresponding to folding speeds equal to or higher than a preset speed upper limit value. The speed upper limit value is set as a positive value and is set to exclude objects with impossible relative speeds.

[0052] For example, as Figure 9 shown, the corresponding relationship between the vehicle speed of the vehicle 80 and the speed upper limit value may also have a constant part where the speed upper limit value is set to be constant throughout the entire range of the vehicle speed. In addition, as Figure 10 shown, the corresponding relationship may also have: an inverse proportion part where, at vehicle speeds less than the first speed, the speed upper limit value is set to decrease as the vehicle speed increases; a constant part where, at vehicle speeds equal to or higher than the first speed, the speed upper limit value is set to be constant. In addition, as Figure 11 shown, the corresponding relationship may also have an inverse proportion part where, throughout the entire range of the vehicle speed, the speed upper limit value is set to decrease as the vehicle speed increases. After the process of S110, the process returns to the process of S70.

[0053] <3. Effects>

[0054] According to the present embodiment described above, the following effects can be obtained.

[0055] (1) Calculate the reference speed as the minimum turning-back speed that is equal to or higher than the speed lower limit value set according to the vehicle speed of the vehicle 80. Then, calculate (n + 1) turning-back speeds obtained by turning back the reference speed in the positive direction 0 to n times, and generate (n + 1) target candidates corresponding to the (n + 1) turning-back speeds. Thereby, it is possible to calculate (n + 1) turning-back speeds within an appropriate range corresponding to the vehicle speed of the vehicle 80. Therefore, it is possible to suppress the processing load and improve the estimation accuracy of the true relative speed. Furthermore, it is possible to suppress the processing load and suppress the non-recognition of objects.

[0056] (2) By making the speed lower limit value negative, the speed lower limit value becomes the relative speed of an object approaching the vehicle 80, so it is possible to recognize an object approaching the vehicle 80.

[0057] (3) By setting the speed lower limit value to be less than or equal to a value obtained by making the vehicle speed of the vehicle 80 negative, it is possible to recognize a stationary object.

[0058] (4) By setting the speed lower limit value to decrease as the vehicle speed of the vehicle 80 increases, it is possible to recognize an object that may collide with the vehicle 80.

[0059] (5) In the correspondence relationship between the vehicle speed of the vehicle 80 and the speed lower limit value, there is a part where the speed lower limit value is inversely proportional to the vehicle speed, so it is possible to accurately recognize an object that may collide with the vehicle 80.

[0060] (6) Exclude the target candidates corresponding to the turning-back speeds that are equal to or higher than the set speed upper limit value from the generated multiple target candidates. By excluding the target candidates with impossible relative speeds, it is possible to reduce the processing load.

[0061] (7) By setting the speed upper limit value to a positive value, the relative speed Vobs = 0 is included between the speed lower limit value and the speed upper limit value, so it is possible to recognize an object that moves in the same direction as the vehicle 80 at the same speed (specifically, a preceding vehicle).

[0062] (Other Embodiments)

[0063] The above describes the embodiments for implementing the present disclosure, but the present disclosure is not limited to the above embodiments and can be implemented with various modifications.

[0064] (a) In the above embodiments, the vehicle 80 is described as an example of a moving body, but the moving body is not limited to the vehicle 80 and can also be a motorcycle, a ship, an airplane, etc. In addition, the radar device 10 is not limited to a millimeter-wave radar and can also be a lidar, etc.

[0065] (b) The object tracking device 20 and its method described in the present disclosure can also be implemented by a dedicated computer provided by a processor and a memory configured to execute one or more functions embodied by a computer program. Alternatively, the object tracking device 20 and its method described in the present disclosure can also be implemented by a dedicated computer provided by a processor constituted by one or more dedicated hardware logic circuits. Alternatively, the object tracking device 20 and its method described in the present disclosure can also be implemented by one or more dedicated computers constituted by a combination of a processor and a memory programmed to execute one or more functions and a processor constituted by one or more hardware logic circuits. In addition, the computer program can also be stored as instructions executed by a computer in a computer-readable non-transitory tangible recording medium. In the method of implementing the functions of each part included in the object tracking device 20, software is not necessarily included, and all of its functions can also be implemented using one or more hardware.

[0066] (c) Multiple functions of one component in the above-described embodiments can be implemented by multiple components, or one function of one component can be implemented by multiple components. In addition, multiple functions of multiple components can be implemented by one component, or one function implemented by multiple components can be implemented by one component. In addition, a part of the structure of the above-described embodiments can be omitted. In addition, at least a part of the structure of the above-described embodiments can be added to or replaced with the structure of other above-described embodiments.

[0067] (d) In addition to the above-described object tracking device, the present disclosure can also be implemented in various forms such as a system including the object tracking device as a component, a program for causing a computer to function as the object tracking device, a non-transitory physical recording medium such as a semiconductor memory recording the program, and an object tracking method.

Claims

1. An object tracking device, wherein, it comprises: an acquisition unit configured to acquire detection information from a sensor mounted on a moving body; an object detection unit configured to detect an object existing around the moving body based on the detection information acquired by the acquisition unit; a distance calculation unit configured to calculate the distance of the object detected by the object detection unit; a first speed calculation unit configured to calculate the relative speed of the object detected by the object detection unit; a direction calculation unit configured to calculate the direction of the object detected by the object detection unit; a second speed calculation unit configured to calculate, for the object first detected by the object detection unit, a folded-back speed that is equal to or greater than a speed lower limit value and is the minimum value as a reference speed, the folded-back speed that is equal to or greater than the speed lower limit value and is the minimum value being the folded-back speed of the relative speed calculated by the first speed calculation unit, and the speed lower limit value being set according to the moving speed of the moving body; a third speed calculation unit configured to calculate a plurality of folded-back speeds obtained by folding back the reference speed calculated by the second speed calculation unit in the positive direction 0 to n times, where n is a natural number, and generate a plurality of target candidates corresponding to the plurality of folded-back speeds; a state estimation unit configured to, for each of the plurality of target candidates generated by the third speed calculation unit, calculate an estimated value of the current state of the target candidate based on the correlation between a predicted value and an observed value of the current state of the target candidate predicted based on the past state of the target candidate, and calculate the likelihood of the calculated estimated value, the observed value including the distance calculated by the distance calculation unit, the relative speed calculated by the first speed calculation unit, and the direction calculated by the direction calculation unit; and a candidate selection unit configured to select, based on the likelihood of the estimated value calculated by the state estimation unit for each of the plurality of target candidates, a target candidate that is presumed to be the real target from the plurality of target candidates.

2. The object tracking device according to claim 1, wherein, the speed lower limit value is set to a negative value.

3. The object tracking device according to claim 1 or 2, wherein, the speed lower limit value is set to be less than or equal to a value that makes the moving speed negative.

4. The object tracking device according to claim 1 or 2, wherein, the speed lower limit value is set to decrease as the moving speed increases.

5. The object tracking device according to claim 1 or 2, wherein, in the correspondence relationship between the moving speed and the speed lower limit value, the speed lower limit value has a part that is inversely proportional to the moving speed.

6. The object tracking device according to claim 1 or 2, wherein, the third speed calculation unit is configured to exclude, from the plurality of target candidates generated, a target candidate corresponding to a folded-back speed that is equal to or greater than a set speed upper limit value.

7. The object tracking device according to claim 6, wherein, the speed upper limit value is set to a positive value.

Images