Object recognition device

Abstract

To provide an object recognition device capable of recognizing an object early from a weak reflection signal while suppressing erroneous recognition of noise as the object.SOLUTION: An object recognition device includes: a target sensor; tracking units S10 and S20; a detection probability calculation unit S30; a presence probability calculation unit S40; and recognition units S50 and S60. The detection probability calculation unit calculates higher detection probability as reflection intensity increases. The existence probability calculation unit calculates existence probability of an object corresponding to an observation value of a reflection position using the calculated detection probability. On condition that the calculated existence probability is equal to or larger than a predetermined value, the recognition unit recognizes a target being tracked as the object.SELECTED DRAWING: Figure 2

Description

Technical Field

[0001] The present disclosure relates to a technology for recognizing an object.

Background Art

[0002] The object detection device described in Patent Document 1 recognizes a reflection point continuously detected for a predetermined period as an object, and determines a reflection point not continuously detected for a predetermined period as a floating object such as fog or exhaust gas.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] Since the reflection signal from an object existing far away is weak, the detection of an object existing far away becomes unstable, and it is difficult to continuously detect an object existing far away for a predetermined period. Therefore, in the above object detection device, the recognition of an object existing far away may be delayed.

[0005] One aspect of the present disclosure provides an object recognition device capable of early recognizing an object from a weak reflection signal while suppressing misrecognition of noise as an object.

Means for Solving the Problems

[0006] An object recognition device according to one aspect of the present disclosure includes a target sensor (10), a tracking unit (20, S10, S20, S100, S110, S200, S210, S300, S310, S400, S410, S600, S610), a detection probability calculation unit (20, S30, S120, S220, S320, S450, S620), an existence probability calculation unit (20, S40, S130, S250, S360, S460, S650), and a recognition unit (20, S50, S60, S260, S270, S370, S380, S480, S490, S660, S670). The target sensor is configured to transmit a search wave to the surroundings, receive a reflected wave generated by the reflection of the search wave by the target, and periodically and repeatedly acquire an observed value of the reflection position of the target and an observed value of the reflection intensity based on the received reflected wave. The tracking unit is configured to sequentially estimate the state quantity of the target in the current processing cycle based on the observed value of the reflection position acquired by the target sensor and the state quantity of the target including at least the reflection position estimated in the previous processing cycle, and track the target. The detection probability calculation unit is configured to calculate the detection probability of the target in the current processing cycle based on the observed value of the reflection intensity acquired in the previous processing cycle or the reflection intensity included in the state quantity of the target. The detection probability calculation unit is configured to calculate a higher detection probability as the reflection intensity is stronger. The existence probability calculation unit is configured to calculate the existence probability of the target with respect to the observed value of the reflection position in the current processing cycle using the detection probability calculated by the detection probability calculation unit. The recognition unit is configured to recognize the target being tracked by the tracking unit as a target indicating an object on the condition that the existence probability calculated by the existence probability calculation unit is equal to or greater than a predetermined value.

[0007] The presence or absence of target detection is mainly affected by thermal noise, and it has been found that the correlation between the reflection intensity and the probability of detecting a target when the target exists at the reflection position (i.e., the detection probability) can be formulated. Therefore, in one aspect of the present disclosure, an object recognition apparatus calculates a detection probability according to the reflection intensity. Then, using the detection probability, the probability that a target exists at the reflection position (i.e., the existence probability) is calculated, and when the existence probability is equal to or greater than a predetermined value, the target being tracked is recognized. Accordingly, it is possible to early recognize a target indicating an object from a weak reflection signal while suppressing misrecognition of noise as an object.

Brief Description of the Drawings

[0008]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Modes for Carrying Out the Invention

[0009] Hereinafter, embodiments of the present disclosure will be described with reference to the drawings. (1. First Embodiment) <1-1. Configuration> The configuration of the object recognition apparatus 30 according to the present embodiment will be described with reference to FIG. 1.

[0010] The object recognition device 30 includes a target sensor 10 and a processing device 20. In the present embodiment, the object recognition device 30 is mounted on an automobile. The target sensor 10 periodically and repeatedly transmits a search wave to the surroundings, and receives a reflected wave generated by the reflection of the search wave by a target. Then, based on the received reflected wave, the target sensor 10 periodically and repeatedly acquires the reflection position of the search wave (i.e., the position of the target) and the reflection intensity of the reflected wave. The target sensor 10 is mounted, for example, at the center of the front bumper of the automobile. In the present embodiment, the physical quantity acquired by the target sensor 10 is referred to as an observation value.

[0011] In the present embodiment, the target sensor 10 is a lidar that emits laser light as a search wave. Note that the target sensor 10 is not limited to a lidar, and may be a millimeter-wave radar that emits millimeter waves as a search wave. The target sensor 10 may be any radar that emits electromagnetic waves as a search wave.

[0012] The processing device 20 is connected to the target sensor 10, and tracks and recognizes a target based on at least one observation value acquired by the target sensor 10. The processing device 20 includes a CPU 21, a ROM 22, and a RAM 23, and the CPU 21 realizes the functions of a tracking unit, a detection probability calculation unit, an existence probability calculation unit, and a recognition unit by executing a program stored in the ROM 22.

[0013] The tracking unit tracks the target detected by the target sensor 10. Specifically, the tracking unit associates the observation value of the target in the current processing cycle with the state quantity of the target in the previous processing cycle, and sequentially estimates the state quantity of the target in the current processing cycle from the associated observation value of the target and the state quantity of the target. The tracking unit associates the observation value of the target with the state quantity of the target that has the highest relevance to the observation value.

[0014] The reflection points that reflect the search wave include, in addition to the object, floating substances such as fog and exhaust gas (i.e., noise). The object corresponds to a three-dimensional object such as a vehicle ahead, a side wall, a guardrail, a falling object, or a pedestrian. The processing device 20 excludes noise from among the reflection points detected by the target sensor 10 and recognizes the reflection points indicating the object as targets.

[0015] The detection probability calculation unit calculates the detection probability P of the target in the current processing cycle based on the observed value of the reflection intensity of the target obtained in the previous processing cycle or the state quantity of the reflection intensity. d The detection probability Pd corresponds to the probability that the target is detected.

[0016] The existence probability calculation unit uses the calculated detection probability P d to calculate the existence probability V of the target for each observed value of the reflection position in the current processing cycle. k The existence probability V k corresponds to the probability that a target exists at the observed value of the reflection position.

[0017] The recognition unit recognizes the target being tracked on the condition that the calculated existence probability V k is equal to or greater than the probability threshold. That is, the recognition unit determines, on the condition that the existence probability V k is equal to or greater than the probability threshold, that the target being tracked is not a floating substance such as fog or exhaust gas (i.e., noise), but a target indicating an object such as a vehicle ahead.

[0018] <1-2. Processing> Next, the object recognition process executed by the processing device 20 will be described with reference to the flowchart of FIG. 2.

[0019] In S10, the processing device 20 acquires the observed values of the reflection intensity and the reflection position for each reflection point from the target sensor 10. Here, each reflection point is regarded as a target before recognition. The targets before recognition include noise such as floating substances in addition to the targets indicating the object.

[0020] Next, in S20, the state quantity of the target is estimated. That is, for each target, the observed value of the target obtained in the current processing cycle is associated with the state quantity of the target estimated in the previous processing cycle that has a high correlation with the observed value. Then, a Kalman filter or the like is applied to the associated observed value of the target and the state quantity of the target to estimate the state quantity of the target in the current processing cycle. The observed value and the state quantity of the target include at least the reflection position. The observed value and the state quantity of the target may include the reflection intensity in addition to the reflection position.

[0021] Next, in S30, for each target, the detection probability Pd of the target in the current processing cycle is calculated from the observed value of the reflection intensity obtained in the previous processing cycle. As a result of the inventor's study, it was found that the presence or absence of target detection is mainly affected by thermal noise, and the correlation between the reflection intensity and the detection probability can be formulated.

[0022] For example, when an object such as a vehicle ahead exists near the target sensor 10, the reflection intensity of the object becomes relatively strong, and the object is likely to be continuously detected. On the other hand, when an object such as a vehicle ahead exists far from the target sensor 10, the reflection intensity of the object becomes relatively weak, and the object is less likely to be continuously detected due to noise such as thermal noise.

[0023] Therefore, the detection probability calculation unit formulates the correlation between the reflection intensity and the detection probability Pd. Specifically, the detection probability calculation unit calculates a higher detection probability P d as the reflection intensity in the previous processing cycle is stronger.

[0024] The detection probability calculation unit approximates the distribution of the reflection intensity to a Gaussian distribution. Specifically, as shown in FIG. 3, the same value as the reflection intensity obtained in the previous processing cycle is used as the expected value of the reflection intensity obtained in the current processing cycle, and a predetermined error range centered on the expected value is set for the distribution of the reflection intensity. The approximated Gaussian distribution is a distribution that spreads within a predetermined error range with the expected value as the median.

[0025] Then, the detection probability calculation unit calculates the detection probability based on the ratio exceeding the intensity threshold set in the Gaussian distribution. That is, the detection probability calculation unit calculates the detection probability P d from the first area S1 and the second area S2 in the Gaussian distribution. The first area S1 corresponds to the area where values equal to or greater than the intensity threshold are distributed in the Gaussian distribution. The second area S2 corresponds to the area where values less than the intensity threshold are distributed in the Gaussian distribution. The intensity threshold is a value for distinguishing the reflection intensity of the target from noise such as thermal noise for detection. The detection probability P d is d calculated from the formula P = S1 / (S1 + S2).

[0026] Note that instead of the observed value of the reflection intensity obtained in the previous processing cycle, the state quantity of the reflection intensity estimated in the previous processing cycle may be used to calculate the detection probability P d . That is, the stronger the state quantity of the reflection intensity estimated in the previous processing cycle, the higher the detection probability P d of the object may be calculated.

[0027] Also, the distribution of the reflection intensity may be approximated by a distribution such as a Rayleigh distribution other than the Gaussian distribution. Further, the detection probability calculation unit may calculate the detection probability P d without using the distribution of the reflection intensity. For example, an expression representing the relationship between the reflection intensity and the detection probability P d may be derived, and the derived expression may be used to calculate the detection probability P d . Also, a correspondence table between the detection probability P d and the reflection intensity may be prepared in advance, and the prepared correspondence table may be used to calculate the detection probability P d .

[0028] Subsequently, in S40, for each target, using the detection probability P d calculated in S30 for the current processing cycle, the existence probability V k of the object corresponding to the observed value of the reflection position in the current processing cycle obtained in S10 is calculated. For example, the existence probability V kand the detection probability P calculated in the current processing cycle d are used to calculate (i) the probability that the target exists when the target is not detected in the current cycle, and (ii) the probability updated based on the observation value when the target is detected in the current cycle.

[0029] (i) corresponds to the probability of the existence of the target considering the case where the target is not detected despite its existence. (ii) corresponds to the probability of existence updated based on the observation value obtained when the target exists. By adding (i) and (ii), the probability of existence V k in the current processing cycle is calculated. The method for calculating the probability of the probability of existence V k is not particularly limited.

[0030] Thereby, the probability of existence V k of the target according to the reflection intensity and the reflection position is calculated. For example, the reflection intensity of the reflected wave from a distant object is relatively weak. Therefore, a distant object may or may not be detected in each processing cycle. Thus, when distinguishing between an object and a floating object and recognizing the object on the condition that it has been continuously detected for a predetermined period, the recognition of the distant object may be delayed. In contrast, in this embodiment, even if the detected object ceases to be detected, the probability of existence V k is not reset but maintained, so that the distant object is recognized earlier.

[0031] Subsequently, in S50, for each target, it is determined whether the probability of existence V k calculated in S40 is equal to or greater than a preset probability threshold. If it is determined that the probability of existence V k is equal to or greater than the probability threshold, the process proceeds to S60. If it is determined that the probability of existence V k is less than the probability threshold, the process proceeds to S70.

[0032] In S60, the target for which it is determined that the probability of existence V k is equal to or greater than the probability threshold is recognized as a target indicating an object. In S70, the probability of existence V kA target determined to have a probability below the probability threshold is not recognized as a target indicating an object.

[0033] <1-3. Effect> According to the first embodiment described in detail above, the following effects are achieved. (1) Detection probability P according to the reflection intensity d is calculated. Then, using the detection probability P d , the existence probability V of the target considering (i) the case where the target is not detected even though it exists, and (ii) the case where a false detection occurs even though the target does not exist k is calculated. And when the existence probability V k is equal to or greater than a predetermined value, the target before recognition during tracking is recognized as a target indicating an object. Therefore, it is possible to early recognize an object from a weak reflection signal while suppressing the misrecognition of noise as an object.

[0034] (2) By calculating the detection probability P d based on the observed value of the reflection intensity, the accuracy of the existence probability V k can be improved. (3) Since thermal noise follows a Gaussian distribution, by approximating the distribution of the reflection intensity to a Gaussian distribution, a weak reflection signal indicating an object can be extracted from the thermal noise, and the object can be recognized early.

[0035] (2. Second Embodiment) <2-1. Differences from the First Embodiment> In the second embodiment, since the basic configuration is the same as that of the first embodiment, the differences will be described below. Note that the same reference numerals as those in the first embodiment indicate the same configuration, and refer to the previous description.

[0036] In the second embodiment, target tracking processing (hereinafter referred to as RFS tracking) is performed by a state estimation filter based on the random finite set (hereinafter referred to as RFS) theory. Examples of the state estimation filter based on the RFS theory include a Probability Hypothesis Density (PHD) filter, a Cardinalized Probability Hypothesis Density (CPHD) filter, a Cardinality Balanced Multi-Bernoulli (CBMeMBer) filter, a Generalized Labeled Multi-Bernoulli (GLMB) filter, a Labeled Multi-Bernoulli (LMB) filter, a Poisson Multi-Bernoulli (PMB) filter, and a Poisson Multi-Bernoulli Mixture (PMBM) filter.

[0037] <2-2. Processing> Next, the object recognition processing executed by the processing device 20 according to the second embodiment will be described with reference to the flowchart of FIG. 4.

[0038] In S100 to S120, the same processing as that in S10 to S30 is executed by RFS tracking. In S130, the existence probability calculation unit calculates the existence probability Vk by RFS tracking from the reflection position and the detection probability Pd. For example, the existence probability calculation unit calculates the existence probability Vk using a PHD filter. Specifically, the existence probability Vk is calculated based on the following equations (1) and (2). The existence probability Vk calculated here corresponds to the expected value of the number of objects per unit area, that is, the intensity, and the higher the value, the higher the probability that an object exists.

[0039]

Equation

[0040]

Equation

[0041] Equation (1) is the probability of existence V in the previous processing cycle k-1 (x k-1 |z 1:k-1 ) to calculate the predicted value V of the probability of existence in this processing cycle k|k-1 (x k |z 1:k-1 ). x represents the state quantity and z represents the observation quantity. k represents the time. The first term on the right side of Equation (1) corresponds to the prediction of how the distribution of objects per unit area in the previous processing cycle has changed in this processing cycle, and corresponds to the integral of the survival probability × propagation function × previous probability of existence. The survival probability is the probability that the target does not disappear. The second term on the right side of Equation (1) corresponds to the branch from the distribution of objects per unit area in the previous processing cycle, and the third term on the right side of Equation (1) corresponds to the distribution of objects per unit area that occurs at time k.

[0042] Equation (2) is based on the predicted value V k|k-1 (x k |z 1:k-1 ) calculated in Equation (1) and the detection probability P d to calculate the probability of existence V k (x k |z 1:k ) in this processing cycle.

[0043] The first term on the right side of Equation (2) corresponds to (i) the expected value of the number of objects per unit area when the target is not detected in this processing cycle. The second term on the right side of Equation (2) corresponds to (ii) the expected value of the number of objects per unit area when the target is detected. K k (z) represents noise. Subsequently, in S140 to S160, the same processing as S50 to S70 is executed.

[0044] <2-3. Effect> According to the second embodiment described in detail above, the effects (1) to (3) of the first embodiment described above are achieved, and further, the following effects are achieved.

[0045] (4) Calculate the existence probability V using RSS supported by theory, thereby enhancing the reliability of the existence probability V calculated according to the reflection intensity. k k

[0046] (3. Third Embodiment) <3-1. Differences from the First Embodiment> In the third embodiment, since the basic configuration is the same as that of the first embodiment, the differences will be described below. The same reference numerals as those in the first embodiment denote the same configurations, and reference is made to the previous description.

[0047] The third embodiment is different from the first embodiment in that the processing device 20 has the function of an area estimation unit. The area estimation unit estimates the occurrence area where occlusion occurs. The detection probability calculation unit corrects the detection probability of the estimated occurrence area.

[0048] <3-2. Processing> Next, the object recognition processing executed by the processing device 20 according to the third embodiment will be described with reference to the flowchart of FIG. 5.

[0049] In S200 to S220, the same processing as S10 to S30 is executed. In S230, the area estimation unit estimates the occurrence area where occlusion occurs. Specifically, among the plurality of reflection positions acquired in S200, the reflection position existing behind other reflection positions is specified, and the area where the reflection position exists is set as the occurrence area.

[0050] Subsequently, in S240, among the detection probabilities calculated in S220, the detection probability in the occurrence area estimated in S230 is decreased. Subsequently, in S250 to S280, the same processing as S40 to S70 is executed.

[0051] <3-3. Effects> According to the third embodiment described in detail above, the effects (1) to (3) of the first embodiment described above are achieved, and further, the following effects are achieved. ​​

[0052] (5) Reducing the detection probability P in the occlusion occurrence area d can improve the accuracy of the detection probability P d and thus improve the accuracy of the existence probability V k .

[0053] (4. Fourth Embodiment) <4-1. Differences from the First Embodiment> Since the basic configuration of the fourth embodiment is the same as that of the third embodiment, the differences will be described below. The same reference numerals as those in the third embodiment denote the same configurations, and reference is made to the previous description.

[0054] In the above-described third embodiment, the processing device 20 estimated the occlusion occurrence area based on the reflection position observed by the target sensor 10. In contrast, in the fourth embodiment, the processing device 20 is different from the third embodiment in that it estimates the occlusion occurrence area based on the observation information of the environment sensor 40 different from the target sensor 10.

[0055] <4-2. Processing> Next, the object recognition processing executed by the processing device 20 according to the fourth embodiment will be described with reference to the flowchart of FIG. 6.

[0056] In S300 to S320, the same processing as in S200 to S220 is executed. In S330, the observation information observed by the environment sensor 40 is acquired. The environment sensor 40 is a sensor that observes the environment around the target sensor 10. The environment sensor 40 is, for example, an in-vehicle camera that captures the front of the vehicle. The observation information is, for example, a captured image captured by the in-vehicle camera.

[0057] In S340, based on the observation information acquired in S330, the occlusion occurrence area is estimated. When the observation information is a captured image, the area where the target is hidden behind other targets on the image is estimated as the occlusion occurrence area. Subsequently, in S350 to S390, the same processing as in S240 to S280 is executed.

[0058] <4-3. Effect> According to the fourth embodiment described in detail above, the same effects as the effects (1) to (3), (5) of the fourth embodiment described above are achieved.

[0059] (5. Fifth Embodiment) <5-1. Differences from the First Embodiment> Since the basic configuration of the fifth embodiment is the same as that of the first embodiment, the differences will be described below. Note that the same reference numerals as those in the first embodiment indicate the same configuration, and reference is made to the previous description.

[0060] In the fifth embodiment, the tracking unit calculates the difference between the observed value of the reflection intensity and the state quantity, and when the difference is equal to or greater than the difference threshold value, the existence probability in the previous processing cycle is retained, which is different from the first embodiment.

[0061] <5-2. Processing> Next, the object recognition processing executed by the processing device 20 according to the fifth embodiment will be described with reference to the flowchart of FIG. 7.

[0062] In S400, the same processing as in S10 is executed. Subsequently, in S410, the state quantity of the reflection intensity of each target estimated in the previous processing cycle is acquired.

[0063] Subsequently, in S420, for each target, the difference between the observed value of the reflection intensity acquired in S400 and the state quantity of the previous reflection intensity having the highest relevance to the observed value is calculated.

[0064] Subsequently, in S430, for each target, it is determined whether the difference calculated in S420 is less than or equal to a preset difference threshold value. If it is determined in S430 that the difference is less than or equal to the difference threshold value, the process proceeds to the processing of S440. In S440 to S460, the same processing as in S20 to S40 is executed.

[0065] On the other hand, in S430, when it is determined that the difference is greater than the difference threshold, the process proceeds to the process of S470. In S470, the existence probability V calculated in the previous processing cycle k is retained. That is, when the difference is greater than the difference threshold, the observed value of the reflection intensity is not associated with the state quantity, and the existence probability V k is not calculated. Subsequently, in S480 to S500, the same processing as in S50 to S70 is executed.

[0066] <5-3. Effect> According to the fifth embodiment described in detail above, the effects (1) to (3) of the first embodiment described above are achieved, and further, the following effects are achieved.

[0067] (6) By preventing the association of the observed value and the state quantity when the difference is greater than the difference threshold, the association of the observed value and the state quantity of different targets is prevented, and the detection probability P d can be suppressed from being calculated. As a result, the accuracy of the detection probability P d can be improved, and the accuracy of the existence probability V k can be improved.

[0068] (6. Sixth Embodiment) <6-1. Differences from the First Embodiment> Since the basic configuration of the sixth embodiment is the same as that of the first embodiment, the differences will be described below. Note that the same reference numerals as those in the first embodiment indicate the same configuration, and reference is made to the previous description.

[0069] In the sixth embodiment, the processing device 20 is different from the first embodiment in that it acquires environmental information from the environmental sensor 40 and corrects the detection probability when the environmental information includes adverse environmental information.

[0070] <6-2. Processing> Next, the object recognition processing executed by the processing device 20 according to the sixth embodiment will be described with reference to the flowchart of FIG. 8.

[0071] In S600 to S620, the same processing as that in S10 to S30 is executed. In S630, the observation information observed by the environment sensor 40 is acquired. The environment sensor 40 is a sensor that observes the environment around the target sensor 10. The environment sensor 40 is, for example, an in-vehicle camera that captures the front of the vehicle. The observation information is, for example, a captured image captured by the in-vehicle camera.

[0072] Subsequently, in S640, when the observation information acquired in S630 includes bad environment information, the detection probability of each target calculated in S620 is decreased. The bad environment information corresponds to environment information that degrades the detection performance of the target sensor 10 and is, for example, rainy weather information. Subsequently, in S650 to S680, the same processing as that in S40 to S70 is executed.

[0073] <6-3. Effect> According to the sixth embodiment described in detail above, the effects (1) to (3) of the first embodiment described above are achieved, and further, the following effects are achieved.

[0074] (7) When the environment around the target sensor 10 is an environment that deteriorates the detection situation of the target, by decreasing the detection probability P d the accuracy of the detection probability P d can be improved. As a result, the accuracy of the existence probability V k can be improved.

[0075] (7. Other Embodiments) As described above, the embodiments of the present disclosure have been described. However, the present disclosure is not limited to the above-described embodiments and can be implemented with various modifications.

[0076] (a) The multiple functions of one component in the above-described embodiment may be realized by multiple components, or one function of one component may be realized by multiple components. Also, the multiple functions of multiple components may be realized by one component, or one function realized by multiple components may be realized by one component. Further, a part of the configuration of the above-described embodiment may be omitted. Also, at least a part of the configuration of the above-described embodiment may be added to or replaced with the configuration of another of the above-described embodiments.

[0077] (b) In addition to the object recognition device described above, the present disclosure can also be realized in various forms such as a system including the object recognition device as a component, a program for causing a computer to function as the object recognition device, a non-transitory tangible recording medium such as a semiconductor memory recording this program, and an object recognition method.

Explanation of Reference Numerals

[0078] 10... object sensor, 20... processing device, 30... object recognition device.

Claims

1. A target sensor (10) configured to transmit a search wave to the surroundings, receive a reflected wave generated by reflection of the search wave by a target, and periodically and repeatedly acquire an observed value of the reflection position of the target and an observed value of the reflection intensity based on the received reflected wave; A tracking unit (20, S10, S20, S100, S110, S200, S210, S300, S310, S400, S410, S600, S610) configured to sequentially estimate the state quantity of the target in the current processing cycle based on the observed value of the reflection position obtained by the target sensor and the state quantity of the target including at least the reflection position estimated in the previous processing cycle, and to track the target; A detection probability calculation unit configured to calculate the detection probability of the target in the current processing cycle based on the observed value of the reflection intensity obtained in the previous processing cycle or the reflection intensity included in the state quantity of the target, and configured to calculate the detection probability higher as the reflection intensity is stronger (20, S30, S120, S220, S320, S450, S620); An existence probability calculation unit (20, S40, S130, S250, S360, S460, S650) configured to calculate the existence probability of the target with respect to the observed value of the reflection position in the current processing cycle using the detection probability calculated by the detection probability calculation unit; A recognition unit (20, S50, S60, S260, S270, S370, S380, S480, S490, S660, S670) configured to recognize the target being tracked by the tracking unit as a target indicating an object on the condition that the existence probability calculated by the existence probability calculation unit is equal to or greater than a predetermined value; An object recognition device.

2. The state quantity of the target includes the reflection position and the reflection intensity, The tracking unit is configured to sequentially estimate the state quantity of the target in the current processing cycle based on the observed value of the reflection position and the observed value of the reflection intensity and the state quantity of the target estimated in the previous processing cycle, The detection probability calculation unit is configured to calculate the detection probability based on the observed value of the reflection intensity. The object recognition device according to Claim 1.

3. The existence probability calculation unit is configured to calculate the existence probability using a random finite set. The object recognition device according to claim 1 or 2.

4. The detection probability calculation unit is configured to approximate the distribution of the reflection intensity with a Gaussian distribution, and calculate the detection probability based on a ratio exceeding an intensity threshold in the distribution of the reflection intensity approximated by the Gaussian distribution. The object recognition device according to any one of claims 1 to 3.

5. Further comprising a region estimation unit (20, S230, S340) for estimating a generation region where occlusion occurs. The detection probability calculation unit is configured to reduce the detection probability in the generation region estimated by the region estimation unit. The object recognition device according to any one of claims 1 to 4.

6. When the difference between the observed value of the reflection intensity and the reflection intensity included in the state quantity of the target is greater than a threshold value, the tracking unit is configured not to associate the observed value of the reflection position and the observed value of the reflection intensity with the state quantity of the target estimated in the previous processing cycle. The object recognition device according to claim 2.

7. The detection probability calculation unit is configured to reduce the detection probability when obtaining bad environment information from an environmental sensor. The object recognition device according to any one of claims 1 to 6.

8. The object recognition device is mounted on an automobile. The object recognition device according to any one of claims 1 to 7.

Images