JP6333437B1 - Object recognition processing device, object recognition processing method, and vehicle control system - Google Patents

Abstract

An object recognition processing device and an object recognition processing method for obtaining a relative position of an object relative to a host vehicle more accurately than a conventional vehicle, and a vehicle control system including the object recognition processing device. In the object recognition processing device and the object recognition processing method, when the second observation data is captured in the prediction data at time tk and the first detection position D1 (tk) is detected, the prediction position P is detected. The wake position T (tk) is generated using (tk) and the update detection position D1r (tk). [Selection] Figure 1

Description

  The present invention relates to an object recognition processing device and an object recognition processing method for acquiring a relative position of an object with respect to a host vehicle, and a vehicle control system including the object recognition processing device.

  2. Description of the Related Art Conventionally, a vehicle to which an automatic driving technique or a preventive safe driving technique is applied includes an in-vehicle ranging device that measures a distance between the own vehicle and an object around the own vehicle. The measurement results from the in-vehicle ranging device show that the collision damage reduction brake system that reduces damage when the host vehicle collides with an object ahead, the adaptive cruise control system that the host vehicle follows the vehicle ahead, and the host vehicle automatically Used in automatic parking systems that park in parking spaces. As described above, the in-vehicle distance measuring device is for improving the safety of driving of the own vehicle and the comfortable driving of the own vehicle, and is used in a vehicle application.

  Here, as an in-vehicle distance measuring device, a radar device (for example, see Patent Document 1) is often used. Generally, a radar apparatus measures the relative position and relative speed of an object with respect to the host vehicle by transmitting the radio wave to an object around the host vehicle and receiving the radio wave reflected from the object. Strictly speaking, the detection position of the object detected by the radar apparatus is not the center position of the object but the reflection point position where the object reflects the radio wave transmitted from the radar apparatus. Therefore, the reflection point varies greatly depending on the positional relationship between the host vehicle and the object.

  The radar apparatus described in Patent Literature 1 detects the distance from the own vehicle to the preceding vehicle and the direction of the preceding vehicle with respect to the own vehicle. The radar apparatus is configured to calculate the position of the preceding vehicle with respect to the length direction of the own vehicle and the center position of the preceding vehicle with respect to the width direction of the own vehicle from the detected distance and direction. With this configuration, a countermeasure against a change in the detection position of the radar due to the reflection point is realized.

JP 2004-233275 A

  In the prior art described in Patent Document 1, a correlation between a detection position detected by a radar device and a predicted position generated using a wake position at a time earlier than the time at which the detection position was detected is obtained. No particular mention is made of the method.

  For example, when the correlation is determined by the distance between the detected position and the predicted position corresponding to the time when the detected position was detected, the detected position varies greatly depending on the reflection point, and as a result, the detected position and the predicted position Cannot be accurately correlated. In such a case, the relative position of the object with respect to the host vehicle cannot be obtained accurately.

  The present invention has been made to solve the above-described problems, and compared with the conventional art, an object recognition processing device and an object recognition processing method that aim to more accurately acquire the relative position of an object with respect to the host vehicle, An object is to obtain a vehicle control system including the object recognition processing device.

The object recognition processing device according to the present invention transmits a detection wave to an object around the host vehicle, receives a detection wave reflected from the object, and receives observation data including a first detection position based on the received detection wave. A first object detection unit that detects as one observation data; a second object detection unit that detects observation data including a second detection position and an object size as second observation data in a detection method different from the first object detection unit; observation data using the track data including the track position of the time tk-1 before the time tk which is detected, to predict the trajectory data including track position at time tk, the prediction comprising the prediction position of the time tk When the second observation data is already detected before the time tk−1 and the first observation data at the time tk is detected, the prediction position included in the prediction data at the time tk is detected. Yo If it contains correction processing unit and the correlation predicted position in the correction data at time tk for generating correction data including the correlation predicted position at time tk by correcting the predicted position based on the object size, the time that there is a correlation A correlation processing unit that generates correlation data including a combination of the prediction position for correlation of tk and the first detection position, and a correlation prediction when the combination of the correlation prediction position and the first detection position is included in the correlation data at time tk The detection position for update at time tk is generated by correcting the first detection position based on the correction of the prediction position when the position is generated, and the prediction position corresponding to the correlation prediction position and the detection position for update are obtained. And an update processing unit that generates wake data including the wake position at time tk .

In the object recognition processing method according to the present invention, a detection wave is transmitted to an object around the host vehicle, a detection wave reflected from the object is received, and observation data including a first detection position is obtained based on the received detection wave. A first object detection step for detecting as one observation data; a second object detection step for detecting observation data including a second detection position and an object size as second observation data by a detection method different from the first object detection step; observation data using the track data including the track position of the time tk-1 before the time tk which is detected, to predict the trajectory data including track position at time tk, the prediction comprising the prediction position of the time tk a prediction processing step of generating data, and the second observation data is detected already at time tk-1 earlier, if the first observation data at time tk is detected, is included in the prediction data A correction processing step of generating correction data including the correlation predicted position by correcting the predicted position based on the predicted position and the object size, if it contains the correlation predicted position in the correction data at time tk, the time that there is a correlation Correlation processing step for generating correlation data including a combination of the prediction position for correlation of tk and the first detection position, and when the correlation data of time tk includes a combination of the prediction position for correlation and the first detection position The detection position for update at time tk is generated by correcting the first detection position based on the correction of the prediction position when the position is generated, and the prediction position corresponding to the correlation prediction position and the detection position for update are obtained. And an update processing step for generating wake data including the wake position at time tk .

  The vehicle control system according to the present invention includes an object recognition processing device and a vehicle control device that controls the host vehicle based on track data generated by an update processing unit of the object recognition processing device.

  According to the present invention, an object recognition processing device and an object recognition processing method for obtaining a relative position of an object with respect to the host vehicle more accurately than before, and a vehicle control system including the object recognition processing device. Can be obtained.

It is a block diagram which shows the structure of the vehicle control system provided with the object recognition processing apparatus in Embodiment 1 of this invention. It is a flowchart which shows the production | generation process of the prediction data performed by the prediction process part in Embodiment 1 of this invention. It is a flowchart which shows the production | generation process of the correction data performed by the correction process part in Embodiment 1 of this invention. It is explanatory drawing which shows the concept of the prediction position for a correlation produced | generated by the correction process part in Embodiment 1 of this invention. It is explanatory drawing which shows the concept of the detection position for a correlation produced | generated by the correction process part in Embodiment 1 of this invention. It is a conceptual diagram which shows the method in which the correlation process part in Embodiment 1 of this invention takes the correlation of the prediction position for a correlation, and a 1st detection position. It is a conceptual diagram which shows the method in which the correlation process part in Embodiment 1 of this invention takes the correlation of a predicted position and the detection position for correlation. It is a flowchart which shows the production | generation process of the track data performed by the update process part in Embodiment 1 of this invention. It is explanatory drawing which shows the concept of the detection position for an update produced | generated by the update process part in Embodiment 1 of this invention. It is explanatory drawing which shows the concept of the prediction position for an update produced | generated by the update process part in Embodiment 1 of this invention.

  Hereinafter, an object recognition processing device and an object recognition processing method according to the present invention, and a vehicle control system including the object recognition processing device will be described with reference to the drawings according to a preferred embodiment. In the description of the drawings, the same portions or corresponding portions are denoted by the same reference numerals, and redundant description is omitted.

Embodiment 1 FIG.
FIG. 1 is a block diagram showing a configuration of a vehicle control system including an object recognition processing device 1 according to Embodiment 1 of the present invention. Note that the arrows in FIG. 1 indicate the flow of signals.

  The vehicle control system shown in FIG. 1 is provided in a host vehicle and includes an object recognition processing device 1 and a vehicle control device 2. The object recognition processing device 1 includes a first object detection unit 11, a second object detection unit 12, and an object recognition processing unit 13.

  The first object detection unit 11 and the second object detection unit 12 detect an object around the host vehicle, and further detect observation data including a detection position that is a relative position of the object with respect to the host vehicle.

  Specifically, the first object detection unit 11 transmits detection waves such as light, electromagnetic waves, and ultrasonic waves to objects around the own vehicle, receives detection waves reflected from the objects, and receives the detected waves. Based on this, the first observation data including the first detection position is detected. The first object detection unit 11 outputs the detected first observation data to the object recognition processing unit 13. In addition to the first detection position, the first observation data may further include accuracy information of the first detection position.

  The first object detection unit 11 is configured using, for example, a sensor such as a millimeter wave radar, a laser radar, or an ultrasonic sensor, and a microcomputer that calculates the first observation data by analyzing the output of the sensor. . The first detection position detected by the first object detection unit 11 is a reflection point position where the object reflects the detection wave transmitted by the first object detection unit 11.

  The second object detection unit 12 receives detection waves such as light and electromagnetic waves transmitted from the object, and based on the received detection waves, the second object detection unit 12 includes a second detection position and assumes that the shape of the detected object is a rectangle. Second observation data further including an object size indicating the length of the object in the width direction (hereinafter referred to as the X direction) and the length direction of the object (hereinafter referred to as the Y direction) is detected. The second object detection unit 12 outputs the detected second observation data to the object recognition processing unit 13. The second observation data may further include the direction (orientation) of the object with respect to the host vehicle in addition to the second detection position and the object size.

  As an example of a method for detecting the object size, there is a method for directly calculating the lengths in the X direction and the Y direction assuming that the shape of the detected object is a rectangle. In this case, the second observation data may further include accuracy information of the object dimension in addition to the second detection position and the object dimension.

  Further, as an example of a method for detecting an object size, there is a method in which an object type and an object size for each type are defined in advance, the detected object type is specified, and an object size corresponding to the type is selected. Examples of the object type include an automobile, a truck, a bicycle, a person, a sign, and a pole. In this case, the second observation data may further include reliability information of the object type detected for selecting the object size in addition to the second detection position and the object size. As a method for detecting the reliability of the object type, for example, there is a method of detecting a high reliability according to the number of times the object type is specified by the second object detection unit 12.

  The second object detection unit 12 calculates the second observation data by analyzing a sensor such as an optical camera, a vehicle-to-vehicle communication device corresponding to a device for communicating between a plurality of vehicles, and the output of the sensor. Configured with a microcomputer. The second detection position detected by the second object detection unit 12 is the center position of the edge of the detected object, that is, the center position on the short side of the rectangle when the shape of the detected object is assumed to be a rectangle. .

  As described above, the second object detection unit 12 detects the second observation data including the second detection position and the object size by a detection method different from that of the first object detection unit 11.

  The object recognition processing unit 13 is, for example, a microcomputer that executes arithmetic processing, a ROM (Read Only Memory) that stores data such as program data and fixed value data, and the stored data are updated and sequentially rewritten. This is realized by a RAM (Random Access Memory). The object recognition processing unit 13 includes a prediction processing unit 14, a correction processing unit 15, a correlation processing unit 16, and an update processing unit 17.

  The prediction processing unit 14 predicts track data at time tk (current time) when observation data is detected using track data described later, and generates a prediction result as prediction data at time tk. Specifically, the prediction processing unit 14 predicts the track data at the time tk using the track data at the time tk-1 (past time) immediately before the time tk, and the prediction result at the time tk. Generate as prediction data. The prediction processing unit 14 outputs the generated prediction data to the correction processing unit 15.

  Next, generation of prediction data performed by the prediction processing unit 14 will be further described with reference to FIG. FIG. 2 is a flowchart showing prediction data generation processing performed by the prediction processing unit 14 according to Embodiment 1 of the present invention.

  In step S <b> 101, the prediction processing unit 14 determines whether or not the second observation data is captured in the track data at time tk−1 input from the update processing unit 17. If it is determined that the second observation data is captured in the track data at time tk-1, the process proceeds to step S102. On the other hand, if it is determined that the second observation data is not captured in the track data at time tk-1, the process proceeds to step S103.

  When the process proceeds to step S102, that is, when the second observation data is captured in the track data at time tk-1, the second object detection unit 12 has already detected the observation data of the object before time tk-1. Will be doing. In this case, the track data at time tk-1 includes the track position T (tk-1) and the object size.

  In step S102, the prediction processing unit 14 generates prediction data at time tk using the wake position T (tk-1) and the object size included in the wake data at time tk-1, and the process ends.

  Specifically, the prediction processing unit 14 generates a predicted position P (tk) at time tk using the track position T (tk-1). Further, since the object size is included in the track data at time tk-1, the prediction processing unit 14 uses the object size as it is as the object size included in the prediction data at time tk.

  When the process proceeds to step S103, that is, when the second observation data is not captured in the track data at time tk-1, the second object detection unit 12 still detects the observation data of the object before time tk-1. It will not be. In this case, the track data at time tk-1 includes the track position T (tk-1).

  In step S103, the prediction processing unit 14 generates prediction data at time tk using the track position T (tk-1) included in the track data at time tk-1, and the process ends.

  As described above, the prediction processing unit 14 uses the wake data including the wake position T (tk-1) at the time tk-1 before the time tk at which the observation data is detected, to use the wake position T ( Prediction data including the predicted position P (tk) is generated by predicting the wake data including tk).

  In addition, as a method of predicting the predicted position P (tk) using the wake position T (tk-1), various known methods can be cited, and for example, the following methods can be cited. That is, for example, the wake data further includes the relative speed of the object with respect to the host vehicle, the difference between the time tk−1 and the time tk, the relative speed and the wake position T (tk) included in the wake data at the time tk−1. -1) and predicting the predicted position P (tk). As described above, for example, the prediction processing unit 14 obtains the wake data including the wake position T (tk-1) and the relative speed of the object at time tk-1, and the elapsed time from time tk-1 to time tk. To generate the predicted position P (tk).

  A series of processes in the flowchart shown in FIG. 2 is performed for each observation data for all the observation data input from the object detection unit.

  As the correction process, the correction processing unit 15 corrects the predicted position P (tk) included in the prediction data at time tk or the second detection position D2 (tk) included in the second observation data at time tk, and the correction. The result is generated as correction data. The correction processing unit 15 outputs the generated correction data to the correlation processing unit 16.

  Subsequently, generation of correction data performed by the correction processing unit 15 will be further described with reference to FIG. FIG. 3 is a flowchart showing correction data generation processing performed by the correction processing unit 15 according to Embodiment 1 of the present invention.

  In step S <b> 201, the correction processing unit 15 determines whether or not the second observation data is captured in the prediction data at the time tk input from the prediction processing unit 14. If it is determined that the second observation data is captured in the prediction data at time tk, the process proceeds to step S202. On the other hand, if it is determined that the second observation data is not captured in the prediction data at time tk, the process proceeds to step S205.

  When the process proceeds to step S202, that is, when the second observation data is captured in the prediction data at time tk, the second object detection unit 12 has already detected the observation data of the object before time tk-1. Will be. In this case, the predicted data at time tk includes the predicted position P (tk) and the object size.

  In step S202, the correction processing unit 15 determines whether the object detection unit that has detected the observation data at time tk is the first object detection unit 11 or the second object detection unit 12. If the object detection unit that has detected the observation data at time tk is the first object detection unit 11, the process proceeds to step S203.

  On the other hand, when the object detection unit that has detected the observation data at time tk is the second object detection unit 12, the process ends. In this case, the correction processing unit 15 outputs the prediction data at time tk to the correlation processing unit 16 as it is without performing correction processing.

  In step S203, the correction processing unit 15 uses the predicted position P (tk) and the object size included in the predicted data at time tk to correct the predicted position P (tk) according to the correction map prepared in advance. One correction amount is calculated, and the process proceeds to step S204.

  In step S204, the correction processing unit 15 generates correction data at time tk using the first correction amount calculated in step S203, and the process ends.

  Specifically, the correction processing unit 15 corrects the predicted position P (tk) at the time tk using the first correction amount, and the correction data including the correlation predicted position Pc (tk), which is the correction result. Generate.

  Here, generation of the predicted position for correlation Pc (tk) performed by the correction processing unit 15 will be further described with reference to FIG. FIG. 4 is an explanatory diagram showing the concept of the correlation predicted position Pc (tk) generated in the correction processing unit 15 according to Embodiment 1 of the present invention.

  In FIG. 4, the second object detection unit 12 has already detected observation data of an object (here, another vehicle) before the time tk−1, and the first object detection unit 11 at the time tk. Considers a case in which observation data including the first detection position D1 (tk) of the object is detected.

  As described above, the prediction processing unit 14 generates prediction data at the time tk using the track data at the time tk-1. In this case, the predicted data at time tk includes the predicted position P (tk) and the object size. The correction processing unit 15 calculates the first correction amount according to a correction map prepared in advance using the predicted position P (tk) and the object size included in the predicted data at time tk.

  Here, if the same object is detected by both the first object detection unit 11 and the second object detection unit 12 at the same time, the deviation between the first detection position and the second detection position in the X direction and the Y direction will be described. Occurs. The amount of deviation can be estimated if the first detection position, the second detection position, and the object dimensions are known.

  Therefore, by performing experiments, simulations, and the like in advance while changing the first detection position, the second detection position, and the object size as parameters, the amount of deviation is grasped as a correction amount. Then, a correction map in which the second detection position, the object dimension, and the correction amount are associated is prepared in advance.

  In step S203, the correction processing unit 15 regards the predicted position P (tk) included in the predicted data at time tk as the second detection position in the correction map, and corresponds to the predicted position P (tk) and the object size. One correction amount is calculated from the correction map.

  The correction processing unit 15 is configured to calculate the correction amount in consideration of the orientation of the object in addition to the predicted position P (tk) and the object size included in the prediction data at the time tk. Also good. Further, in the case of such a configuration, in order to correct the predicted position P (tk) to the closest point (nearest point) to the host vehicle and the object based on the predicted position P (tk), the object size, and the direction of the object. May be calculated.

  In step S204, the correction processing unit 15 shifts the predicted position P (tk) according to the first correction amount calculated in step S203 on the XY coordinates composed of the X direction and the Y direction, as shown in FIG. Thus, the correlation predicted position Pc (tk) is generated.

  Thus, when the second observation data is captured in the prediction data P (tk) and the first observation data is detected, the correction processing unit 15 includes the prediction position P (tk) and the object included in the prediction data. Correction data including the correlation predicted position Pc (tk) is generated by correcting the predicted position P (tk) based on the dimensions.

  Returning to the description of FIG. 3, when the process proceeds to step S <b> 205, that is, when the second observation data is not captured in the prediction data at time tk, the second object detection unit 12 detects the object before time tk−1. The observation data has not been detected yet. In this case, the predicted position P (tk) is included in the track data at time tk-1.

  In step S205, the correction processing unit 15 determines whether the object detection unit that has detected the observation data at time tk is the first object detection unit 11 or the second object detection unit 12. If the object detection unit that has detected the observation data at time tk is the second object detection unit 12, the process proceeds to step S206.

  On the other hand, when the object detection unit that has detected the observation data at time tk is the first object detection unit 11, the process ends. In this case, the correction processing unit 15 outputs the prediction data at time tk to the correlation processing unit 16 as it is without performing correction processing.

  In step S206, the correction processing unit 15 uses the second detection position D2 (tk) and the object size included in the second observation data at time tk to determine the second detection position D2 (tk) according to the correction map. A second correction amount for correction is calculated, and the process proceeds to step S207.

  In addition to the second detection position D2 (tk) and the object size included in the second observation data at time tk, the correction processing unit 15 further calculates the correction amount in consideration of the direction of the object. It may be configured. Moreover, when comprised in this way, based on 2nd detection position D2 (tk), an object dimension, and direction of an object, 2nd detection position D2 (tk) is a point (nearest point) nearest to the own vehicle and an object. You may calculate the correction amount for correct | amending.

  In step S207, the correction processing unit 15 generates correction data at time tk using the second correction amount calculated in step S206, and the process ends.

  Specifically, the correction processing unit 15 corrects the second detection position D2 (tk) at the time tk using the second correction amount, and includes the correlation detection position D2c (tk) that is the correction result. Generate data.

  Here, generation of the correlation detection position D2c (tk) performed by the correction processing unit 15 will be further described with reference to FIG. FIG. 5 is an explanatory diagram showing the concept of the correlation detection position D2c (tk) generated in the correction processing unit 15 according to Embodiment 1 of the present invention.

  In FIG. 5, the first object detection unit 11 has already detected the observation data of the object (here, another vehicle) before the time tk−1, while the second object detection unit 12 has detected the object. In this case, the second object detection unit 12 detects the observation data including the second detection position D2 (tk) of the object at time tk.

  As described above, the prediction processing unit 14 generates prediction data at the time tk using the track data at the time tk-1. In this case, the prediction position P (tk) is included in the prediction data at time tk. The correction processing unit 15 calculates the second correction amount according to the correction map using the second detection position D2 (tk) and the object size included in the second observation data at time tk.

  In step S206, the correction processing unit 15 regards the second detection position D2 (tk) included in the second observation data at time tk as the second detection position in the correction map, and determines the second detection position D2 (tk) and A second correction amount corresponding to the object size is calculated from the correction map.

  In step S207, the correction processing unit 15 shifts the second detection position D2 (tk) according to the second correction amount calculated in step S206 on the XY coordinates composed of the X direction and the Y direction, as shown in FIG. As a result, the correlation detection position D2c (tk) is generated.

  As described above, when the second observation data is not captured in the prediction data and the second observation data is detected, the correction processing unit 15 includes the second detection position D2 (tk) included in the second observation data and Correction data including the correlation detection position D2c (tk) is generated by correcting the second detection position D2 (tk) based on the object size.

  As can be seen from the above, the correction processing unit 15 detects the reflection point position as the detection position of the object, the detection result by the first object detection unit 11 of the detection method, and the second detection method different from the first object detection unit 11. It plays a role in matching the detection result of the object detection unit. Therefore, the correlation processing unit 16 to be described later can accurately select a combination of correlated object prediction results and object detection results. Further, it is possible to improve the accuracy of the wake data generated by the update processing unit 17 described later.

  A series of processes in the flowchart illustrated in FIG. 3 is performed for each prediction data with respect to all the prediction data input from the prediction processing unit 14.

  The correlation processing unit 16 individually combines the correction data at all times tk and the observation data at all times tk, and outputs a combination having a correlation from these combinations to the update processing unit 17 as correlation data. .

  Here, generation of correlation data performed by the correlation processing unit 16 will be further described with reference to FIG. FIG. 6 is a conceptual diagram illustrating how the correlation processing unit 16 according to Embodiment 1 of the present invention obtains a correlation between the correlation predicted position Pc (tk) and the first detection position D1 (tk).

  In FIG. 6, at time tk, the observation data including the first detection position D1 (tk) _1 and the observation data including the first detection position D1 (tk) _2 are detected by the first object detection unit 11. Furthermore, a case is considered in which correction data including the correlation predicted position Pc (tk) _1 and correction data including the correlation predicted position Pc (tk) _2 are generated by the correction processing unit 15.

  In this case, the correlation processing unit 16 calculates the distance by individually combining all the predicted positions for correlation Pc (tk) and all the first detection positions D1 (tk), and the distance is the largest among these combinations. Short combinations are generated as correlation data. For example, in the situation shown in FIG. 6, the combination of the correlation predicted position Pc (tk) _1 and the first detection position D1 (tk) _2, the correlation prediction position Pc (tk) _2, and the first detection position D1 (tk). A combination of _1 is generated as correlation data.

  As described above, when the correction data includes the correlation predicted position Pc (tk), the correlation processing unit 16 includes a combination of the correlation predicted position Pc (tk) having the correlation and the first detection position D1 (tk). Generate correlation data.

  Subsequently, generation of correlation data performed by the correlation processing unit 16 will be further described with reference to FIG. FIG. 7 is a conceptual diagram illustrating how the correlation processing unit 16 according to Embodiment 1 of the present invention obtains a correlation between the predicted position P (tk) and the correlation detection position D2c (tk).

  In FIG. 7, at time tk, prediction data including the prediction position P (tk) _1 and prediction data including the prediction position P (tk) _2 are generated by the prediction processing unit 14, and further, a detection position for correlation A case is considered in which the correction processing unit 15 generates correction data including D2c (tk) _1 and correction data including the detection position for correlation D2c (tk) _2.

  In this case, the correlation processing unit 16 calculates the distance by individually combining all the predicted positions P (tk) and all the first detection positions D1 (tk), and among these combinations, the combination having the shortest distance Are generated as correlation data. For example, in the situation shown in FIG. 7, a combination of the predicted position P (tk) _1 and the detection position for correlation D2c (tk) _2, and a combination of the predicted position P (tk) _2 and the detection position for correlation D2c (tk) _1. And generated as correlation data.

  As described above, when the correction data includes the correlation detection position D2c (tk), the correlation processing unit 16 includes the correlation data including the combination of the predicted position P (tk) having the correlation and the detection position D2c (tk) for correlation. Is generated.

  In addition, when correction processing is not performed by the correction processing unit 15 at time tk, the prediction data at all times tk and the observation data at all times tk are individually combined. Similarly, a combination with correlation is output to the update processing unit 17 as correlation data.

  The update processing unit 17 generates wake data using the correlation data input from the correlation processing unit 16.

  Here, the generation of wake data performed by the update processing unit 17 will be described with reference to FIG. FIG. 8 is a flowchart showing wake data generation processing performed by the update processing unit 17 according to Embodiment 1 of the present invention.

  In step S301, the update processing unit 17 determines whether the correlation data input from the correlation processing unit 16 includes a combination of the correlation predicted position Pc (tk) and the first detection position D1 (tk). If the combination is included in the correlation data, the process proceeds to step S302. If the combination is not included in the correlation data, the process proceeds to step S304.

  In step S302, the update processing unit 17 uses the first correction amount calculated by the correction processing unit 15 and corresponding to the correlation predicted position Pc (tk) to determine the first detection position D1 (tk) at time tk. Correction is performed to generate an update detection position D1r (tk) as the correction result, and the process proceeds to step S303.

  Here, generation of the update detection position D1r (tk) performed by the update processing unit 17 will be described with reference to FIG. FIG. 9 is an explanatory diagram showing the concept of the update detection position D1r (tk) generated by the update processing unit 17 according to the first embodiment of the present invention. In FIG. 9, the same case as in FIG. 4 is considered.

  As described above, the correction processing unit 15 generates the correlation predicted position Pc (tk) by correcting the predicted position P (tk) using the first correction amount. In addition, the correlation processing unit 16 generates a combination of the correlation predicted position Pc (tk) and the first detection position D1 (tk) as correlation data.

  As illustrated in FIG. 9, the update processing unit 17 predicts the first detection position D1 (tk) for correlation according to the first correction amount used when correcting the predicted position P (tk) on the XY coordinates. The detection position for updating D1r (tk) is generated by shifting the predicted position P (tk) when the position Pc (tk) is generated in the direction opposite to the shifted direction.

  In step S303, the update processing unit 17 uses the predicted position P (tk) corresponding to the correlation predicted position Pc (tk) and the update detection position D1r (tk) as shown in FIG. The wake-up data including T (tk) is generated, and the process ends. Examples of a method for generating the wake position T (tk) using the predicted position P (tk) and the update detection position D1r (tk) include a method using an α-β filter, a Kalman filter, and the like.

  As described above, when the correlation data includes a combination of the correlation predicted position Pc (tk) and the first detection position D1 (tk), the update processing unit 17 generates the correlation predicted position Pc (tk). The detection position for updating D1r (tk) is generated by correcting the first detection position D1 (tk) based on the correction of the predicted position P (tk). Further, the update processing unit 17 uses the prediction position P (tk) corresponding to the correlation prediction position Pc (tk) and the update detection position D1r (tk), and the wake data including the wake position T (tk). Is generated.

  In step S304, the update processing unit 17 determines whether or not the correlation data input from the correlation processing unit 16 includes a combination of the predicted position P (tk) and the correlation detection position D2c (tk). If the combination exists in the correlation data, the process proceeds to step S305. If the combination does not exist in the correlation data, the process proceeds to step S307.

  In step S305, the update processing unit 17 corrects the predicted position P (tk) at the time tk using the second correction amount calculated by the correction processing unit 15 and corresponding to the correlation detection position D2c (tk). Then, an update predicted position Pr (tk) that is the correction result is generated, and the process proceeds to step S306.

  Here, the generation of the update predicted position Pr (tk) performed by the update processing unit 17 will be described with reference to FIG. FIG. 10 is an explanatory diagram showing the concept of the update predicted position Pr (tk) generated by the update processing unit 17 according to Embodiment 1 of the present invention. In FIG. 10, the same case as in FIG. 5 is considered.

  As described above, the correction processing unit 15 generates the correlation detection position D2c (tk) by correcting the second detection position D2 (tk) using the second correction amount. In addition, the correlation processing unit 16 generates a combination of the predicted position P (tk) and the detection position for correlation D2c (tk) as correlation data.

  As shown in FIG. 10, the update processing unit 17 detects the predicted position P (tk) for correlation according to the second correction amount used when correcting the second detection position D2 (tk) on the XY coordinates. The update predicted position Pr (tk) is generated by shifting the second detection position D2 (tk) when the position D2c (tk) is generated in the direction opposite to the shifted direction.

  In step S306, as shown in FIG. 10, the update processing unit 17 uses the update predicted position Pr (tk) and the second detection position D2 (tk) corresponding to the correlation detection position D2c (tk). The wake data including the wake position T (tk) is generated, and the process ends. In addition, as a method of generating the wake position T (tk) using the update predicted position Pr (tk) and the second detection position D2 (tk), for example, a method using an α-β filter, a Kalman filter, or the like can be given. It is done.

  As described above, when the correlation data includes a combination of the predicted position P (tk) and the detection position for correlation D2c (tk), the update processing unit 17 generates the second detection position D2c (tk) when the correlation detection position D2c (tk) is generated. An updated predicted position Pr (tk) is generated by correcting the predicted position P (tk) based on the correction of the detection position D2 (tk). In addition, the update processing unit 17 includes the wake position T (tk) using the update predicted position Pr (tk) and the second detection position D2 (tk) corresponding to the correlation detection position D2c (tk). Generate wake data.

  In step S307, the update processing unit 17 uses the prediction position P (tk) included in the prediction data at time tk and the detection position included in the observation data at time tk to obtain wake data including the wake position T (tk). Generate and the process ends. Examples of a method for generating the wake position T (tk) using the predicted position P (tk) and the detected position include a method using an α-β filter, a Kalman filter, and the like.

  In addition, the update process part 17 makes the average value of each motion specification of prediction data and observation data the value of each motion specification of the track data after update. The object type included in the observation data is set as the object type included in the updated track data, and the object size included in the observation data is set as the object size included in the updated track data.

  The series of processes in the flowchart shown in FIG. 8 is performed for each combination for all combinations included in the correlation data input from the correlation processing unit 16.

  The vehicle control device 2 controls the host vehicle based on the track position included in the track data generated by the update processing unit 17. Specifically, the vehicle control device 2 uses the information on the wake position included in the wake data as control of the own vehicle, and reduces the damage when the own vehicle collides with an object ahead. And control of an adaptive cruise control system that follows the vehicle ahead.

  As described above, according to the first embodiment, as the first configuration, when the second observation data is captured in the prediction data at time tk and the first detection position D1 (tk) is detected, the prediction is performed. A predicted position for correlation Pc (tk) is generated by correcting the predicted position P (tk) based on the predicted position P (tk) and the object size included in the data, and a predicted position for correlation Pc (tk) is generated. The detection position D1r (tk) for update is generated by correcting the first detection position D1 (tk) based on the correction of the predicted position P (tk) at the time, and the prediction position P (tk) and the update detection position D1r are generated. (Tk) is used to generate the wake position T (tk).

  Further, as a second configuration, when the second observation data is not captured in the prediction data at time tk and the second detection position D2 (tk) is detected, the second observation data includes the second observation data D2 (tk). The detection position D2c (tk) for correlation is generated by correcting the second detection position D2 (tk) based on the detection position D2 (tk) and the object size, and the correlation detection position D2c (tk) is generated. An updated predicted position Pr (tk) is generated by correcting the predicted position P (tk) based on the correction of the second detected position D2 (tk), and the updated predicted position Pr (tk) and the second detected position D2 (Tk) is used to generate the wake position T (tk). As described above, by realizing the first configuration, the second configuration, or the configuration in which the first configuration and the second configuration are combined, the relative position of the object with respect to the host vehicle can be more accurately compared to the conventional case. Can be acquired.

Embodiment 2. FIG.
In Embodiment 2 of the present invention, an object configured to generate a wake position in consideration of accuracy information of each of the first detection position and the second detection position with respect to the configuration of Embodiment 1 above. The recognition processing device 1 will be described. In the second embodiment, description of points that are the same as those of the first embodiment will be omitted, and points different from the first embodiment will be mainly described.

  The first object detection unit 11 detects first observation data that further includes accuracy information of the first detection position in addition to the first detection position.

  Here, as a method for detecting the accuracy information of the first detection position, for example, the following method can be considered. That is, a test relating to the operation of the first object detection unit 11 is performed in advance, the relationship between the first detection position and its accuracy is evaluated, and a first table in which the first detection position and its accuracy are associated is created. . The first object detection unit 11 calculates the accuracy corresponding to the detected first detection position from the first table created in advance, and detects the calculation result as accuracy information.

  In addition to the second detection position, the second object detection unit 11 detects second observation data that further includes accuracy information of the second detection position.

  Here, as a method for detecting the accuracy information of the second detection position, for example, the following method can be considered. That is, a test relating to the operation of the second object detection unit 12 is performed in advance, the relationship between the second detection position and its accuracy is evaluated, and a second table that associates the second detection position with its accuracy is created. . The second object detection unit 12 calculates the accuracy corresponding to the detected second detection position from the second table created in advance, and detects the calculation result as accuracy information.

  In step S303 of FIG. 8, the update processing unit 17 adds the accuracy information of the first detection position D (tk) and the prediction position P in addition to the prediction position P (tk) and the update detection position D1r (tk). The wake position T (tk) is generated using the accuracy information of (tk).

  Here, as a method for calculating the accuracy information of the predicted position P (tk), for example, the following method can be considered. That is, since the second observation data is captured in the prediction data including the predicted position P (tk), the accuracy corresponding to the predicted position P (tk) is obtained from the above second table with the predicted position P (tk). ) As accuracy information.

  In order to generate the wake position T (tk), for example, the wake position T (tk) may be calculated according to the following equation.

Wake position T (tk)
= A * predicted position P (tk) + (1-a) * updated detection position D1r (tk)
However, a = accuracy of predicted position P (tk) / (accuracy of predicted position P (tk) + accuracy of first detection position D1 (tk))

  In this way, the update processing unit 17 calculates the accuracy information of the predicted position P (tk) corresponding to the correlation predicted position Pc (tk). In addition to the predicted position P (tk) corresponding to the correlation predicted position Pc (tk) and the update detected position D1r (tk), the update processing unit 17 also provides accuracy information of the first detected position D1 (tk). And the accuracy information of the predicted position P (tk) corresponding to the predicted position for correlation Pc (tk), the wake position T (tk) is generated.

  In step S306 of FIG. 8, the update processing unit 17 adds accuracy information of the second detection position D (tk) and the prediction in addition to the update prediction position Pr (tk) and the second detection position D2 (tk). The track position T (tk) is generated using the accuracy information of the position P (tk).

  Here, as a method for calculating the accuracy information of the predicted position P (tk), for example, the following method can be considered. That is, since the second observation data is not captured in the prediction data including the predicted position P (tk), the accuracy corresponding to the predicted position P (tk) is obtained from the first table with the accuracy corresponding to the predicted position P (tk). ) As accuracy information.

  In order to generate the wake position T (tk), for example, the wake position T (tk) may be calculated according to the following equation.

Wake position T (tk)
= A × prediction position for update Pr (tk) + (1−a) × second detection position D2 (tk)
However, a = accuracy of predicted position P (tk) / (accuracy of predicted position P (tk) + accuracy of second detection position D2 (tk))

  In this way, the update processing unit 17 calculates the accuracy information of the predicted position P (tk) corresponding to the updated predicted position Pr (tk). Further, the update processing unit 17 sets the second detection position D2 (tk) in addition to the predicted update position Pr (tk) and the second detection position D2 (tk) corresponding to the correlation detection position D2c (tk). The wake position T (tk) is generated using the accuracy information and the accuracy information of the predicted position P (tk) corresponding to the updated predicted position Pr (tk).

  The detection position used when calculating the wake position may be corrected using the reliability information of the object type or the accuracy information of the detection position.

  The weighting of prediction data and observation data is changed by changing the estimation position estimation error included in the prediction data or the detection position estimation error included in the observation data using the reliability information of the object type. Then, wake data may be generated. By configuring in this way, for example, when the object type is wrong and the reliability information of the object type is output correctly, the value corrected by increasing the estimation error can be corrected. It is possible to reduce the influence of accuracy degradation on the wake data.

  Note that the weighting of the prediction data and the observation data is changed by changing the estimation error of the prediction position included in the prediction data or the detection error of the detection position included in the observation data using the accuracy information of the object dimensions. The wake data may be generated. By configuring in this way, for example, when the error of the object dimension is large and the accuracy information of the object dimension has been output accurately, the corrected value is erroneously corrected by increasing the estimation error. The influence of the decrease in accuracy on can be reduced.

  As described above, according to the second embodiment, the accuracy information of the predicted position P (tk) is calculated with respect to the configuration of the first embodiment, and the predicted position P (tk) and the update detection position D1r ( In addition to tk), the track position T (tk) is generated using the accuracy information of the first detection position D1 (tk) and the accuracy information of the predicted position P (tk).

  Further, with respect to the configuration of the first embodiment, accuracy information of the predicted position P (tk) is calculated, and in addition to the updated predicted position Pr (tk) and the second detection position D2 (tk), The track position T (tk) is generated using the accuracy information of the two detection positions D2 (tk) and the accuracy information of the predicted position P (tk). Even in such a configuration, the same effect as in the first embodiment can be obtained.

  DESCRIPTION OF SYMBOLS 1 Object recognition processing apparatus, 2 Vehicle control apparatus, 11 1st object detection part, 12 2nd object detection part, 13 Object recognition processing part, 14 Prediction processing part, 15 Correction processing part, 16 Correlation processing part, 17 Update processing part .

Claims (12)

A detection wave is transmitted to an object around the host vehicle, the detection wave reflected from the object is received, and observation data including a first detection position is detected as first observation data based on the received detection wave. A first object detection unit;
A second object detection unit for detecting observation data including a second detection position and an object size as second observation data by a detection method different from that of the first object detection unit;
The observation data using the track data including the track position of the time tk-1 before the time tk which is detected, to predict the trajectory data including the track position of the time tk, the predicted position of the time tk A prediction processing unit that generates prediction data including
When the second observation data has already been detected before the time tk-1 and the first observation data at the time tk is detected, the predicted position and the object included in the prediction data at the time tk A correction processing unit that generates correction data including the predicted position for correlation at the time tk by correcting the predicted position based on dimensions;
A correlation processing unit that generates correlation data including a combination of the correlation predicted position at the time tk and the first detection position having a correlation when the correction data at the time tk includes the correlation predicted position;
When the correlation data at the time tk includes a combination of the predicted correlation position and the first detection position, the first detection position is determined based on the correction of the predicted position when the correlation predicted position is generated. The detection position for update at the time tk is generated by correction, and the wake including the wake position at the time tk using the prediction position corresponding to the prediction position for correlation and the detection position for update. An update processing unit for generating data;
An object recognition processing apparatus. Preparing in advance a correction map in which the second detection position, the object size, and the correction amount are associated;
The correction processing unit
A first correction amount corresponding to the predicted position and the object size included in the prediction data is calculated from the correction map, and the predicted position for correlation is generated by shifting the predicted position according to the first correction amount. The object recognition processing apparatus according to claim 1. The update processing unit
The detection for update is performed by shifting the first detection position according to the first correction amount corresponding to the correlation prediction position in a direction opposite to the direction in which the prediction position is shifted when the correlation prediction position is generated. The object recognition processing apparatus according to claim 2, wherein a position is generated. The first observation data further includes accuracy information of the first detection position,
The second observation data further includes accuracy information of the second detection position,
The update processing unit
The accuracy information of the predicted position corresponding to the predicted position for correlation is calculated, and in addition to the predicted position corresponding to the predicted position for correlation and the detection position for update, the accuracy information of the first detected position and the The object recognition processing device according to any one of claims 1 to 3, wherein the wake position is generated using accuracy information of the predicted position corresponding to the correlation predicted position. A detection wave is transmitted to an object around the host vehicle, the detection wave reflected from the object is received, and observation data including a first detection position is detected as first observation data based on the received detection wave. A first object detection unit;
A second object detection unit for detecting observation data including a second detection position and an object size as second observation data by a detection method different from that of the first object detection unit;
The observation data using the track data including the track position of the time tk-1 before the time tk which is detected, to predict the trajectory data including the track position of the time tk, the predicted position of the time tk A prediction processing unit that generates prediction data including
If the second observation data is not yet detected before the time tk−1 and the second observation data at the time tk is detected, the second observation data included in the second observation data at the time tk . A correction processing unit that generates correction data including the detection position for correlation at the time tk by correcting the second detection position based on the detection position and the object size;
If it contains the correlation detection position on the correction data of the time tk, and the correlation processing unit that generates correlation data comprising the combination of the predicted position and the correlation position detection of the time tk a correlation,
When the correlation data at the time tk includes a combination of the predicted position and the detection position for correlation, the predicted position is corrected based on the correction of the second detection position when the detection position for correlation is generated. Thus, the update predicted position at the time tk is generated, and the wake including the track position at the time tk using the predicted update position and the second detection position corresponding to the detection position for correlation. An update processing unit for generating data;
An object recognition processing apparatus. Preparing in advance a correction map in which the second detection position, the object size, and the correction amount are associated;
The correction processing unit
The correlation is obtained by calculating a second correction amount corresponding to the second detection position and the object size included in the second observation data from the correction map, and shifting the second detection position according to the second correction amount. The object recognition processing apparatus according to claim 5, wherein the object detection position is generated. The update processing unit
The update prediction is performed by shifting the predicted position according to the second correction amount corresponding to the correlation detection position in a direction opposite to the direction in which the second detection position is shifted when the correlation detection position is generated. The object recognition processing device according to claim 6 which generates a position. The first observation data further includes accuracy information of the first detection position,
The second observation data further includes accuracy information of the second detection position,
The update processing unit
Accuracy information of the predicted position corresponding to the updated predicted position is calculated, and in addition to the updated predicted position and the second detected position corresponding to the correlation detected position, accuracy information of the second detected position The object recognition processing device according to any one of claims 5 to 7, wherein the wake position is generated using the prediction position accuracy information corresponding to the update predicted position. The prediction processing unit
The predicted position is generated using the wake data including the wake position at the time tk-1 and the relative velocity of the object, and the elapsed time from the time tk-1 to the time tk. The object recognition processing device according to any one of claims 8 to 9. A detection wave is transmitted to an object around the host vehicle, the detection wave reflected from the object is received, and observation data including a first detection position is detected as first observation data based on the received detection wave. A first object detection step;
A second object detection step of detecting observation data including a second detection position and an object size as second observation data by a detection method different from the first object detection step;
The observation data using the track data including the track position of the time tk-1 before the time tk which is detected, to predict the trajectory data including the track position of the time tk, the predicted position of the time tk A prediction processing step for generating prediction data including:
If the second observation data has already been detected before the time tk-1 and the first observation data at the time tk is detected, the second observation data is based on the predicted position and the object size included in the prediction data. A correction processing step of generating correction data including the predicted position for correlation by correcting the predicted position;
A correlation processing step of generating correlation data including a combination of the correlation prediction position at the time tk and the first detection position having a correlation when the correction data at the time tk includes the correlation prediction position;
When the correlation data at the time tk includes a combination of the predicted correlation position and the first detection position, the first detection position is determined based on the correction of the predicted position when the correlation predicted position is generated. The detection position for update at the time tk is generated by correction, and the wake including the wake position at the time tk using the prediction position corresponding to the prediction position for correlation and the detection position for update. Update processing steps to generate data;
An object recognition processing method comprising: A detection wave is transmitted to an object around the host vehicle, the detection wave reflected from the object is received, and observation data including a first detection position is detected as first observation data based on the received detection wave. A first object detection step;
A second object detection step of detecting observation data including a second detection position and an object size as second observation data by a detection method different from the first object detection step;
The observation data using the track data including the track position of the time tk-1 before the time tk which is detected, to predict the trajectory data including the track position of the time tk, the predicted position of the time tk A prediction processing step for generating prediction data including:
If the second observation data is not yet detected before the time tk−1 and the second observation data at the time tk is detected, the second observation data included in the second observation data at the time tk . A correction processing step of generating correction data including the detection position for correlation at the time tk by correcting the second detection position based on the detection position and the object size;
If it contains the correlation detection position on the correction data of the time tk, and correlation processing step of generating correlation data comprising a combination of the predicted position and the correlation position detection of the time tk a correlation,
When the correlation data at the time tk includes a combination of the predicted position and the detection position for correlation, the predicted position is corrected based on the correction of the second detection position when the detection position for correlation is generated. Thus, the update predicted position at the time tk is generated, and the wake including the track position at the time tk using the predicted update position and the second detection position corresponding to the detection position for correlation. Update processing steps to generate data;
An object recognition processing method comprising: The object recognition processing device according to any one of claims 1 to 9,
A vehicle control device that controls the host vehicle based on the wake data generated by the update processing unit of the object recognition processing device;
Vehicle control system equipped with.

Images