JP5024165B2 - Visual object estimation device - Google Patents

Description

  The present invention relates to a visual recognition object estimation device that estimates a visual recognition object viewed by a driver driving a host vehicle.

A technology has been developed in which a visual object visually recognized by a driver is detected from the line-of-sight direction of the driver and a danger is notified by an alarm or the like according to the visual object. As a technique for detecting such a visual recognition object visually recognized by the driver, conventionally, a technique for detecting a target object based on a driver's line-of-sight vector and using it for driving safety is known (for example, Patent Document 1). reference). In this technique, a driver's face and eyeball are detected from an image obtained by capturing the driver's line of sight, a driver's line of sight vector is calculated, and a visual recognition object is specified from the calculated line of sight vector. And the driving assistance according to the object position of the visual recognition object is performed.
JP 2005-284975 A

  However, in the travel safety device disclosed in Patent Document 1, since a detection error occurs when detecting the line-of-sight vector, an error range corresponding to the detection error occurs in the line-of-sight vector. At this time, if there are a plurality of objects within the error range of the line-of-sight vector, it is difficult for the driver to identify which object the driver is viewing. As described above, in the travel safety device disclosed in Patent Document 1, it is difficult to estimate the object that the driver visually recognizes, for example, whether the visually recognized object is an oncoming vehicle or a pedestrian. There has been a problem that it is difficult to provide driving support according to the visual recognition object visually recognized by the driver.

  Therefore, an object of the present invention is to provide a visual object estimation device that can accurately detect a visual object visually recognized by a driver.

The visual object estimation device according to the present invention that has solved the above problems includes a visual line direction detection unit that detects a visual line direction of a driver, an object detection unit that detects an object that exists in the visual line direction, and an object that acquires object information related to the object An information acquisition means; and a visual object estimation means for estimating a driver's visual recognition object from the objects based on object information relating to the object when a plurality of objects are detected . The driver's visual object is estimated based on the degree of danger to the own vehicle, and the estimated collision time of the object existing in the line of sight with respect to the own vehicle is used as the degree of danger to the own vehicle. Detection and acquisition by the object information acquisition means are performed at a plurality of times, and an estimation degree is set for each of a plurality of objects at a plurality of times. Based on the estimated degree set for each of the plurality of objects in each time, and estimates the visual object driver.

  In the visual recognition object estimation device according to the present invention, when a plurality of objects are detected, the visual recognition object of the driver is estimated from the objects based on object information related to the objects. For this reason, even if a plurality of visually recognized objects are detected, it is possible to detect an object visually recognized by the driver in consideration of each detected object. Therefore, it is possible to accurately detect a visually recognized object visually recognized by the driver.

  Usually, the driver of the own vehicle has a high tendency to visually recognize an object having a high degree of danger to the own vehicle. Therefore, the visual recognition object visually recognized by the driver can be detected more accurately by estimating the visual recognition object of the driver based on the degree of danger of the object to the host vehicle.

  Furthermore, it can be set as the aspect which uses the time when the own vehicle reaches | attains the assumption collision position of the object which exists in a gaze direction, and the own vehicle as a danger level with respect to the own vehicle.

  As described above, as the degree of danger to the own vehicle, the estimated collision time of the object existing in the line-of-sight direction with respect to the own vehicle, or until the own vehicle reaches the assumed collision position between the object existing in the line-of-sight direction and the own vehicle. By using the time, it is possible to easily and reliably determine the risk level for the host vehicle.

  In addition, a travel environment information acquisition unit that acquires travel environment information related to the travel environment of the host vehicle can be further provided, and the visual object estimation unit can estimate the driver's visual object from the travel environment and the object attributes.

  Objects in the vicinity of the host vehicle have characteristics such as those that tend to be visually recognized by the driver and those that do not. For this reason, the visual recognition object visually recognized by the driver can be detected with higher accuracy by estimating the visual recognition object of the driver from the driving environment and the attribute of the object.

  Further, the visual object estimation means may be configured to estimate the visual object of the driver based on the object detection history.

  In the traveling vehicle, the surrounding scenery changes every moment, but the driver often keeps track of the visual target that the driver visually recognizes. Therefore, even if time passes, it becomes easy to rise as a candidate for a visual recognition object. For this reason, by estimating the visual recognition object of the driver based on the detection history of the object, the visual recognition object visually recognized by the driver can be detected with higher accuracy.

  And the visual recognition object estimation means can be made into the aspect which estimates the visual recognition object of a driver based on the detection history over a long term compared with the case where the speed of the own vehicle is low when the speed of the own vehicle is high. .

  When the speed of the host vehicle is high, the degree of dependence on the time change when estimating the visually recognized object is higher than when the speed of the host vehicle is low. For this reason, when the speed of the host vehicle is high, the driver's visual object is estimated more accurately by estimating the driver's visual object based on a longer detection history than when the host vehicle speed is low. Can do.

  According to the visual recognition object detection device according to the present invention, it is possible to accurately detect the visual recognition object visually recognized by the driver.

  Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings. In the description of the drawings, the same elements are denoted by the same reference numerals, and redundant description is omitted. For the convenience of illustration, the dimensional ratios in the drawings do not necessarily match those described.

  FIG. 1 is a block configuration diagram of the visual object estimation device according to the first embodiment of the present invention. As shown in FIG. 1, the visual object estimation device according to the present embodiment includes a visual object estimation ECU 1. A driver imaging camera 2, driver gaze camera 3, radar sensor 4, wheel speed sensor 5, and display 7 are connected to the visual object estimation ECU 1. Furthermore, the visual recognition object estimation device includes a driver irradiation device 6.

  The driver imaging camera 2 is provided, for example, at a position in front of the driver's seat in the host vehicle, and captures the face of the driver driving the host vehicle. The driver imaging camera 2 transmits driver image information related to the captured image to the visual object estimation ECU 1.

  The driver's line-of-sight camera 3 is provided, for example, directly above the driver seat on the ceiling of the host vehicle, and photographs the front of the host vehicle. The front position of the host vehicle is a direction that can be the driver's line-of-sight direction. The driver line-of-sight camera 3 transmits line-of-sight image information regarding the captured image to the visual object estimation ECU 1.

  The radar sensor 4 is attached to, for example, the front grill portion of the host vehicle, and detects the direction of the obstacle in front of the host vehicle and the distance between the host vehicle and the obstacle. The obstacles here include fixed objects such as buildings and road signs, and moving objects such as pedestrians and other vehicles. As the radar sensor 4, for example, a millimeter wave radar sensor that transmits and receives millimeter wave information is used. The radar sensor 4 transmits object information related to an object in front of the host vehicle to the visual object estimation ECU 1.

  The wheel speed sensor 5 is provided, for example, on the wheel of the host vehicle, and detects the rotational speed of the wheel. The wheel speed sensor 5 acquires the vehicle speed from the detected rotational speed of the wheel, and transmits the acquired vehicle speed to the visual object estimation ECU 1.

  The driver irradiation device 6 is provided on the side of the driver imaging camera 2. This driver irradiation device 6 irradiates near infrared rays toward the driver's face. By irradiating the driver's face with the driver irradiation device 6, the driver's line of sight can be easily detected at night.

  The visual object estimation ECU 1 includes an image processing unit 11, an object distance acquisition unit 12, a visual object candidate estimation unit 13, a visual object candidate storage unit 14, and a visual object estimation unit 15.

  The image processing unit 11 performs image processing on the driver image information transmitted from the driver imaging camera 2 and detects the line-of-sight direction of the driver. Further, the image processing unit 11 performs image processing on the line-of-sight image information transmitted from the driver line-of-sight camera 3, and detects an object (object) within the visual field of the driver by matching the line-of-sight direction of the detected driver. The image processing unit 11 outputs detected object information regarding the detected object to the visual object candidate estimating unit 13.

  Based on the object information transmitted from the radar sensor 4, the object distance acquisition unit 12 calculates the distance between the object and the host vehicle in front of the host vehicle and the direction of the object with respect to the traveling direction of the host vehicle. The object distance acquisition unit 12 outputs the object state information regarding the calculated object distance and direction to the visual object candidate estimation unit 13.

  Based on the detected object information output from the image processing unit 11 and the object state information output from the object distance acquisition unit 12, the visual recognition object candidate estimation unit 13 calculates the risk level of each object in front of the host vehicle. The procedure for calculating the degree of risk will be described later. Moreover, the visual recognition object candidate estimation part 13 estimates a visual recognition object candidate from each object based on the calculated risk. The visual object candidate estimation unit 13 temporarily stores visual object candidate information related to the estimated visual object candidate in the visual object candidate storage unit 14.

  The visual object estimation unit 15 reads the visual object candidate information once stored in the visual object candidate storage unit 14. The visual object estimation unit 15 estimates the visual object of the driver based on the read visual object candidate information and the vehicle speed information transmitted from the wheel speed sensor 5. The visual object estimation procedure will be described later. The visual recognition object estimation unit 15 generates driving support information using the estimated visual object after estimating the visual object of the driver. The visual recognition object estimation unit 15 transmits the generated driving support information to the display 7.

  The display 7 displays an image based on the driving support information transmitted from the visual object estimation unit 15. Here, for example, with respect to an image captured by the driver's line-of-sight camera 3, information on an image in which the driver selects an object to be watched out of objects other than the visible object and highlights the selected object is the driving support information. Become.

  Next, a processing procedure in the visual object estimation ECU 1 will be described. FIG. 2 is a flowchart showing a processing procedure in the visual object estimation ECU. As shown in FIG. 2, the visual object estimation ECU 1 first acquires foreground information (S1). The foreground information includes driver image information relating to an image taken by the driver imaging camera 2, line-of-sight image information relating to an image taken by the driver line-of-sight camera 3, the own vehicle detected by the radar sensor 4, the object ahead, and the direction and direction. Contains object information.

  After obtaining the foreground information, the image processing unit 11 detects the driver's line of sight based on the driver image information transmitted from the driver imaging camera 2 (S2). Subsequently, in the image processing unit 11, the detected driver line of sight is superimposed on the line-of-sight image obtained from the line-of-sight image information transmitted from the driver line-of-sight camera 3 (S3).

  Then, the line-of-sight candidate area is specified by performing image processing on the line-of-sight image on which the driver's line of sight is superimposed (S4). Subsequently, all objects included in the line-of-sight candidate area are detected (S5), and the object detection information is output to the visual object candidate estimation unit 13. The object distance acquisition unit 12 calculates the distance and direction between the host vehicle and each object (S6), and outputs object state information regarding the calculated distance and direction to the visual object candidate estimation unit 13.

  The visual object candidate estimation unit 13 calculates the risk level for each object based on the object detection information output from the image processing unit 11 and the object state information output from the object distance detection unit (S7). As the risk calculation procedure, for example, the following procedure can be used.

  The simplest risk level calculation procedure is to determine that an object having a shorter distance to the host vehicle has a higher risk level.

  Another risk level calculation procedure is to determine that the risk level is higher as the collision estimated time TTC (Time to Collision) is shorter. The collision estimation time TTC here is a collision estimation time calculated as a time until the vehicle is expected to collide with the object based on the relative distance and relative speed between the vehicle and the object. A driver driving the host vehicle tends to turn his eyes to an object having a short collision estimated time TTC. In the example using the collision estimation time TTC, the visual recognition object is estimated using the degree of risk obtained by using the tendency of the driver.

  For example, as shown in FIG. 3, it is assumed that a building B1, a sign B2, and a pedestrian B3 exist as objects included in the driver's line-of-sight candidate area. In addition, the points P1 to P4 in the figure indicate the points at 12 o'clock, 3 o'clock, 6 o'clock, and 9 o'clock of the circular line-of-sight candidate areas, respectively, and the point P5 indicates the center of the line-of-sight candidate area.

  The relative distance to the host vehicle is longer in the order of the sign B2, the building B1, and the pedestrian B3. Moreover, the relative speed with respect to the own vehicle is the shortest with respect to the pedestrian B3, and is the same relative speed with respect to the sign B2 and the building B1. When the estimated collision time TTC for each object is calculated under such conditions, the estimated collision time TTC for the building B1 is 20 seconds, the estimated collision time TTC for the sign is 3 seconds, and the estimated collision time TTC for the pedestrian B3 is 2 seconds. Suppose. In this case, the sign B2 is the shortest relative distance to the host vehicle, but the estimated collision time TTC is the smallest for pedestrians. Therefore, in the example shown in FIG. 3, it determines with the risk degree of pedestrian B3 being the highest.

Further, as another risk level calculation procedure, an assumed collision position CP (Conflict point) between the own vehicle and each object is obtained, and the shorter the arrival time until the own vehicle is assumed to reach the assumed collision position CP, It is determined as an object with a high degree of danger. The assumed collision position CP refers to a collision position where the host vehicle and each object are expected to collide when the host vehicle travels. Calculation of the assumed collision position CP for each object can be performed using the following equations (1) to (3).

  When the object is a fixed object, TCP2 = 0. In addition, the constant α can be a time obtained by adding a certain margin to a time when it is assumed that it is difficult to stop the host vehicle by a driver's operation. Furthermore, the constant β can be a time obtained by adding a certain margin to the time when it is assumed that it is difficult to avoid a collision between the host vehicle and the object due to the stopping ability between the host vehicle and the object. An estimated collision position CP for each object calculated in this way is calculated, and an object having a closer estimated collision position CP is determined to have a higher degree of danger.

  After the risk level is calculated in this way, a visual object candidate is estimated (S8). Here, all the detected objects are targeted as the visual recognition object candidates. If the visual object candidate is estimated, the candidate degree of each visual object candidate is estimated (S9). The candidate degree of the visual object candidate is estimated corresponding to the risk level, and the object degree of the visual object candidate is estimated to be higher as the object has a higher risk level.

  When the candidate degree of the visual object candidate is estimated, the estimated visual object candidate and the candidate degree for each visual object candidate are associated with each other and stored in the visual object candidate storage unit 14. As shown in FIG. 4, in the visual object candidate storage unit 14, in addition to the currently estimated visual object candidate and the candidate degree of the visual object candidate, the visual object candidate estimated in the past and the candidate degree of the visual object candidate are stored. Is stored as a detection history.

  Then, the visual object estimation unit 15 estimates the visual object candidate (S10). The visual object candidate is estimated as follows. First, the visual object estimation unit 15 reads the visual object candidates stored in the visual object candidate storage unit 14 and the candidate degrees of the visual object candidates for a plurality of times. Subsequently, as shown in FIG. 5, the visual object candidates and the candidate degrees of the visual object candidates at each time are arranged along the time axis. In FIG. 5, the past is as the time axis moves to the right. Then, based on the vehicle speed transmitted from the wheel speed sensor 5, the candidate degree of the visual object candidate at each time is weighted. As a weighting method, the time component of the visual object candidate is weighted more heavily as the vehicle speed is faster.

  For example, as shown in FIG. 6A, an object included in the line-of-sight candidate area that was seen at time t0 appears to move as shown in FIG. 6B at time t1. Thus, the object in the line-of-sight candidate area appears to move. The visual object candidates and the candidate object candidate degrees at each time are arranged in time series as shown in FIG. In addition, there are a visual object candidate that deviates from the line of sight and a visual object candidate that enters the line of sight as the line of sight moves. For example, at time t0, the objects A to E are visual object candidates that enter the line of sight, but at time t1, the object D disappears from the line of sight, and the object F newly enters the line of sight.

  Then, the estimated degree of each object is calculated according to the speed of the host vehicle. For example, when the speed of the host vehicle is higher than a predetermined threshold, the estimated degree of the visual object candidate is calculated using the candidate degrees of the three visual object candidates at time t0 to time t2 as the object estimation degree. . Further, when the speed of the host vehicle is lower than a predetermined threshold value, the estimated degree of the visual object candidate is calculated using the candidate degrees of the two visual object candidates at time t0 and time t1 as the estimated degree of the object. .

  The estimated degree of each object is obtained by adding the candidate degrees of each object. Here, when the visual recognition object is estimated using the candidate degrees of the three visual object candidates, the candidate degrees of the objects A to F are 2.1, 2.2, 1.7, 0.6, and 0, respectively. .5, 0.3. In this case, the object B is estimated as a visually recognized object. As described above, when the vehicle speed of the host vehicle is high, the visually recognized object is estimated by adding the time component to the determination condition.

  Further, when the visual recognition object is estimated using the candidate degrees of the two visual object candidates, the candidate degrees of the objects A to F are 1.6, 1.5, 1.0, 0.3, 0,. 5,0.1. In this case, the object A is estimated as a visually recognized object. As described above, when the vehicle speed of the host vehicle is high, the visually recognized object is estimated by reducing the time component added to the determination condition.

  When the host vehicle travels, the surrounding scenery changes from moment to moment, but the driver often keeps track of the visual target to be visually recognized by the driver. Therefore, even if time passes, it becomes easy to rise as a candidate for a visual recognition object. For this reason, by estimating the visual recognition object of the driver based on the detection history of the object, the visual recognition object visually recognized by the driver can be detected with higher accuracy.

  When the visual recognition object is estimated in this way, driving assistance corresponding to the estimated visual recognition object is provided (S11). As driving assistance here, while highlighting an object other than a visual recognition object, a line-of-sight image is displayed on the display 7 or an object other than the visual recognition object is determined and a gaze target object to be watched by the driver is determined. For example, the gaze target object can be highlighted.

  As described above, in the visual recognition object estimation device according to the present embodiment, when a plurality of objects are detected in the driver's visual line destination candidate area, the visual recognition object of the driver is selected from the objects based on the risk level of each object. Estimated. Further, the driver of the own vehicle is generally more likely to visually recognize an object having a high degree of danger to the own vehicle. For this reason, even if a plurality of objects are detected, a visually recognized object visually recognized by the driver can be accurately detected.

  Further, the risk level of each object is obtained using the estimated collision time and the arrival time to the assumed collision position. For this reason, the risk level of each object can be obtained with high accuracy.

Further, the risk level of the past object is stored in the visual object candidate storage unit 14 as a detection history, and the visual object is estimated based on this detection history. For this reason, it is possible to detect a visual object visually recognized by the driver with higher accuracy. Moreover, in estimating the visual recognition object, the visual recognition object of the driver is estimated based on the detection history over a long period as the vehicle speed increases. When the vehicle speed is high, the degree of dependence on the time change when estimating the visually recognized object is higher than when the speed of the host vehicle is low. Therefore, when the speed of the host vehicle is high, the driver's visual object can be estimated more accurately by estimating the driver's visual object based on the detection history over a longer period than when the host vehicle speed is low. it can.

  Next, a second embodiment of the present invention will be described. FIG. 7 is a block configuration diagram of the visual object estimation device according to the present embodiment. As shown in FIG. 7, the visual object estimation device according to the present embodiment is different from the visual object estimation device according to the first embodiment in that the visual object estimation ECU 20 includes a traveling environment acquisition unit 21 and an object attribute storage. The difference is that the portion 22 is provided. The visual object estimation ECU 1 is different in that a road-vehicle communication unit 8 is connected. Other points are common to the first embodiment.

  As shown in FIG. 6, the road-to-vehicle communication unit 8 connected to the visual object estimation ECU 20 according to the present embodiment is capable of communicating with communication equipment installed on the road, and is a traveling environment transmitted from the communication equipment. Receive information. The road-to-vehicle communication unit 8 transmits the received traveling environment information to the visual object estimation ECU 20.

  The traveling environment acquisition unit 21 acquires the traveling environment around the host vehicle based on the traveling environment information transmitted from the road-to-vehicle communication unit 8. The traveling environment acquisition unit 21 outputs acquired traveling environment information based on the acquired traveling environment to the visual object candidate estimation unit 13. The traveling environment here includes, for example, whether the road on which the vehicle is traveling is a single road scene or a temporary stop scene.

  The object attribute storage unit 22 stores, for each object, an object attribute that can be a visual recognition target object around the host vehicle. Various object attributes stored in the object attribute storage unit 22 can be provided. The object attribute storage unit 22 outputs the stored object attribute to the visual object candidate estimation unit 13 based on the signal from the visual object candidate estimation unit 13.

  Examples of the object attributes stored in the object attribute storage unit 22 include fixed objects such as buildings, sky, and road signs as well as moving objects such as pedestrians as shown in FIG. For each of the above, the object attributes of the single road scene and the pause scene are stored in points. The larger the object attribute stored here, the higher the degree of becoming a visually recognized object. When the object attribute is a moving object such as a pedestrian, the object and attribute can be subdivided according to the moving direction. Furthermore, in the case of a pedestrian or the like, the object attributes can be subdivided according to whether the pedestrian is an adult or a child, and the orientation of the pedestrian's face. As a tendency in this case, when the object is a person, the object attribute is set lower than the direction facing the own vehicle, and the object attribute is set lower for the adult than the child. Further, when the object attribute is a vehicle, the object attribute of the vehicle having the same traveling direction as that of the host vehicle is set lower than that of the oncoming vehicle.

  Next, a processing procedure in the visual object estimation device according to the present embodiment will be described. The processing in the visual object estimation device according to the present embodiment is mainly different from the first embodiment in the procedure for estimating the visual object candidate, so that the following description will focus on the visual object candidate estimation procedure. .

  In the visual object estimation ECU 20, the visual object candidate estimation unit 13 estimates the visual object candidate based on the travel environment information transmitted from the road-to-vehicle communication unit 8. Acquires whether it is a road scene or a pause scene. Next, based on the object detection information output from the image processing unit 11, surrounding objects in the host vehicle are acquired. Subsequently, the object attribute corresponding to the acquired object is acquired from the object attribute storage unit 22.

  Then, the visual recognition object candidate estimation unit 13 estimates the acquired object and the object attribute for the object as the visual recognition object candidate candidate degree. Thereafter, the candidate degree of the visual object candidate estimated in the visual object candidate storage unit 14 is stored, and the visual object is estimated by the same procedure as in the first embodiment.

  Thus, in this embodiment, the object attribute of each object is used in estimating the candidate degree of the visual object candidate. By using such object attributes, it is possible to accurately estimate the candidate degree of the visual object candidate.

The preferred embodiment of the present invention has been described above, but the present invention is not limited to the above embodiment. For example, in each of the above embodiments, when estimating the visual recognition object, the visual recognition object candidate is first estimated and then the visual recognition object is estimated based on the detection history. A visual object candidate having the highest candidate degree of visual object candidate can be estimated as a visual object as it is.
Moreover, in the said embodiment, when estimating a visual recognition object based on a detection log | history, the detection log | history of 2 times or 3 times is used, but it can also set to other frequency | count. Furthermore, in the said 2nd Embodiment, although the road on which the own vehicle drive | works is used as a driving | running environment, other elements, such as a weather and day and night, can also be used. Further, in the first embodiment, when the risk level is calculated, the estimated collision time TTC or the time until the host vehicle reaches the assumed collision position CP is used. It can also be set as an aspect.

It is a block block diagram of the visual recognition object estimation apparatus which concerns on 1st Embodiment. It is a flowchart which shows the process sequence in visual recognition object estimation ECU. It is a schematic diagram which shows the example of a gaze image. It is a table | surface which shows the example of the candidate degree of a visual recognition object candidate. It is the figure which put in order the candidate degree of a plurality of visual recognition object candidates along time series. (A) is a schematic diagram which shows the example of the gaze image in the time t0, (b) is a schematic diagram which shows the example of the gaze image in the time t1. It is a block block diagram of the visual recognition object estimation apparatus which concerns on 2nd Embodiment. It is a table | surface which shows the example of the relationship between an object and an object characteristic.

Explanation of symbols

  DESCRIPTION OF SYMBOLS 1,20 ... Visual object estimation ECU, 2 ... Driver imaging camera, 3 ... Driver gaze camera, 4 ... Radar sensor, 5 ... Wheel speed sensor, 6 ... Driver irradiation apparatus, 7 ... Display, 8 ... Road-to-vehicle communication part, 11 DESCRIPTION OF SYMBOLS ... Image processing part, 12 ... Object distance acquisition part, 13 ... Visual object candidate estimation part, 14 ... Visual object candidate storage part, 15 ... Visual object estimation part, 21 ... Driving environment acquisition part, 22 ... Object attribute storage part.

Claims (4)

Gaze direction detection means for detecting the gaze direction of the driver;
Object detection means for detecting an object present in the line-of-sight direction;
Object information acquisition means for acquiring object information relating to the object;
A visual object estimation means for estimating a visual object of the driver from the object based on object information related to the object when a plurality of the objects are detected;
Equipped with a,
The visual recognition object estimation means estimates the visual recognition object of the driver based on the degree of danger of the object to the host vehicle,
As the risk level for the host vehicle, the estimated collision time of the object existing in the line-of-sight direction with respect to the host vehicle is used.
The detection by the gaze direction detection means and the object detection means and the acquisition by the object information acquisition means are performed at a plurality of times,
For each of the plurality of objects at the plurality of times, an estimation degree is set.
A visual object estimation device that estimates the visual object of the driver based on an estimation degree set for each of the plurality of objects at each of the plurality of times . The visual object estimation device according to claim 1 , wherein a time for the host vehicle to reach an assumed collision position between an object existing in the line-of-sight direction and the host vehicle is used as the risk level for the host vehicle. The vehicle further comprises traveling environment information acquisition means for acquiring traveling environment information related to the traveling environment of the host vehicle,
The visual object estimation device according to claim 1, wherein the visual object estimation unit estimates the visual object of the driver from the driving environment and the attribute of the object. The visual recognition object estimation device according to any one of claims 1 to 3, wherein the visual recognition object estimation unit sets the estimation degree larger as a vehicle speed at each time is higher.