JP6187155B2 - Gaze target estimation device - Google Patents

Description

  The present invention relates to a gaze target estimation device that estimates a gaze target that a driver of a vehicle is gazing at.

  For example, Patent Document 1 discloses a travel information providing device that provides necessary information using a driver's line of sight. This travel information providing apparatus measures the driver's line-of-sight direction using an in-vehicle camera or the like. Further, the position of traffic environment objects such as other vehicles, pedestrians, two-wheeled vehicles, signs, displays, signals, pedestrian crossings, and temporary stop lines existing around the host vehicle is detected. Based on the measured line-of-sight direction and the detected position of the traffic environment object, the visibility of the driver with respect to the traffic environment object is calculated. Since the calculated traffic environment object with low visibility is highly likely not to be recognized by the driver, the driver is alerted by highlighting it on the screen, for example.

  However, when measuring the driver's gaze direction using camera images etc., the gaze direction measurement value has the cumulative effect of errors due to various factors such as pupil center estimation error and face orientation estimation error. receive. Also, when the driver's posture changes, the relative positional relationship between the line-of-sight direction and the object changes.

  In view of such points, Patent Document 2 discloses an apparatus that performs automatic calibration so that the estimated line-of-sight direction matches the actual line-of-sight direction. In the device of Patent Document 2, when a specific operation during driving (for example, an operation for changing lanes) is performed, it is considered that attention to a specific position (a rearview mirror or a side mirror for rearward confirmation) increases. The line-of-sight data from a predetermined time before the start time of the specific operation to the end time is extracted. Then, a gaze position at a specific position is calculated based on the distribution of the extracted gaze data, and a correction coefficient for correcting the gaze data so that the gaze position matches a gaze position predetermined for a specific operation Is calculated.

Japanese Patent No. 5171629 JP 2010-29262 A

  However, even when the driver performs a specific operation, the driver does not pay attention to only a specific position corresponding to the specific operation. For example, when changing lanes, as a preparation, the driver checks the situation behind the host vehicle with a rearview mirror or side mirror within a short period of time while paying attention to the preceding preceding vehicle and road conditions. Then, immediately before changing the lane, it is visually confirmed whether the lane can be changed safely by looking back at the rear side of the lane changing side. Such a driver's confirmation operation is a general one, and the confirmation operation, that is, the movement of the line of sight often differs depending on the driver, depending on the road condition or the driver's habit.

  Therefore, even if gaze data from a predetermined time before the start time of a specific operation to the end time is extracted as in the device of Patent Document 2, the extracted gaze data has a specific position (room mirror or side mirror). In addition to the line-of-sight data that is gazing, the line-of-sight data in various directions outside the passenger compartment is included. Therefore, it is difficult to determine the specific position corresponding to the specific operation with high accuracy from the extracted line-of-sight data.

  The present invention has been made in view of the above-described points, and provides a gaze target estimation device capable of estimating a gaze target that a driver of a vehicle is gaze with higher accuracy. Objective.

In order to achieve the above-described object, a gaze target estimation device according to the present invention includes:
Detecting means (10) for detecting a head position including a gaze direction of each of the left and right eyes of the driver of the vehicle and a face direction of the driver;
Based on the gaze direction and the head position including the face direction, gaze point data indicating the position of the gaze point being watched by the driver is calculated, and the gaze point data is calculated from the driver. A calculation means (14) including a distance to
Extraction means (16, 18) for extracting a plurality of gazing point data having a distance equal to or less than a predetermined threshold value from gazing point data repeatedly calculated by the calculation means;
Gazing point data regarded as gazing at least one specific gazing target attached to the vehicle from a plurality of gazing point data distributions , targeting a plurality of gazing point data extracted by the extracting means ; and Gazing point data corresponding to a specific gaze object based on gazing point data classified as gaze point data considered to be gazing at a specific gaze target object Range setting means (20) for determining the range of the position;
When the position indicated by the gazing point data calculated by the calculation means belongs to the range determined by the range setting means, the estimation that the object being watched by the driver of the vehicle is a specific gazing target And means (22).

  As described above, the gaze target estimation device according to the present invention obtains gaze point data indicating the position of the gaze point that the driver of the vehicle is gazes at. This gazing point data includes the distance from the driver to the gazing point. The gaze point object estimation device determines a range of positions indicated by gaze point data corresponding to a specific gaze object based on a plurality of gaze point data. When the newly calculated gazing point data belongs to a predetermined range, it is estimated that the driver of the vehicle is gazing at a specific gazing target.

  Here, the gaze target estimation device according to the present invention uses a vehicle accessory as a specific gaze target, and as gaze point data for defining a range corresponding to the specific gaze target, a predetermined threshold value or less. We narrow down to gaze point data with distance. In this way, by filtering the gaze point data by distance, it is possible to reliably extract the gaze point data that is gazing at a specific gaze target object that is an accessory of the vehicle, and the preceding vehicle, pedestrian, It is possible to exclude gaze point data such as road conditions when the driver is gazing outside the vehicle. Accordingly, it is possible to exclude as much as possible gaze point data that becomes noise in determining a range corresponding to a specific gaze target. For this reason, the range corresponding to a specific gaze object can be determined with higher accuracy than before.

  Note that the reference numerals in the parentheses merely show an example of a correspondence relationship with a specific configuration in an embodiment described later in order to facilitate understanding of the present invention, and limit the scope of the present invention. It is not intended.

  Further, the technical features described in the claims of the claims other than the features described above will become apparent from the description of embodiments and the accompanying drawings described later.

It is the block diagram which showed the structure of the whole system containing the gaze target object estimation apparatus which concerns on embodiment. It is a flowchart which shows the process performed in a gaze target object estimation apparatus. It is a figure which shows the result of having plotted the gaze point data accumulate | stored during the fixed period on the graph which makes the angle of a horizontal direction and the angle of a perpendicular direction 2 axes | shafts. It is a figure which shows the result of having extracted the gaze point data which has the distance below a threshold value from the gaze point data represented by FIG. 3, and plotted the position which the extracted gaze point data showed on the graph. It is a figure which shows the result of having classified the gaze point data corresponding to a specific gaze object, and other gaze point data.

  Hereinafter, a gaze target estimating device according to an embodiment of the present invention will be described with reference to the drawings. FIG. 1 is a block diagram illustrating a configuration of the entire system including the gaze target estimation device 12 according to the present embodiment.

  In FIG. 1, the gaze direction and head position detection device 10 separately detects the gaze directions of the left and right eyes of the driver of the vehicle, and detects the head position including the direction of the driver's face. Information indicating the result is output to the gaze target estimation device 12.

  With regard to detection of the gaze direction, for example, a camera that captures the left and right eyes of the driver is provided. The direction can be detected. Alternatively, the line-of-sight direction may be detected by applying a so-called corneal irradiation method. In this case, the driver is irradiated with infrared rays using an infrared LED or the like, and the left and right eyes of the driver are photographed with a camera capable of photographing infrared light. Then, in the photographed image, the position of the reflected light on the cornea (Purkinje image) is used as a reference point, and the line-of-sight directions of the left and right eyes are detected from the positional relationship of the pupil with respect to the reference point. In addition, other line-of-sight detection technologies such as the sclera tracker method using the boundary between the cornea and the sclera, the search coil method using a contact lens incorporating a coil, and the EOG method using the cornea-retinal potential are applied. Thus, the line-of-sight directions of the left and right eyes may be detected.

  When a camera is used to detect the line of sight, the camera may be a stationary type installed in the passenger compartment, or a wearing type attached to a camera or hat worn by the driver.

  Regarding the detection of the head position, a camera for photographing the driver's face is provided, and the head position including the direction of the driver's face is detected by recognizing the driver's face from the captured image. The head position detection camera may be a stationary type or a wearable type as in the case of detection of the line-of-sight direction. Further, the head position detection camera may be shared with the line-of-sight detection camera, or may be a separate one.

  Moreover, you may detect a driver | operator's head position using means other than a camera. For example, a gyro sensor or an acceleration sensor is incorporated into an attachment such as a hat or glasses worn by the driver on the head, and the orientation of the driver's face is detected by detecting rotation or translation of the driver's head. The head position including the head position may be detected.

  Information indicating the gaze direction and head position detected by the gaze direction and head position detection device 10 is input to the gaze target estimation device 12. The gaze target estimation device 12 is composed of a general microcomputer including a CPU, a ROM, a RAM, an I / O, and the like. In FIG. 1, functions that are realized in the gaze target estimation device 12 when the CPU executes a program stored in the ROM are shown by functional blocks. Hereinafter, each functional block will be described.

  The gazing point data calculation unit 14 generates gazing point data indicating the position of the gazing point that the driver is gazing based on the information indicating the gazing direction and the head position given from the sight line direction and the head position detection device 10. calculate. In this embodiment, the point of gaze is a point where the line of sight of the left and right eyes detected independently intersect. In calculating the gaze point data, first, the gaze point position is obtained in the driver-based coordinate system with the origin between the eyebrows of the driver (between both eyes) using the gaze directions of the left and right eyes.

  And the standard head position when the driver is seated in the driver's seat facing the front using the head position including the direction of the driver's face as the gaze point position in the driver reference coordinate system. It is converted into the point of gaze position in the vehicle reference coordinate system that is the origin. Specifically, the gaze point position in the driver reference coordinate system is rotated in the direction of the driver's face, the Yaw direction, the Roll direction, and the Pitch direction, and the deviation of the driver's head position from the origin, Translate horizontally, vertically, and back and forth. Thereby, the gazing point position in the vehicle-based coordinate system is obtained.

  In this embodiment, the gazing point position is represented by the angle of the gazing point in the horizontal direction (left and right direction), the angle of the gazing point in the vertical direction (up and down direction), and the distance to the gazing point. Therefore, the gazing point data calculated by the gazing point data calculation unit 14 includes a horizontal angle, a vertical angle, and a distance to the gazing point.

  Incidentally, if the detected left and right eye sight do not intersect, then enter the middle direction of the left and right viewing direction or the gaze direction, the advance of the information which is efficacious to the driver watch object estimating apparatus 12 In addition, the gaze direction of the effect may be the gaze direction. In such a case, the distance to the gazing point is a predetermined fixed distance.

  The calculation of the gazing point data in the gazing point data calculation unit 14 is repeatedly performed at predetermined intervals. The gaze point data calculation unit 14 outputs the calculated gaze point data to the gaze point data storage unit 16 and the gaze target estimation unit 22 each time gaze point data is calculated.

  The gazing point data accumulation unit 16 accumulates gazing point data repeatedly output from the gazing point data calculation unit 14. Using the gazing point data stored in the gazing point data storage unit 16, the gazing target range setting unit 20 described later determines the range of the position indicated by the gazing point data corresponding to the specific gazing target. The gazing point data used for setting the range corresponding to the specific gazing target is deleted from the gazing point data storage unit 16. However, the gazing point data may be left as it is without being erased. In this case, when the upper limit of the storage capacity of the gazing point data storage unit 16 is reached, the oldest gazing point data may be overwritten with the latest gazing point data.

  The gazing point data extraction unit 18 extracts gazing point data having a distance equal to or less than a predetermined threshold from the gazing point data accumulated in the gazing point data accumulation unit 16. The gaze point data extraction unit 18 gives the extracted gaze point data to the gaze target range setting unit 20.

  Here, the gazing point data calculated by the gazing point data calculation unit 14 is cumulatively affected by errors of various factors as described above. Furthermore, when the driver's head position or posture changes, the magnitude of the error changes, or the relative positional relationship between the line-of-sight direction and the gaze target changes. Therefore, it is necessary to perform a calibration process for matching the position of the gazing point indicated by the gazing point data with the position of the actual gaze target.

  In order to perform the calibration process, at least one of the left side mirror (left door mirror), the right side mirror (right door mirror), the room mirror (back mirror), the meter, the navigation device, and the head-up display screen is used in the present embodiment. One vehicle accessory is defined as a specific gaze target. Then, in the gaze object range setting unit 20, as described above, the gaze point data for determining the range corresponding to the specific gaze object is equal to or less than the predetermined threshold value extracted by the gaze point data extraction unit 18, as described above. Gaze point data having a distance of. In this way, by applying the filtering by the distance to the gazing point data, it is possible to reliably extract the gazing point data that is gazing at a specific gaze target that is an accessory of the vehicle. On the other hand, it is possible to exclude the gazing point data when the driver is gazing outside the vehicle, such as the preceding vehicle, pedestrians, and road conditions, from the gazing point data for determining the range corresponding to the specific gazing object. It becomes possible. For this reason, in determining the range corresponding to a specific gaze object, it is possible to exclude gaze point data that becomes noise as much as possible. As a result, a range corresponding to a specific gaze object can be determined with higher accuracy than before.

  The threshold used by the gazing point data extraction unit 18 to extract the gazing point data may be set to a constant value based on the vehicle body structure of the vehicle to which the gazing target estimation device 12 according to the present embodiment is applied. Thus, by setting the threshold value based on the vehicle body structure, it is possible to consider the positional relationship between the driver and a specific gaze target that is an accessory of the vehicle. More specifically, it is possible to determine a threshold value capable of distinguishing gazing point data gazing at a specific gazing target and gazing point data gazing at any object outside the vehicle.

  Alternatively, the threshold value may be set and updated according to the driver's seating position and driving posture. For example, a seating detection unit that detects seating of the vehicle on the driver's seat such as a pressure sensor is provided. Then, when seating of the vehicle in the driver's seat is detected, a threshold may be set based on the seat position of the driver's seat (front-rear direction position, vertical position, backrest angle, etc.). In this way, by taking into account the driver's seat position, it is possible to identify the gazing point data when gazing at a specific gazing object and the gazing point data when gazing outside the passenger compartment. As the threshold value, a more appropriate threshold value can be set.

  Further, when a threshold value is set according to the seat position when sitting in the driver's seat, it is desirable to update the set threshold value when the seat position is changed. For example, when driving for a long time, the driver may change the angle of the backrest or adjust the position in the front-rear direction in order to reduce fatigue. In such a case, the threshold value is reset (updated) based on the adjusted sheet position.

  Further, the threshold value may be set and updated in consideration of the driving posture of the driver in addition to the seat position of the driver seat. This is because even if the seat position is the same, the positional relationship between the driver and a specific gaze object may differ depending on the seating position on the seat and the posture of the upper body. In this case, for example, the driver's driving posture is detected by using a load sensor that detects a load distribution applied to the seat surface and the backrest of the driver's seat, a camera that shoots the driver, and the like. Then, when seating in the driver's seat is detected, a threshold is set in consideration of the driver's driving posture in addition to the seat position. Thereafter, even if the seat position is not changed, when the driving posture of the driver changes, the threshold value is updated based on the driving posture of the driver after the change. In this way, it is possible to set a more appropriate threshold value by considering the driving posture of the driver.

  When a plurality of specific gaze objects are determined, the threshold value may be set individually for each of the plurality of gaze objects. This is because the distance from the driver to the gaze object is different for each gaze object. In this case, the gazing point data extraction unit 18 divides the coordinate system of the gazing point data into a plurality of areas each including a plurality of gazing objects, and uses a threshold corresponding to the corresponding gazing object for each area. What is necessary is just to extract gaze point data.

  The gaze target range setting unit 20 sets a plurality of gaze point data provided by the gaze point data extraction unit 18 when it is time to set the position range indicated by the gaze point data corresponding to the specific gaze target. Based on the above, the range of the position indicated by the gaze point data corresponding to the specific gaze object is determined.

  The timing at which the range of the position indicated by the gaze point data corresponding to the specific gaze target is to be set is preferably the timing when a predetermined time has elapsed. In this way, by periodically resetting (updating) the range of the position indicated by the gaze point data corresponding to the specific gaze target, even if the driver's head position changes, the gaze point The position indicated by the data can be matched with the position where the driver actually watches.

  Further, the time when a predetermined condition is satisfied may be set as a timing at which the range of the position indicated by the gaze point data corresponding to the specific gaze object should be set. The predetermined conditions include, for example, a driving operation for a right turn, a driving operation for a left turn, a driving operation for a reverse drive, a driving operation for changing lanes, an operation of a navigation device, etc. Can be adopted. When these operations are performed, the driver has a high probability of confirming a specific target object such as left and right side mirrors, a room mirror, and a navigation device. Therefore, the gazing point data storage unit 16 accumulates gazing point data including gazing point data when a specific gaze target is gazed. Accordingly, the gaze target range setting unit 20 sets or updates a range corresponding to a specific gaze target based on gaze point data including gaze point data when the specific gaze target is gazeed. be able to.

  Alternatively, as the predetermined condition, for example, it may be adopted that the vehicle is in a predetermined traveling state such as deceleration, turning, or stopping. Thus, even when the vehicle is in a predetermined traveling state, the driver of the vehicle often confirms a specific gaze object that is an accessory of the vehicle.

  Furthermore, as the predetermined condition, regarding the gazing point data stored in the gazing point data storage unit 16, it may be adopted that the number of data included in the vicinity range of the specific gazing target is equal to or greater than the predetermined number. As described above, when the number of data of the gazing point data included in the vicinity range of the gazing target is equal to or larger than the predetermined number, the gazing point data includes a large amount of gazing point data when gazing at the specific gazing object. It is because it can be estimated that it is included.

  Then, the gaze target range setting unit 20 sets the position indicated by the horizontal angle and the vertical angle in the gaze point data as the position indicated by the gaze point data. In other words, the position of the gazing point data can be expressed on a plane having two angles of a horizontal angle and a vertical angle. Then, the gaze point data corresponding to the specific gaze object and the other gaze point data are classified from the distribution of the positions of the plurality of gaze point data. The gaze target range setting unit 20 determines a range corresponding to a specific gaze target so as to cover the positions of a plurality of gaze point data classified as gaze point data corresponding to the specific gaze target.

  The gaze point data given to the gaze target range setting unit 20 is narrowed down to gaze point data having a distance equal to or smaller than a predetermined threshold by the gaze point data extraction unit 18. Therefore, regarding the distance, it is considered that the influence on classifying (selecting) gaze point data corresponding to a specific gaze object and other gaze point data is small. Therefore, in the present embodiment, as described above, the position indicated by the horizontal angle and the vertical angle in the gazing point data is the position indicated by the gazing point data. However, when determining the position of each gazing point data, the distance of the gazing point data may be used.

  A method for classifying gaze point data corresponding to a specific gaze object and other gaze point data from the distribution of the positions of a plurality of gaze point data will be briefly described. For example, it is assumed that the position indicated by the gaze point data corresponding to the gaze target indicates a mixed Gaussian distribution approximate to the Gaussian distribution. Then, clustering is performed using the EM algorithm so that the position indicated by the gazing point data given from the gazing point data extraction unit 18 is divided into a class indicating a mixed Gaussian distribution and a class of other outliers.

  Note that the position indicated by the gaze point data corresponding to the gaze object may be assumed to be other than the Gaussian distribution. In addition to the EM algorithm, other methods such as SVM (support vector machine) may be used as a method for classifying the gazing point data corresponding to the specific gazing target object and the other gazing point data. .

  The gaze target estimation unit 22 receives the gaze point data that is repeatedly calculated from the gaze point data calculation unit 14. Further, the gaze target estimation unit 22 receives a range corresponding to the specific gaze target from the gaze target range setting unit 20. The gaze target estimating unit 22 gazes the driver of the vehicle when the position indicated by the gaze point data calculated by the gaze point data calculation unit 14 belongs to the range determined by the gaze target range setting unit 20. The target object is estimated to be a specific gaze target object.

  Furthermore, the gaze target estimation unit 22 determines the range of other gaze objects based on the range determined by the gaze target range setting unit 20, and the position indicated by the gaze point data indicates the position of the other gaze target. When belonging to the range, the gaze target object may be estimated as the target object being watched by the driver. As other gaze objects, for example, a region such as the front, left front, and right front outside the vehicle, a right door window, a left door window, and the like may be set. Although not shown, a device for recognizing other vehicles around the host vehicle, pedestrians, road signs, traffic lights, and the like may be provided, and the recognized object may be set as another gaze object.

  The estimation result in the gaze target estimation unit 22 is output to the notification device or the vehicle control device 24. The notification device or the vehicle control device 24 notifies the driver of the direction to be watched or the target object based on the estimation result of the target object, or is careful about the target object around the vehicle that the driver is not aware of. Arouse or perform vehicle control such as vehicle speed reduction or steering.

  Next, the process in the gaze target estimation device 12 according to the present embodiment will be described using the flowchart of FIG. Note that the processing shown in the flowchart of FIG. 2 is repeatedly executed periodically as long as the vehicle continues to operate.

  First, in step S100, information indicating the head position including the gaze direction and the face direction detected by the gaze direction and head position detection device 10 is acquired. In subsequent step S110, gazing point data indicating the position of the gazing point being watched by the driver is calculated based on the information indicating the line-of-sight direction and the head position. In step S120, the calculated gazing point data is accumulated.

  FIG. 3 shows an example of the result of plotting the gazing point data accumulated during a certain period on a graph having the horizontal angle and the vertical angle as two axes. Usually, the driver of the vehicle mainly looks at the front direction of the vehicle front, the front left direction, and the front right direction, but also turns the gaze on the left and right side mirrors at a certain rate for confirmation of the left and right rear of the vehicle. Therefore, when the driver's gaze position is plotted, as shown in FIG. 3, in addition to the collection of gaze positions when looking mainly from the front direction to the left front direction and right front direction, the left side mirror is viewed. A collection of gazing point positions when looking at the camera and a gazing point position when looking at the right side mirror can be made.

  In the subsequent step S130, the condition for extracting the gazing point data looking at the specific gazing target object that is an accessory of the vehicle and setting and updating the threshold for excluding the gazing point data looking outside the passenger compartment is established. Determine whether or not. In this determination process, as described above, it is determined that the condition is satisfied when the driver is seated in the driver's seat, when the seat position of the driver's seat is changed, or when the driving posture is changed. If it is determined that the condition is satisfied, the process proceeds to step S140, and the threshold value is set or updated. Accordingly, it is possible to set a threshold value that can accurately distinguish the gazing point data when gazing at a specific gazing target and the gazing point data when gazing outside the passenger compartment.

  In step S150, it is determined whether or not it is time to set the position range indicated by the gaze point data corresponding to the specific gaze target. In this determination process, as described above, when a predetermined time has elapsed from the previous setting, it is determined that it is the range setting timing of the specific gaze target. Alternatively, when a predetermined operation is performed by the driver, when the vehicle is in a predetermined state, or with respect to the accumulated gazing point data, the number of data included in the vicinity range of the specific gaze object is a predetermined number. When it becomes above, you may determine with it being the range setting timing of a specific gaze object. If it is determined that it is time to set the range of the specific gaze target, the process proceeds to step S160. On the other hand, when it is determined that it is not the time to set the range of the specific gaze target, the process proceeds to step S180.

  In step S160, gazing point data having a distance equal to or smaller than the threshold is extracted from the accumulated gazing point data. In this way, by applying the filtering by the distance to the gazing point data, it is possible to reliably extract the gazing point data that is gazing at a specific gaze target. On the other hand, it is possible to exclude gazing point data when the driver is gazing outside the vehicle.

  FIG. 4 shows the result of extracting gaze point data having a distance equal to or smaller than the threshold from the gaze point data shown in FIG. 3 and plotting the position indicated by the extracted gaze point data on a graph. In the graph of FIG. 3, the gazing point data when looking at the left front direction and the right front direction from the front direction of the vehicle is mostly when looking outside the vehicle, and is therefore excluded from the graph of FIG. 4. Yes. On the other hand, most of the gazing point data when looking at the left side mirror and the right side mirror that are accessories of the vehicle remains in the graph of FIG.

  In the subsequent step S170, using the EM algorithm or the like from the distribution of the positions of the plurality of gaze point data as shown in FIG. Classify the data.

  FIG. 5 shows an example of the classification result. In the example illustrated in FIG. 5, the left side mirror and the right side mirror are specific gaze objects, and the gaze point data is classified into the left and right side mirrors and the other. After such a classification is performed, as shown in FIG. 5, a specific gaze target is covered so as to cover the positions of a plurality of gaze point data classified as gaze point data corresponding to a specific gaze target object. A range corresponding to the object is set.

  In the subsequent step S180, the object being watched by the driver is estimated based on the gazing point data calculated in step S110 and the range set in step S170. Finally, in step S190, for example, it is determined whether or not driving of the vehicle has ended due to the ignition switch being turned off. If it is determined that the operation has been completed, the processing shown in the flowchart of FIG. 2 is terminated. If it is determined that the operation has not yet been completed, the processing from step S100 is repeated.

  The preferred embodiments of the present invention have been described above. However, the present invention is not limited to the above-described embodiments, and various modifications can be made without departing from the spirit of the present invention. is there.

  For example, in the above-described embodiment, the gazing point position is represented by the horizontal angle of the gazing point, the vertical angle of the gazing point, and the distance to the gazing point. The position of the gazing point may be represented by a dimensional coordinate.

  Further, although the threshold value is set and changed in step S140 of the flowchart of FIG. 2, the threshold value may be a constant value, and in this case, the processing of steps S130 and S140 is omitted.

DESCRIPTION OF SYMBOLS 10 Gaze direction and head position detection device 12 Gaze target estimation device 14 Gaze point data calculation unit 16 Gaze point data storage unit 18 Gaze point data extraction unit 20 Gaze target range setting unit 22 Gaze target range estimation unit 24 Notification device or Vehicle control device

Claims (12)

Detecting means (10) for detecting a head position including a gaze direction of each of the left and right eyes of the driver of the vehicle and a face direction of the driver;
Based on the head position including the gaze direction and the face direction, gaze point data indicating the position of the gaze point being watched by the driver is calculated, and the gaze point data is the driver A calculating means (14) including a distance from the gazing point;
Extracting means (16, 18) for extracting a plurality of gazing point data having a distance equal to or less than a predetermined threshold from the gazing point data repeatedly calculated by the calculating unit;
The gazing point data regarded as gazing at least one specific gazing object attached to the vehicle from the distribution of the plurality of gazing point data, targeting the plurality of gazing point data extracted by the extracting means. When, with classifying and other gazing point data, based on the gazing point data classified as gazing point data is considered to gazing at the specific watch object, corresponding to the specific watch object Range setting means (20) for determining the range of the position indicated by the gazing point data;
When the position indicated by the gazing point data calculated by the calculating unit belongs to the range determined by the range setting unit, the object that the driver of the vehicle is gazing at is the specific gazing object. A gaze target estimating device comprising: an estimation means (22) for estimating The calculating means, as the gaze point data, calculates the distance to the horizontal angle, vertical angle, and the gaze point from the driver,
The range setting means is based on gazing point data classified as gazing point data regarded as gazing at the specific gazing object, on a plane having the horizontal angle and the vertical angle as two axes. The gaze object estimation device according to claim 1, wherein a range corresponding to the specific gaze object is defined.   The extraction means includes storage means (16) for storing the gazing point data repeatedly calculated by the calculation means, and sets a distance equal to or less than a predetermined threshold from the gazing point data stored in the storage means. The gaze target estimation apparatus according to claim 1, wherein gaze point data is extracted.   The gaze target estimating apparatus according to any one of claims 1 to 3, wherein the range setting unit periodically updates a range corresponding to the specific gaze target.   The range setting means updates a range corresponding to the specific gaze object when a predetermined operation is performed by a driver of the vehicle. The gaze target estimation device according to claim 1.   The range setting means updates a range corresponding to the specific gaze object when the vehicle is in a predetermined traveling state. Gaze target estimation device.   When the number of data included in the vicinity range of the specific gaze object is greater than or equal to a predetermined number with respect to the gaze data stored in the accumulation means, the range setting unit sets the specific gaze object The gaze target estimating apparatus according to claim 3, wherein a corresponding range is updated.   The threshold for the extraction means to extract the gazing point data is set based on a vehicle body structure in consideration of a positional relationship between the driver and the specific gazing target. The gaze target object estimation device according to any one of 1 to 7. Seating detection means for detecting seating of the vehicle in the driver's seat;
8. The apparatus according to claim 1, further comprising a setting unit configured to set the threshold based on a seat position of the driver seat when seating of the vehicle in the driver seat is detected. The gaze target estimation device according to claim 1.   The gaze target estimation according to claim 9, further comprising an updating unit (S140) that updates the threshold based on a changed seat position when the seat position of the driver's seat of the vehicle is changed. apparatus. Attitude detection means for detecting the attitude of the driver of the vehicle;
The update means (S140) which updates the threshold based on the attitude of the driver after the change when the attitude of the driver detected by the attitude detection means changes. The gaze target object estimation device according to 10.   The gaze target estimation according to any one of claims 8 to 11, wherein when a plurality of the specific gaze objects are determined, the threshold is individually set for each of the plurality of gaze objects. apparatus.

Images