Object recognition control apparatus and object recognition control method

Abstract

To properly recognize objects.SOLUTION: A control unit 20 that is an object recognition control apparatus, comprises: a video data acquisition unit 21 for acquiring video data captured by a camera 11 that images a range including the front of a vehicle; a light distribution information acquisition unit 22 for acquiring light distribution information indicating a light distribution state of headlights of the vehicle; a recognition processing unit 23 for recognizing the video data acquired by the video data acquisition unit 21 as a specific object if a score indicating the likelihood of the specific object is equal to or greater than a threshold; and a presentation processing unit 24 for presenting information of the specific object recognized by the recognition processing unit 23 to a driver of the vehicle. The light distribution information acquisition unit 22 acquires variable light distribution information of the headlights of the vehicle. The recognition processing unit 23 changes the threshold in a partial light-out range in the lighting of the vehicle so as to be lower than a threshold in other lighting range.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to an object recognition control device and an object recognition control method. [Background technology]

[0002] In the recognition of people in a photographed image, if a score indicating human resemblance is equal to or greater than a predetermined threshold in a recognition process using a person recognition dictionary, the object is determined to be a person. There is known a technique for determining that an object is a pedestrian if a score indicating pedestrian resemblance is high using a recognition dictionary for a photographed image (see, for example, Patent Document 1). [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Laid-Open No. 2016-015029 Summary of the Invention [Problem to be solved by the invention]

[0004] A recognition system using a far-infrared camera recognizes people, animals, and other objects in places that are difficult for humans to see and notifies the driver of the presence of the object. For this reason, the area illuminated by the vehicle's headlights is one that the driver can visually respond to, so recognition accuracy is required. On the other hand, areas not illuminated by the vehicle's headlights could be dangerous if the recognition system misses an object.

[0005] The present invention has been made in view of the above, and has as its object to appropriately recognize objects. [Means for solving the problem]

[0006] In order to solve the above-mentioned problems and achieve the object, the object recognition control device of the present invention comprises a video data acquisition unit that acquires video data captured by an imaging unit that captures an area including the front of the vehicle; a light distribution information acquisition unit that acquires light distribution information indicating the light distribution state of the vehicle's headlights; a recognition processing unit that recognizes the video data acquired by the video data acquisition unit as the specific object if a score indicating the likelihood of the video data being a specific object is equal to or greater than a threshold; and a presentation processing unit that presents information about the specific object recognized by the recognition processing unit to the driver of the vehicle, and the recognition processing unit changes the threshold based on the light distribution information of the vehicle's headlights acquired by the light distribution information acquisition unit.

[0007] The object recognition control method executed by the object recognition control device of the present invention includes the steps of acquiring video data captured by a capturing unit that captures an area including the front of the vehicle, acquiring light distribution information indicating the light distribution state of the headlights of the vehicle, varying a threshold value that determines a score indicating the likelihood of the acquired video data being a specific object based on the light distribution information of the headlights of the vehicle, and recognizing the object as the specific object if the score is equal to or greater than the threshold value, and presenting information about the recognized specific object to the driver of the vehicle. [Effects of the Invention]

[0008] According to the present invention, an effect is achieved in that an object can be appropriately recognized. [Brief explanation of the drawings]

[0009] [Figure 1] FIG. 1 is a block diagram showing an example of the configuration of an object recognition device having an object recognition control device according to a first embodiment. [Figure 2] FIG. 2 is a flowchart showing the flow of processing in the object recognition control device according to the first embodiment. [Figure 3] FIG. 3 is a schematic diagram illustrating the illumination range of the low beam. [Figure 4]FIG. 4 is a schematic diagram illustrating the illumination range of the high beam. [Figure 5] FIG. 5 is a flowchart showing the flow of processing in the object recognition control device according to the second embodiment. [Figure 6] FIG. 6 is a schematic diagram illustrating an example of an illumination range of the variable light distribution. [Figure 7] FIG. 7 is a schematic diagram illustrating another example of the illumination range of the variable light distribution. [Figure 8] FIG. 8 is a flowchart showing the flow of processing in the object recognition control device according to the third embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0010] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An object recognition control device and an object recognition control method according to embodiments of the present invention will be described in detail below with reference to the accompanying drawings. However, the present invention is not limited to the following embodiments.

[0011] [First embodiment] <Object Recognition Device> FIG. 1 is a block diagram showing an example configuration of an object recognition device 10 having a control unit 20, which is an object recognition control device according to a first embodiment. The object recognition device 10 recognizes a specific object by performing recognition processing on video data using an object recognition dictionary. The object recognition device 10 notifies the vehicle driver of information related to the recognized specific object. When the vehicle driver has good visibility, the object recognition device 10 reduces the sensitivity of object recognition compared to when the driver has poor visibility.

[0012] The object recognition device 10 includes a camera (photographing unit) 11, a recognition dictionary storage unit 12, a display unit 13, a CAN (Controller Area Network) interface unit (hereinafter referred to as "IF unit") 14, and a control unit (object recognition control device) 20. The object recognition device 10 may be implemented as a function of, for example, a device with a safe driving support function that is pre-installed in a vehicle, a navigation device, a drive recorder, or the like.

[0013] The camera 11 is a camera that captures an area including the area ahead of the vehicle. The camera 11 is mounted at the front of the vehicle. The video data acquisition unit 21 captures an area including at least the area illuminated by the vehicle's headlights. The camera 11 is a far-infrared camera, but may be configured as a visible light camera or a combination of a far-infrared camera and a visible light camera, as long as it can detect objects in an area not illuminated by the vehicle's headlights. The camera 11 is disposed, for example, in a position that allows it to capture an area ahead, which is the direction of travel of the vehicle. The camera 11 continuously captures video from the time the engine starts until it stops, that is, while the vehicle is operating. The camera 11 outputs the captured video data to the video data acquisition unit 21 of the control unit 20. The video data is a moving image composed of, for example, 30 frames per second. In this embodiment, the camera 11 is disposed at the front of the vehicle.

[0014] The recognition dictionary storage unit 12 stores dictionary data for recognizing various objects from video data. For example, the recognition dictionary storage unit 12 performs machine learning on various videos in which specific objects are captured, and stores an object recognition dictionary that can verify that an object included in the video data is a specific object. The recognition dictionary storage unit 12 is, for example, a semiconductor memory element such as a ROM (Read Only Memory) or a flash memory, or a storage device such as an external storage device connected via a network. The recognition dictionary storage unit 12 stores, for example, a dictionary for recognizing specific objects such as people and animals.

[0015] The display unit 13 is a device that displays various types of information, and examples thereof include a display device specific to the object recognition device 10, or a display device shared with other systems including a drive recorder and a navigation system. The display unit 13 is a display including, for example, a liquid crystal display (LCD) or an organic electroluminescence (EL) display. The display unit 13 displays an image based on a video signal output from the presentation processing unit 24 of the control unit 20. In this embodiment, the display unit 13 is disposed in a position that is visible to the driver of the vehicle while driving.

[0016] The IF unit 14 is an interface for acquiring light distribution information of the vehicle headlights via the CAN.

[0017] The light distribution information is information indicating the light distribution state of the vehicle's headlights. The light distribution information includes information indicating whether the vehicle's headlights are on or off. The light distribution information includes information indicating whether the vehicle's headlights are on in high beam or low beam. The light distribution information may also include information indicating whether the vehicle's headlights are on or off.

[0018] <Object Recognition Control Device> The control unit 20 is an arithmetic processing device (control device) configured with, for example, a CPU (Central Processing Unit). The control unit 20 loads a stored program into memory and executes instructions included in the program. The control unit 20 includes an internal memory (not shown) that is used for temporary storage of data in the control unit 20. For this reason, the control unit 20 causes the object recognition method performed by the object recognition device 10. The control unit 20 is also a computer that runs the program according to the present invention. The control unit 20 has a video data acquisition unit 21, a light distribution information acquisition unit 22, a recognition processing unit 23, and a presentation processing unit 24, which are connected to a bus 20X.

[0019] The video data acquisition unit 21 acquires video data captured by the camera 11. More specifically, the video data acquisition unit 21 acquires video data output by the camera 11 that captures the forward direction, i.e., the traveling direction of the vehicle. The video data includes at least the illumination range of the vehicle's headlights. If the camera 11 is a far-infrared camera, the video data acquisition unit 21 acquires video data showing the heat distribution captured by the far-infrared camera.

[0020] The light distribution information acquisition unit 22 acquires light distribution information indicating the light distribution state of the vehicle's headlights. More specifically, the light distribution information acquisition unit 22 acquires light distribution information indicating whether the vehicle's headlights are in high beam or low beam. The light distribution information acquisition unit 22 acquires the light distribution information of the vehicle's headlights via the IF unit 14. When the light distribution of the vehicle's headlights changes in response to a right / left turn operation, the light distribution information acquisition unit 22 acquires light distribution information corresponding to the change in the light distribution direction. Furthermore, the position in the video data on which the recognition processing unit 23 performs recognition processing and the illumination range corresponding to the light distribution state of the headlights are associated in advance.

[0021] The recognition processing unit 23 recognizes a specific object from the video data acquired by the video data acquisition unit 21. When the camera 11 is a far-infrared camera, the recognition processing unit 23 recognizes a person or an animal as the specific object. The recognition processing unit 23 performs object recognition processing to recognize the specific object by performing pattern matching on the video data using an object recognition dictionary stored in the recognition dictionary storage unit 12. A known method can be used in the object recognition processing. The recognition processing unit 23 may perform processing continuously while the object recognition device 10 is running.

[0022] The specific objects include people, two-wheeled vehicles including bicycles ridden by people, other vehicles, and moving objects such as animals. When the camera 11 is a far-infrared camera, the specific objects are people and animals.

[0023] When the score indicating the likelihood of a specific object in the video data acquired by the video data acquisition unit 21 is greater than or equal to a predetermined threshold, the recognition processing unit 23 recognizes it as a specific object. Let the threshold of the score during normal times be A (a specified value). The threshold of the score is changed according to the light distribution information acquired by the light distribution information acquisition unit 22. Let the changed threshold of the score be B (B < A). When the threshold of the score is increased, it becomes more difficult to recognize a specific object, and misrecognition decreases, but there is a possibility of recognition omission. When the threshold of the score is decreased, it becomes easier to recognize a specific object, and the occurrence of recognition omission decreases, but the possibility of misrecognition increases.

[0024] The recognition processing unit 23 changes the threshold of the score based on the light distribution information acquired by the light distribution information acquisition unit 22 to recognize a specific object. In the present embodiment, the recognition processing unit 23 changes the threshold of the score based on the light distribution information of the vehicle's headlight acquired by the light distribution information acquisition unit 22. The recognition processing unit 23 changes the threshold of the score based on the light distribution information acquired by the light distribution information acquisition unit 22 to recognize a specific object.

[0025] The threshold of the score when the vehicle's headlight is in the low beam state is set to a value lower than the threshold of the score when it is in the high beam state. In other words, when the light distribution information indicates that the vehicle's headlight is in the low beam state, the recognition processing unit 23 uses a threshold B lower than the normal threshold A of the score to recognize it as a specific object. When the vehicle's headlight is in the high beam state, since the driver's field of vision is good and can see far, the recognition of the specific object by the recognition processing unit 23 emphasizes accuracy. When the light distribution information indicates that the vehicle's headlight is in the high beam state, the specific object is recognized using the normal threshold A of the score.

[0026] For example, the maximum score indicating the likelihood of a specific object is 1.0. As an example, threshold A is 0.9 and threshold B is 0.7. For example, when the vehicle's headlights are on low beam, the recognition processing unit 23 determines that the detected object is a specific object if the score is 0.7 or higher. For example, when the vehicle's headlights are on high beam, the recognition processing unit 23 determines that the detected object is a specific object if the score is 0.9 or higher. For example, when the recognition processing unit 23 performs object recognition on the condition that the score is 0.9 or higher, when the vehicle's headlights are on low beam, the recognition processing unit 23 switches to recognition processing that uses a lower score threshold.

[0027] The score threshold may be set for each specific object, in other words, for each object recognition dictionary.

[0028] The score threshold may be changed stepwise or linearly.

[0029] The presentation processing unit 24 presents information about the specific object recognized by the recognition processing unit 23 to the vehicle driver. By presenting information about the specific object recognized by the presentation processing unit 24, the vehicle driver is alerted to the specific object. The presentation processing unit 24 presents information about the recognized specific object to the driver using a display on the display unit 13 or an audio output from an audio output unit (not shown). Therefore, when presenting information about the specific object to the driver by display on the display unit 13, the presentation processing unit 24 functions as a display control unit. More specifically, the presentation processing unit 24 generates video for presenting information to the vehicle driver. For example, the presentation processing unit 24 generates presentation video data that displays a frame line over the area of ​​the recognized specific object in the video data. The presentation processing unit 24 may also generate presentation video data that displays text or an icon that alerts the driver to the specific object. The presentation processing unit 24 may also generate audio, such as a warning sound, to be output together with the video. The presentation processing unit 24 outputs a video signal for displaying the generated video to the display unit 13, thereby displaying the video. The presentation processing unit 24 may also output an audio signal for outputting the generated audio together with the video to a speaker (not shown).

[0030] <Processing in the control unit> Next, the flow of processing in the control unit 20 will be described with reference to FIG. 2. FIG. 2 is a flowchart showing the flow of processing in the control unit 20, which is an object recognition control device according to the first embodiment. When the object recognition device 10 is started, the processing of the flowchart shown in FIG. 2 is initiated. The processing of FIG. 2 is started under any condition. For example, the start of the processing may be when the vehicle equipped with the object recognition device 10 becomes available, for example, when the engine of the vehicle is started, or when the operation of the object recognition device 10 is started by a user operation. The processing of step S101 is executed on the condition that the vehicle's headlights are on. The processing of step S101 may be executed on the condition that the surroundings of the vehicle are dark and the vehicle's headlights are on. As a result, when the headlights are on during the day when the surroundings of the vehicle are bright, in other words, when daytime running lights are being used, the processing of step S101 is not executed. The processing of step S101 may be executed on the condition that the vehicle's headlights have started to be turned on by an automatic light function. Furthermore, the processing of step S101 may be executed on the condition that the vehicle is moving.

[0031] Whether the surroundings of the vehicle are dark may be determined based on sensor data acquired by an illuminance sensor (not shown) arranged in the vehicle. Whether the surroundings of the vehicle are dark may also be determined based on brightness or luminance obtained by performing image processing on video data. Whether the surroundings of the vehicle are dark may also be determined by comparing sunset information acquired from an external device (not shown) with current time information that can be acquired by the object recognition device 10. The method for determining whether the surroundings of the vehicle are dark is not limited to these, and any known method can be used.

[0032] 2 starts, the control unit 20 starts image capture and object recognition processing (step S101). More specifically, the control unit 20 causes the camera 11 to start image capture. The control unit 20 acquires image data output by the camera 11 using the image data acquisition unit 21. The control unit 20 recognizes specific objects around the vehicle from the image data acquired by the image data acquisition unit 21 using the recognition processing unit 23. The score threshold for the object recognition processing in step S101 is the normal score threshold A. For example, if the recognition processing unit 23 determines that the score is 0.9 or higher, the control unit 20 determines that the detected object is a specific object. The control unit 20 proceeds to step S102.

[0033] When the process of step S101 starts, control unit 20 determines whether the headlights are on high beam (step S102). More specifically, control unit 20 acquires light distribution information via IF unit 14 using light distribution information acquisition unit 22. If the acquired light distribution information indicates that the headlights are on high beam (Yes in step S102), control unit 20 proceeds to step S103. If the acquired light distribution information indicates that the headlights are not on high beam (No in step S102), control unit 20 proceeds to step S104.

[0034] If the result indicates that the headlights are not on high beam (No in step S102), the control unit 20 causes the recognition processing unit 23 to execute specific object recognition processing using threshold B, which is a threshold lower than threshold A for the normal score (step S104). More specifically, the recognition processing unit 23 performs pattern matching on the video data, and if the score indicating the likelihood of the object being a specific object is 0.7 or higher, the control unit 20 proceeds to step S105.

[0035] If the headlights are on high beam (Yes in step S102), the control unit 20 causes the recognition processing unit 23 to execute specific object recognition processing using the normal score threshold A (step S103). More specifically, the recognition processing unit 23 performs pattern matching on the video data, and if the score indicating the likelihood of the object being a specific object is 0.9 or higher, the recognition processing unit 23 determines that the detected object is a specific object. As a result, when the headlights are on low beam, the detected object is more likely to be recognized as a specific object than when the headlights are on normal beam or high beam. The control unit 20 proceeds to step S105.

[0036] The control unit 20 determines whether or not a specific object has been recognized based on the object recognition result of the recognition processing unit 23 (step S105). If the control unit 20 determines that a specific object has been recognized (Yes in step S105), the control unit 20 proceeds to step S106. If the control unit 20 does not determine that a specific object has been recognized (No in step S105), the control unit 20 proceeds to step S107.

[0037] The control unit 20 causes the presentation processing unit 24 to display a frame line within the range of the recognized specific object (step S106). More specifically, the control unit 20 causes the presentation processing unit 24 to superimpose a frame line surrounding the specific object around the vehicle onto the video data and display it on the display unit 13. The control unit 20 proceeds to step S107.

[0038] The control unit 20 determines whether or not to end the photography and object recognition process (step S107). For example, when the control unit 20 detects an operation to end the object recognition process, or when the vehicle is stopped and the engine is turned off, the control unit 20 determines to end the photography and object recognition process. When the control unit 20 determines to end the photography and object recognition process (Yes in step S107), the control unit 20 ends the process. When the control unit 20 does not determine to end the photography and object recognition process (No in step S107), the control unit 20 executes the process of step S102 again.

[0039] <Effects> As described above, in this embodiment, the score threshold in the object recognition process is changed based on the light distribution information of the vehicle. According to this embodiment, it is possible to appropriately recognize a specific object based on the light distribution information of the vehicle.

[0040] In this embodiment, when the vehicle's headlights are on low beam, the score threshold for the object recognition process is set lower than when the vehicle's headlights are on high beam. According to this embodiment, when the vehicle's headlights are on high beam, the headlights are illuminated over a long distance, making it easier for the vehicle driver to visually identify people and other objects from a long distance. Therefore, this embodiment can reduce erroneous recognition of specific objects and present accurate recognition results for specific objects.

[0041] In this embodiment, when the vehicle's headlights are on low beam, the score threshold for the object recognition process is set lower than when the vehicle's headlights are on high beam. According to this embodiment, when the vehicle's headlights are on low beam, it may be more difficult for the driver to visually identify people and the like than when the vehicle's headlights are on high beam. In this embodiment, when the headlights are on low beam, which does not illuminate a long distance, the occurrence of specific objects not being recognized is reduced, and more appropriate recognition results for specific objects can be presented.

[0042] In this embodiment, a far-infrared camera is used as the camera 11. According to this embodiment, even when the surroundings of the vehicle are dark, people, animals, and the like can be appropriately recognized as specific objects.

[0043] In this embodiment, information about the recognized specific object is presented to the driver. This embodiment can appropriately present the recognition result of the specific object to the driver of the vehicle even in a low beam state where the driver of the vehicle has difficulty seeing a large area.

[0044] [Second embodiment] An object recognition device 10 according to this embodiment will be described with reference to FIGS. 3 to 5. FIG. 3 is a schematic diagram illustrating the low beam illumination range. FIG. 4 is a schematic diagram illustrating the high beam illumination range. FIG. 5 is a flowchart showing the processing flow in a control unit 20, which is an object recognition control device according to a second embodiment. In this embodiment, the object recognition device 10 reduces the object recognition sensitivity in the headlight illumination range compared to the non-illuminated range. The object recognition device 10 has a basic configuration similar to that of the object recognition device 10 according to the first embodiment. In the following description, components similar to those of the object recognition device 10 are denoted by the same or corresponding symbols, and detailed description thereof will be omitted.

[0045] The light distribution information acquired from the vehicle via the IF unit 14 includes information indicating the illumination range of the headlights. The light distribution information may include information indicating the illumination range of the high beam and information indicating the illumination range of the low beam.

[0046] The recognition processing unit 23 recognizes a specific object by using a score threshold value in the non-illuminated area of ​​the vehicle headlights that is lower than the score threshold value in the illuminated area. In other words, the recognition processing unit 23 recognizes the non-illuminated area of ​​the headlights as a specific object by using a threshold value B that is lower than the score threshold value A in normal times.

[0047] The recognition processing unit 23 may recognize a specific object by changing the score threshold between the non-irradiation range of the high beam and the non-irradiation range of the low beam. More specifically, the recognition processing unit 23 may recognize a specific object with the score threshold in the non-irradiation range of the low beam being lower than the score threshold in the non-irradiation range of the high beam. For example, the recognition processing unit 23 may recognize a specific object using a threshold BH that is lower than the normal score threshold A in the non-irradiation range of the high beam. For example, the recognition processing unit 23 may recognize a specific object using a threshold BL (BL < BH) that is lower than the normal score threshold A in the non-irradiation range of the low beam. For example, let BH be 0.8 and BL be 0.7. Note that the irradiation range of the high beam and the irradiation range of the low beam may also be recognized as specific objects using the normal threshold A.

[0048] As shown in FIG. 3, in the case of the low beam, the video data 100 includes an irradiation range 101 and a non-irradiation range 102 of the low beam in the video data 100. In FIG. 3, the boundary line L between the irradiation range 101 and the non-irradiation range 102 is shown as a solid line. The irradiation range 101 is located below the boundary line L than the non-irradiation range 102.

[0049] As shown in FIG. 4, in the case of the high beam, the video data 100 includes an irradiation range 103 and a non-irradiation range 104 of the high beam in the video data 100. In FIG. 4, the boundary line H between the irradiation range 103 and the non-irradiation range 104 is shown as a solid line. The irradiation range 103 is located below the boundary line H than the non-irradiation range 104. The irradiation range 103 of the high beam is wider in front than the irradiation range 101 of the low beam. The irradiation range 103 of the high beam includes a range that is more separated from the vehicle at the central portion in the left-right direction in the figure of the video data 100 compared to the irradiation range 101 of the low beam.

[0050] In the video data 100 shown in Fig. 3, the illuminated area 101 is recognized as a specific object using threshold A, and the non-illuminated area 102 is recognized as a specific object using threshold BL. In the video data 100 shown in Fig. 4, the illuminated area 103 is recognized as a specific object using threshold A, and the non-illuminated area 104 is recognized as a specific object using threshold BH.

[0051] If the high beam illumination range is changed so as not to illuminate preceding and oncoming vehicles when the high beam is turned on, the score threshold is set according to the changed illumination range.

[0052] The recognition processing unit 23 may be configured to execute the recognition process only in the non-illuminated area of ​​the vehicle's headlights. In other words, the recognition processing unit 23 may not execute the recognition process in the illuminated area of ​​the vehicle's headlights. In this case, the recognition processing unit 23 executes the recognition process in the non-illuminated area 102 shown in FIG. 3 and the non-illuminated area 104 shown in FIG. 4. The recognition processing unit 23 does not execute the recognition process in the illuminated area 101 shown in FIG. 3 and the illuminated area 103 shown in FIG. 4. In this case, the processing load can be reduced and processing resources can be further utilized to improve the accuracy of the recognition process.

[0053] <Processing in the control unit> Next, the flow of processing in the control unit 20 will be described with reference to Fig. 5. The processing in steps S111, S112, and S115 to S117 is the same as that in steps S101, S102, and S105 to S107 in the flowchart shown in Fig. 3.

[0054] If the headlights are not on high beam (No in step S112), the control unit 20 causes the recognition processing unit 23 to execute specific object recognition processing using different score thresholds inside and outside the low beam illumination range (step S113). More specifically, the recognition processing unit 23 performs pattern matching on the video data, and in the low beam non-illumination range, if the score indicating the likelihood of the object being a specific object is BL or higher, determines that the detected object is a specific object. The recognition processing unit 23 performs pattern matching on the video data, and in the low beam illumination range, if the score indicating the likelihood of the object being a specific object is A or higher, determines that the detected object is a specific object. This reduces the occurrence of missed recognition of specific objects in the low beam non-illumination range compared to the illumination range. The control unit 20 proceeds to step S115.

[0055] If the headlights indicate high beams (Yes in step S112), the control unit 20 causes the recognition processing unit 23 to execute specific object recognition processing using different score thresholds inside and outside the high beam illumination range (step S114). More specifically, the recognition processing unit 23 performs pattern matching on the video data, and determines that the detected object is a specific object if the score indicating the likelihood of the object being a specific object is BH or higher in the non-high beam illumination range. The recognition processing unit 23 performs pattern matching on the video data, and determines that the detected object is a specific object if the score indicating the likelihood of the object being a specific object is A or higher in the high beam illumination range. This allows for more accurate detection of specific objects in the non-high beam illumination range compared to the non-low beam illumination range. The control unit 20 proceeds to step S115.

[0056] As described above, in this embodiment, the score threshold in the non-illuminated area of ​​the vehicle's headlights is set to a lower value than the score threshold in the illuminated area. The area illuminated by the vehicle's headlights makes it easy for the driver to visually identify people and the like. Therefore, in this embodiment, even in the area not illuminated by the vehicle's headlights, the score threshold in the non-illuminated area is changed in accordance with the light distribution of the headlights, making it possible to present a more appropriate recognition result of a specific object.

[0057] Since the non-illuminated area of ​​the high beam in the direction of travel of the vehicle is located far from the vehicle, the size of the detected specific object on the image is small. Therefore, in this embodiment, it is possible to reduce false recognition in the non-illuminated area of ​​the high beam and present more accurate information about the specific object to the driver of the vehicle. Furthermore, even if a specific object is present in the non-illuminated area of ​​the high beam in the direction of travel of the vehicle, there is still time to avoid danger, so it is useful to reduce false recognition in the non-illuminated area and present more accurate information about the specific object to the user.

[0058] Furthermore, since the low beam non-illumination area in the vehicle's traveling direction is located closer to the vehicle than the high beam non-illumination area, the size of the detected specific object on the image is larger than that of a specific object located further away. Therefore, in this embodiment, the occurrence of missed recognition in the low beam non-illumination area can be reduced, and information that makes it easier for the driver of the vehicle to notice the presence of a specific object near the vehicle can be presented. Furthermore, if a specific object is present in the low beam non-illumination area in the vehicle's traveling direction, there is little time to avoid danger, so it is useful to reduce the occurrence of missed recognition in the non-illumination area and present more appropriate information about the specific object to the user.

[0059] [Third embodiment] 6 to 8, the object recognition device 10 according to this embodiment will be described. FIG. 6 is a schematic diagram illustrating an example of an illumination range of variable light distribution. FIG. 7 is a schematic diagram illustrating another example of an illumination range of variable light distribution. FIG. 8 is a flowchart showing the flow of processing in an object recognition control device according to a third embodiment. The object recognition device 10 increases the sensitivity of object recognition in a partially turned-off area of ​​the headlight compared to other illumination areas. The object recognition device 10 has the same basic configuration as the object recognition device 10 of the first embodiment.

[0060] In this embodiment, the vehicle headlights perform variable light distribution control. More specifically, variable light distribution control is a function that detects oncoming vehicles or preceding vehicles using a camera or radar, and partially turns off the high beams or turns off the low beams, thereby illuminating the high beam range and setting a partially turned-off range as needed. When the headlights are performing variable light distribution control, the partially turned-off range is not illuminated by the headlights. Therefore, if a pedestrian or the like is present in the partially turned-off range, it may not be possible to properly detect it. Furthermore, if a pedestrian or the like is present between the vehicle and the oncoming vehicle, it may be difficult to visually confirm the pedestrian or the like due to glare.

[0061] The light distribution information acquisition unit 22 acquires variable light distribution information of the vehicle headlights. More specifically, the light distribution information acquisition unit 22 acquires variable light distribution information that indicates the light distribution state when the vehicle headlights are subjected to variable light distribution control.

[0062] The variable light distribution information includes information on the partial extinguishing range of the headlights. Furthermore, the position in the video data on which the recognition processing unit 23 performs the recognition process is associated in advance with the illumination range and partial extinguishing range when the headlights perform variable light distribution control.

[0063] The recognition processing unit 23 sets a threshold value for the partially unlit area of ​​the vehicle headlights to a lower value than the threshold values ​​for other illumination areas. In other words, the recognition processing unit 23 recognizes the partially unlit area of ​​the vehicle headlights as a specific object using a threshold value B that is lower than the threshold value A for the score during normal times.

[0064] The recognition processing unit 23 may set the threshold value in the partial extinguished light range of the vehicle's headlight to a value lower than the threshold value in other non-irradiated ranges.

[0065] FIG. 6 is a diagram showing an example of the irradiation range of variable light distribution when an oncoming vehicle is detected. As shown in FIG. 6, when it is variable light distribution when an oncoming vehicle is detected, a partial extinguished light range 107 is generated so that the oncoming vehicle is not irradiated. The video data 100 includes a variable light distribution irradiation range 105, a non-irradiation range 106, and a partial extinguished light range 107 in the video data 100. In FIG. 6, the boundary line KL between the irradiation range 105, the non-irradiation range 106, and the partial extinguished light range 107 is shown as a solid line. The partial extinguished light range 107 is located on the upper right side of the irradiation range 105.

[0066] FIG. 7 is a diagram showing an example of the irradiation range of variable light distribution when a preceding vehicle is detected. As shown in FIG. 7, when it is variable light distribution when a preceding vehicle is detected, a partial extinguished light range 110 is generated so that the preceding vehicle is not irradiated. The video data 100 includes a variable light distribution irradiation range 108, a non-irradiation range 109, and a partial extinguished light range 110 in the video data 100. In FIG. 7, the boundary line KH between the irradiation range 108, the non-irradiation range 109, and the partial extinguished light range 110 is shown as a solid line. The partial extinguished light range 110 is located on the upper center side of the irradiation range 108.

[0067] In the video data 100 shown in FIG. 6, the irradiation range 105 is recognized as a specific object using the threshold value A, and the partial extinguished light range 107 is recognized as a specific object using the threshold value BL. The non-irradiation range 106 may also be recognized as a specific object using the threshold value BL. In the video data 100 shown in FIG. 7, the irradiation range 108 is recognized as a specific object using the threshold value A, and the partial extinguished light range 110 is recognized as a specific object using the threshold value BH (BL < BH). The non-irradiation range 109 may also be recognized as a specific object using the threshold value BH.

[0068] <Processing in the control unit> Next, the flow of processing in the control unit 20 will be described with reference to Fig. 8. The processing in steps S121, S123, and S125 to S127 in Fig. 8 is the same as that in steps S101, S103, and S105 to S107 in the flowchart shown in Fig. 3.

[0069] The control unit 20 determines whether the headlights are performing variable light distribution control (step S122). More specifically, if the light distribution information acquired by the light distribution information acquisition unit 22 indicates that the headlights are not performing variable light distribution control (No in step S122), the control unit 20 proceeds to step S123. If the acquired light distribution information indicates that the headlights are performing variable light distribution control (Yes in step S122), the control unit 20 proceeds to step S124.

[0070] If it indicates that the headlights are performing variable light distribution control (Yes in step S122), the control unit 20 causes the recognition processing unit 23 to execute specific object recognition processing (step S124), using threshold B, which is lower than normal score threshold A, for the partially unlit area and score threshold A for the other illuminated areas. More specifically, the recognition processing unit 23 performs pattern matching on the video data for the partially unlit area, and determines that the detected object is a specific object if the score indicating the specific object-likelihood is 0.7 or higher. The recognition processing unit 23 performs pattern matching on the video data for the illuminated areas other than the partially unlit area, and determines that the detected object is a specific object if the score indicating the specific object-likelihood is 0.9 or higher. The control unit 20 proceeds to step S125.

[0071] As described above, in this embodiment, the occurrence of missed recognition in the partially turned-off range of variable light distribution control is reduced, and information that makes it easy for the driver of the vehicle to notice the presence of a specific object near the vehicle can be presented to the driver. Furthermore, when a specific object is present in a partially turned-off area, there is little time to avoid danger. According to this embodiment, the occurrence of missed recognition in the partially turned-off area is reduced, and more appropriate information about the specific object can be presented to the user.

[0072] In this embodiment, the threshold value in the partially unlit area can be set to a lower value than the threshold value in other non-illuminated areas. According to this embodiment, for example, when a pedestrian or the like is present in the partially unlit area, the occurrence of missed detection can be reduced. According to this embodiment, for example, when a pedestrian or the like is present between the vehicle and an oncoming vehicle and visual confirmation is difficult due to glare, the occurrence of missed detection can be reduced.

[0073] In the present embodiment, the partially unlit area 107 shown in Fig. 6 is an area where glare caused by the headlights of oncoming vehicles is more likely to occur than the partially unlit area 110 shown in Fig. 7. For this reason, a threshold value for the partially unlit area 107 shown in Fig. 6 may be lower than the threshold value for the partially unlit area 110 shown in Fig. 7 to recognize a specific object. The same applies when the partially unlit area 107 shown in Fig. 6 and the partially unlit area 110 shown in Fig. 7 occur simultaneously.

[0074] <Modification> In the above description, the light distribution information is described as being acquired from the vehicle via the IF unit 14, but this is not limiting. The light distribution information may be detected based on luminance or brightness by performing image processing on video data. If the size and shape of the bright area are as shown in FIG. 4, it may be detected that the high beam is on, and if the size and shape of the bright area are as shown in FIG. 3, it may be detected that the low beam is on.

[0075] In the above description, the headlight illumination range has been described as being included in the light distribution information acquired from the vehicle via the IF unit 14, but this is not limiting. The headlight illumination range may be detected based on luminance or brightness by performing image processing on the video data. More specifically, for example, image processing is performed on the video data to recognize bright ranges and dark ranges. The bright range is detected as the illuminated range, and the dark range is detected as the non-illuminated range.

[0076] In the above description, the recognition processing unit 23 has been described as changing the threshold value based on the light distribution information of the vehicle's headlights acquired by the light distribution information acquisition unit 22. The recognition processing unit 23 may also change the threshold value based on vehicle information including wiper operation information, raindrop detection information, fog lamp information, and in-vehicle camera system information. More specifically, when the vehicle information indicates bad weather, such as rain, snow, or fog, the score threshold value may be lower than usual. The vehicle information is acquired from the vehicle via the IF unit 14. This makes it easier for the driver to recognize specific objects that are difficult for the driver to recognize visually when visibility is poor due to bad weather.

[0077] Furthermore, the recognition processing unit 23 may change the score threshold in stages depending on the amount of rainfall, the amount of snowfall, poor visibility due to fog, and the brightness around the vehicle. [Explanation of symbols]

[0078] 10 Object Recognition Device 11 Camera (photography section) 12 Recognition dictionary storage unit 13 Display section 20 Control unit (object recognition control device) 21 Video data acquisition unit 22 Light distribution information acquisition unit 23 Recognition processing section 24 Presentation processing unit 100 video data 101 Irradiation range 102 Non-illuminated area 103 Irradiation range 104 Non-illuminated area

Claims

1. a video data acquisition unit that acquires video data captured by an image capture unit that captures an area including the area ahead of the vehicle; a light distribution information acquisition unit that acquires light distribution information indicating a light distribution state of a headlight of the vehicle; a recognition processing unit that recognizes the video data acquired by the video data acquisition unit as the specific object when a score indicating a likelihood of the video data being a specific object is equal to or greater than a threshold; a presentation processing unit that presents information about the specific object recognized by the recognition processing unit to a driver of the vehicle; Equipped with the light distribution information acquisition unit acquires variable light distribution information of a headlamp of the vehicle, The recognition processing unit sets the threshold value in a partially turned-off range of the headlights of the vehicle to a lower value than the threshold value in other illumination ranges. Object recognition control device.

2. An object recognition control method executed by an object recognition control device, a video data acquisition step of acquiring video data captured by an image capturing unit that captures an area including the area ahead of the vehicle; a light distribution information acquisition step of acquiring light distribution information indicating a light distribution state of a headlamp of the vehicle; a recognition processing step of recognizing the acquired video data as a specific object when a score indicating a likelihood of the video data being a specific object is equal to or greater than a threshold; a presentation processing step of presenting information about the recognized specific object to a driver of the vehicle; Including, the light distribution information acquisition step acquires variable light distribution information of a headlamp of the vehicle, The recognition processing step sets the threshold value in a partially turned-off range of the headlights of the vehicle to a lower value than the threshold value in other illumination ranges. Object recognition control method.

Images