WO2022145054A1 - 画像処理装置、画像処理方法、及び記録媒体 - Google Patents
画像処理装置、画像処理方法、及び記録媒体 Download PDFInfo
- Publication number
- WO2022145054A1 WO2022145054A1 PCT/JP2021/000031 JP2021000031W WO2022145054A1 WO 2022145054 A1 WO2022145054 A1 WO 2022145054A1 JP 2021000031 W JP2021000031 W JP 2021000031W WO 2022145054 A1 WO2022145054 A1 WO 2022145054A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- image
- processing
- image processing
- distant
- unit
- Prior art date
Links
- 238000012545 processing Methods 0.000 title claims abstract description 679
- 238000003672 processing method Methods 0.000 title claims description 7
- 238000000034 method Methods 0.000 claims description 92
- 230000008569 process Effects 0.000 claims description 43
- 230000002194 synthesizing effect Effects 0.000 claims description 15
- 238000003384 imaging method Methods 0.000 abstract description 2
- 230000015572 biosynthetic process Effects 0.000 description 30
- 238000003786 synthesis reaction Methods 0.000 description 30
- 238000010586 diagram Methods 0.000 description 16
- 238000011156 evaluation Methods 0.000 description 14
- 230000006870 function Effects 0.000 description 13
- 238000004891 communication Methods 0.000 description 11
- 238000006243 chemical reaction Methods 0.000 description 9
- 239000000284 extract Substances 0.000 description 8
- 230000010354 integration Effects 0.000 description 8
- 230000008859 change Effects 0.000 description 6
- 238000012937 correction Methods 0.000 description 6
- 239000000203 mixture Substances 0.000 description 5
- 230000009467 reduction Effects 0.000 description 5
- 230000000694 effects Effects 0.000 description 3
- 238000012986 modification Methods 0.000 description 3
- 230000004048 modification Effects 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 238000001514 detection method Methods 0.000 description 2
- 238000001454 recorded image Methods 0.000 description 2
- 230000011218 segmentation Effects 0.000 description 2
- 238000006467 substitution reaction Methods 0.000 description 2
- 230000009466 transformation Effects 0.000 description 2
- 208000004547 Hallucinations Diseases 0.000 description 1
- 230000001133 acceleration Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 239000002131 composite material Substances 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 239000003973 paint Substances 0.000 description 1
- 230000002093 peripheral effect Effects 0.000 description 1
Images
Classifications
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B60—VEHICLES IN GENERAL
- B60R—VEHICLES, VEHICLE FITTINGS, OR VEHICLE PARTS, NOT OTHERWISE PROVIDED FOR
- B60R1/00—Optical viewing arrangements; Real-time viewing arrangements for drivers or passengers using optical image capturing systems, e.g. cameras or video systems specially adapted for use in or on vehicles
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T3/00—Geometric image transformations in the plane of the image
- G06T3/40—Scaling of whole images or parts thereof, e.g. expanding or contracting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/50—Depth or shape recovery
- G06T7/536—Depth or shape recovery from perspective effects, e.g. by using vanishing points
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/70—Determining position or orientation of objects or cameras
- G06T7/73—Determining position or orientation of objects or cameras using feature-based methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/25—Determination of region of interest [ROI] or a volume of interest [VOI]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/56—Context or environment of the image exterior to a vehicle by using sensors mounted on the vehicle
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/56—Context or environment of the image exterior to a vehicle by using sensors mounted on the vehicle
- G06V20/58—Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/56—Context or environment of the image exterior to a vehicle by using sensors mounted on the vehicle
- G06V20/588—Recognition of the road, e.g. of lane markings; Recognition of the vehicle driving pattern in relation to the road
-
- G—PHYSICS
- G08—SIGNALLING
- G08G—TRAFFIC CONTROL SYSTEMS
- G08G1/00—Traffic control systems for road vehicles
- G08G1/09—Arrangements for giving variable traffic instructions
- G08G1/0962—Arrangements for giving variable traffic instructions having an indicator mounted inside the vehicle, e.g. giving voice messages
- G08G1/09623—Systems involving the acquisition of information from passive traffic signs by means mounted on the vehicle
-
- G—PHYSICS
- G08—SIGNALLING
- G08G—TRAFFIC CONTROL SYSTEMS
- G08G1/00—Traffic control systems for road vehicles
- G08G1/16—Anti-collision systems
- G08G1/166—Anti-collision systems for active traffic, e.g. moving vehicles, pedestrians, bikes
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N7/00—Television systems
- H04N7/18—Closed-circuit television [CCTV] systems, i.e. systems in which the video signal is not broadcast
- H04N7/183—Closed-circuit television [CCTV] systems, i.e. systems in which the video signal is not broadcast for receiving images from a single remote source
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B60—VEHICLES IN GENERAL
- B60R—VEHICLES, VEHICLE FITTINGS, OR VEHICLE PARTS, NOT OTHERWISE PROVIDED FOR
- B60R2300/00—Details of viewing arrangements using cameras and displays, specially adapted for use in a vehicle
- B60R2300/30—Details of viewing arrangements using cameras and displays, specially adapted for use in a vehicle characterised by the type of image processing
- B60R2300/307—Details of viewing arrangements using cameras and displays, specially adapted for use in a vehicle characterised by the type of image processing virtually distinguishing relevant parts of a scene from the background of the scene
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B60—VEHICLES IN GENERAL
- B60R—VEHICLES, VEHICLE FITTINGS, OR VEHICLE PARTS, NOT OTHERWISE PROVIDED FOR
- B60R2300/00—Details of viewing arrangements using cameras and displays, specially adapted for use in a vehicle
- B60R2300/60—Details of viewing arrangements using cameras and displays, specially adapted for use in a vehicle characterised by monitoring and displaying vehicle exterior scenes from a transformed perspective
- B60R2300/602—Details of viewing arrangements using cameras and displays, specially adapted for use in a vehicle characterised by monitoring and displaying vehicle exterior scenes from a transformed perspective with an adjustable viewpoint
- B60R2300/605—Details of viewing arrangements using cameras and displays, specially adapted for use in a vehicle characterised by monitoring and displaying vehicle exterior scenes from a transformed perspective with an adjustable viewpoint the adjustment being automatic
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30248—Vehicle exterior or interior
- G06T2207/30252—Vehicle exterior; Vicinity of vehicle
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/07—Target detection
Definitions
- the present invention relates to an image processing device, an image processing method, and a recording medium.
- Image processing technology using a computer is widespread.
- efforts are being made to perform image processing on images taken by a camera mounted on a vehicle to help provide various services.
- Patent Document 1 among the images taken by the rear side camera mounted on the vehicle, the area of interest of interest to the driver is enlarged and displayed in a composite manner, so that the driver can easily recognize the image.
- An image processing device that presents information is disclosed.
- the image processing includes image recognition that recognizes the content reflected in the image.
- image recognition there is area recognition (also referred to as area division or segmentation).
- Area recognition is a technique for estimating the type of subject represented in an area for each area included in the image by inputting an image.
- Non-Patent Document 1 is an example of such area recognition.
- the inventor of the present invention has found the following problems in image processing.
- the part occupied by a distant subject in the image is small.
- such a small part cannot be easily recognized by image processing. That is, there is a problem that it is difficult to accurately recognize a distant subject when performing image processing on a captured image.
- Cited Document 1 merely displays an area of interest that is of interest to the driver in an easy-to-see manner for the driver. That is, it does not solve the above-mentioned problem of accurately recognizing a distant subject.
- One of the objects of the present invention is an image processing apparatus, an image processing method, and an image processing method capable of accurately recognizing a distant subject when the above-mentioned problems are solved and image processing is performed on a captured image. It is to provide a recording medium.
- the image processing device includes an image acquisition means for acquiring an image taken by an image pickup device, a first processing means for performing a first image processing on the image, and a distant portion of the image.
- a distant specifying means to be specified a second processing means for performing a second image processing different from the first image processing on a distant portion of the image, a processing result of the first image processing, and the second image processing. It is provided with an output means for outputting based on the processing result of image processing.
- the image processing apparatus acquires an image taken by the image pickup apparatus, performs the first image processing on the image, identifies a distant portion of the image, and describes the image.
- a second image processing different from the first image processing is performed on the distant portion of the image, and output is performed based on the processing result of the first image processing and the processing result of the second image processing.
- an image taken by an image pickup apparatus is acquired by a computer, the first image processing is performed on the image, a distant portion of the image is specified, and the distant portion of the image is specified.
- FIG. 1 is a diagram showing a configuration of an image processing system according to the first embodiment.
- the image processing system includes an image processing device 10 and an image pickup device 20.
- the image processing device 10 and the image pickup device 20 are communicably connected.
- the image processing device 10 acquires an image taken by the image pickup device 20, performs image processing on the image, and outputs the image based on the processing result.
- the image processing device 10 is realized, for example, as a computer mounted on a vehicle.
- the present invention is not limited to this, and the image processing apparatus 10 may be realized as a server installed in, for example, a data center or the like.
- the image pickup device 20 captures an image.
- the image pickup device 20 is, for example, a camera of a drive recorder mounted on a vehicle. In this case, the image pickup device 20 generates an image of the surroundings of the vehicle, for example, the front.
- the image pickup apparatus 20 may be, for example, a camera installed on the roadside of a road or a camera installed inside a facility. Further, the image captured by the image pickup apparatus 20 may be a still image or an image (moving image) of a plurality of frames continuously in time.
- the image processing device 10 and the image pickup device 20 may be connected by a wired communication such as a wired LAN or an internal bus communication, or may be connected by a wireless communication such as a wireless LAN or a short-range communication.
- a wired communication such as a wired LAN or an internal bus communication
- a wireless communication such as a wireless LAN or a short-range communication.
- the image processing device 10 and the image pickup device 20 may be connected by an internal bus of the vehicle, but the present invention is not limited to this.
- a plurality of image processing devices 10 and image pickup devices 20 may exist.
- the image processing device 10 and the image pickup device 20 do not necessarily have to be connected one-to-one, and may be connected one-to-many or many-to-many.
- a plurality of image pickup devices 20 may be connected to one image processing device 10.
- FIG. 2 is a diagram showing a functional block of the image processing device 10 in the first embodiment.
- the image processing device 10 includes an image acquisition unit 110, an image processing unit 120, a scene recognition unit 130, and an output unit 140.
- the image acquisition unit 110 functions as a means for acquiring an image captured by the image pickup device 20.
- the image processing unit 120 functions as a means for performing image processing on the acquired image and generating a processing result.
- the image processing unit 120 further includes a first processing unit 121, a distant identification unit 122, a second processing unit 123, and a compositing unit 124.
- the first processing unit 121 functions as a means for performing a predetermined first image processing on the acquired image.
- the range to be processed by the first image processing by the first processing unit 121 is, for example, the entire image.
- the present invention is not limited to this, and the first processing unit 121 may use mask processing or the like to exclude a part of the range of the image (for example, a distant portion of the image) from the processing target of the first image processing. ..
- the first processing unit 121 can perform area recognition as the first image processing.
- area recognition is mainly performed as the first image processing.
- the first processing unit 121 generates the first processing result as a result of performing the above-mentioned first image processing.
- the distant identification unit 122 functions as a means for identifying a distant portion from the acquired image.
- the distant portion is a portion of the image including a distant subject.
- the distant portion is represented by, for example, a rectangle including a distant subject in the image.
- the present invention is not limited to this, and the distant portion may be represented by a polygon other than a rectangle, a circle, an ellipse, or any other shape.
- the distant portion is not limited to a continuous single shape in such an image, but may be a plurality of discrete shapes.
- the distant identification unit 122 generates distant identification information expressing the distant portion in a predetermined format as a result of specifying the distant portion.
- the distant identification information is, for example, the coordinates of each point of the rectangle in the image.
- the second processing unit 123 functions as a means for performing a second image processing different from the first image processing on the distant portion of the acquired image based on the above-mentioned distant identification information.
- the second processing unit 123 uses, as the second image processing, a process of enlarging a distant portion of an image, performing a predetermined process on the enlarged image, and reducing the processing result of the predetermined process. be able to. Further, for example, the second processing unit 123 can use, as the second image processing, an image processing in which a setting different from that of the first image processing is applied to a distant portion of the image.
- the second processing unit 123 generates a second processing result as a result of performing the predetermined second image processing as described above.
- the synthesizing unit 124 functions as a means for synthesizing the above-mentioned first processing result and the second processing result.
- the synthesizing unit 124 generates a synthesizing process result which is the result of synthesizing.
- the scene recognition unit 130 functions as a means for performing scene recognition based on at least one of the above-mentioned first processing result and second processing result and the above-mentioned synthesis processing result.
- the scene recognition is a process of recognizing the meaning of the scene represented in the image.
- the scene recognition unit 130 generates a scene recognition result as a result of performing scene recognition.
- the output unit 140 functions as a means for performing a predetermined output based on at least one of the above-mentioned first processing result and the second processing result, the above-mentioned synthesis processing result, and the above-mentioned scene recognition result. ..
- FIG. 3 is a flowchart showing the operation of the image processing device 10 in the first embodiment.
- the image acquisition unit 110 of the image processing device 10 acquires a captured image from the image pickup device 20 (step S110 in FIG. 3). For example, the image acquisition unit 110 acquires an image including the road in front of the vehicle as shown in FIG.
- the first processing unit of the image processing device 10 performs predetermined first image processing on the image acquired by the image acquisition unit 110 to generate a first processing result (step S120 in FIG. 3).
- the first processing unit 121 can perform area recognition as the first image processing.
- area recognition also referred to as region division or segmentation
- the first processing unit 121 uses the acquired image as an input image to perform region recognition (also referred to as region division or segmentation) on the input image, and each region included in the input image is represented in that region. Estimate the type of subject and generate the processing result.
- FIG. 5 is a diagram showing an example of the processing result of area recognition.
- the area recognition processing result is, for example, an image having the same resolution as the input image, and is expressed in the form of an image in which the subject type ID to which the corresponding pixel of the input image belongs is stored in each pixel. Will be done.
- the subject type ID is an identifier that identifies the type of the subject.
- the subject type ID is a numerical value of 1, 2, 3, 4, 5, and 6, and each corresponds to a person, an automobile, a building, other, a road, and the sky as the subject type.
- the type of subject is not limited to the example shown in FIG. 5, for example, a two-wheeled vehicle, a sign, a traffic light, a white line, a stop line, an obstacle, a pedestrian crossing, a parking lot (parking space on the road shoulder), a paint on the road, a sidewalk, and a drive. It may include ways (vehicle passages on sidewalks connecting roadways and facilities), railroad tracks, and vegetation.
- the first processing unit 121 calculates the reliability when estimating the type of the subject represented in each area, and includes the reliability in the processing result. May be good.
- the reliability for each pixel may be separately generated as additional information and included in the processing result.
- the first processing unit 121 executes the first image processing as described above to generate the first processing result.
- the distant identification unit 122 of the image processing device 10 identifies the distant portion based on the image acquired by the image acquisition unit 110 and generates the distant identification information (step S130 in FIG. 3).
- FIG. 6 is a flowchart showing the operation of the distant identification unit 122.
- the image acquired by the image acquisition unit 110 is an image including a road.
- the distant identification unit 122 estimates the vanishing point of the road for the image acquired by the image acquisition unit 110 (step S131 in FIG. 6).
- the vanishing point of the road will be described.
- the vanishing point of a road is such a point where distant roads are aggregated in the image.
- FIG. 7 is a diagram showing an example of a vanishing point of a road in such an image including a road.
- the point VP represents the vanishing point of the road.
- the distant identification unit 122 performs area recognition on the acquired image.
- the distant identification unit 122 extracts the point at the uppermost part of the image from the area where the type of the subject is estimated to be a road in the processing result of the area recognition, and estimates it as the vanishing point of the road. This is because the image pickup apparatus 20 usually captures the road in such an arrangement that the farther the road is, the higher the image is.
- FIG. 8 is a diagram showing an example of the vanishing point of the road estimated in this way. In FIG. 8, the point VP represents the estimated vanishing point of the road.
- simple area recognition may be used as the area recognition performed by the distant identification unit 122.
- the distant identification unit 122 may use the area recognition in which the type of the subject is limited to a small number (for example, a road and two others) as a simple area recognition.
- the distant identification unit 122 may reduce the image and perform area recognition on the reduced image as a simple area recognition. By using such a simple area recognition, it is possible to reduce the processing load in the distant specific unit 122.
- the distant identification unit 122 does not perform area recognition, and the first processing unit 121 does not perform area recognition.
- the generated area recognition processing result may be used. By omitting the area recognition in this way, the processing load in the distant specific unit 122 can be reduced.
- the distant identification unit 122 estimates a line representing the road edge for each of the left and right road edges on the road.
- the distant identification unit 122 estimates the point where the lines representing the left and right road edges intersect as the vanishing point of the road.
- FIG. 9 is a diagram showing an example of a line representing such a roadside and a vanishing point of the road.
- the line LL represents the left road edge
- the line RL represents the right road edge
- the point VP represents the vanishing point of the road.
- various methods can be used as a method for estimating the line representing the roadside by the distant identification unit 122.
- the distant identification unit 122 performs area recognition on the acquired image, extracts an area in which the type of the subject is presumed to be a road from the processing result of the area recognition, and left and right of the extracted area. Each end may be approximated by a straight line, and the straight line may be used as a line representing the road end.
- the distant identification unit 122 may detect the white line and / and the guardrail from the acquired image, approximate the detected white line and / and the guardrail with a straight line, and use the straight line as a line representing the roadside. good.
- the distant identification unit 122 has been described as approximating a line representing a roadside with a straight line, but the present invention is not limited to this, and a curved line may be used for approximation.
- a curved line may be used for approximation.
- the distant identification unit 122 may select whether to use a straight line approximation or a curved line approximation for each image.
- the distant identification unit 122 may perform approximation with a straight line and approximation with a curve, and may select the one having the smaller approximation error.
- the image processing device 10 can acquire the shooting position of the image and the road map information
- the distant identification unit 122 has a straight line on the road displayed in the image based on the shooting position and the road map information. It is possible to estimate whether it is a curve or a curve, and select whether to use a straight line approximation or a curved line approximation depending on the result of the estimation.
- the distant identification device 20 is specified.
- the unit 122 estimates whether the road displayed in the image is a straight line or a curved line based on the traveling data, and depending on the result of the estimation, whether the approximation is a straight line or a curved line. You may choose whether to use. In this way, by selecting whether the distant identification unit 122 uses a straight line approximation or a curved line approximation, the vanishing point of the road is estimated accurately according to the shape of the road displayed in the image. be able to.
- the distant identification unit 122 determines the distant portion based on the estimated vanishing point of the road (step S132 in FIG. 6).
- the distant identification unit 122 determines a portion occupying a predetermined ratio in the image centered on the estimated vanishing point as the distant portion.
- FIG. 10 is a diagram showing a distant portion determined in this way.
- the point VP represents the vanishing point of the road and the partial FR represents the distant portion.
- the shape of the distant portion is, for example, a rectangle having the same aspect ratio as the acquired image.
- the shape of the distant portion is not limited to this, and may be a rectangle having an aspect ratio different from that of the acquired image, or may be a polygon other than the rectangle, a circle, an ellipse, or any other shape. ..
- the predetermined ratio is, for example, 1/16 of the area of the acquired image.
- the present invention is not limited to this, and other ratios may be used.
- the distant identification unit 122 lists a plurality of portions including the estimated vanishing point of the road as candidates for the distant portion.
- FIG. 11 is a diagram showing an example of a candidate for a distant portion listed by the distant identification unit 122.
- the point VP is the estimated vanishing point of the road
- the partial FR1 is the portion including the vanishing point of the road in the lower half
- the partial FR2 is the portion including the vanishing point of the road in the upper half
- the partial FR3 is the vanishing point of the road.
- the part FR4 is the part including the vanishing point of the road in the left half.
- the distant identification unit 122 can list such a plurality of rectangular portions as candidates for the distant portion. However, not limited to this, the distant identification unit 122 can list any number of other portions occupying any shape, size, and position as candidates for the distant portion.
- the distant identification unit 122 evaluates a plurality of candidates for the distant portion based on a predetermined criterion, and determines the candidate with the highest evaluation as the distant portion.
- the distant identification unit 122 can evaluate a plurality of distant portion candidates based on the processing result of the area recognition. As an example, the distant identification unit 122 can give a high evaluation when a large number of areas of the type of subject (for example, a road) to be watched are included in the candidates of the distant portion. By evaluating in this way, it is possible to determine a portion containing a large number of subjects to be watched as a distant portion.
- the remote identification unit 122 may use the simple area recognition as shown in the first example of the vanishing point estimation as the area recognition. Further, in the first example of vanishing point estimation, when the first processing unit 121 has already performed area recognition as the first image processing, the distant identification unit 122 does not perform area recognition and the first processing unit 121 The processing result of the area recognition generated by may be used. Furthermore, in the first example of vanishing point estimation, when the distant identification unit 122 has already performed the area recognition, the distant identification unit 122 does not perform further area recognition, and the processing result of the area recognition already performed is performed. You may use it. By omitting the area recognition in this way, the processing load in the distant specific unit 122 can be reduced.
- the image acquisition unit 110 acquires images (moving images) of a plurality of frames continuously in time from the image pickup apparatus 20.
- These multi-frame images (moving images) are images including roads.
- the distant identification unit 122 performs the processing as described in the first example of the distant identification for each frame of the images (moving images) of the plurality of frames, and generates a plurality of distant portions that are continuous in time. do.
- the distant identification unit 122 integrates a plurality of distant portions that are continuous in time, and determines one distant portion.
- the integration of a plurality of distant parts is performed, for example, by statistically processing the plurality of distant parts.
- the distant identification unit 122 excludes a distant portion whose position and size are extremely different from those of the other distant portions among a plurality of distant portions which are continuous in time.
- the distant identification unit 122 calculates a representative position and size for the remaining distant portion (for example, calculates the average of the positions and sizes).
- the portion having the position and size calculated in this way is defined as the distant portion.
- the present invention is not limited to this, and the distant identification unit 122 may use other statistical processing as the integration of the distant part.
- the integration of a plurality of distant portions can be performed at predetermined time units.
- the distant identification unit 122 may divide a plurality of frames of images (moving images) into time units of 5 seconds and integrate the images (moving images) in the time units of 5 seconds.
- the present invention is not limited to this, and the remote identification unit 122 may be integrated for each other fixed or variable time unit.
- the distant part 122 temporarily has a road area in the image due to a vehicle, a person, an obstacle, or the like. Even when concealed, it is possible to accurately identify the distant part of the road.
- the distant identification unit 122 specifies a predetermined range as a distant portion.
- the distant identification unit 122 can use a rectangular portion having a size of 1/16 of the area of the acquired image and whose center coincides with the center of the image as a predetermined range.
- the present invention is not limited to this, and the distant identification portion 122 may use any other portion occupying any shape, size, and position as a predetermined range.
- the above-mentioned predetermined range can be set by, for example, the user or manager of the image processing device 10, or the user or manager of the image pickup device 20. After the image pickup device 20 is installed, these users or managers may check the image taken by the image pickup device 20 and set a range considered to represent a distant portion. As an example, when the image pickup device 20 is a camera of a drive recorder mounted on a vehicle, the user or the manager considers that the angle of view or the like is confirmed from the image taken by the camera and represents a distant part of the road. The range may be set. When there are a plurality of image pickup devices 20, the predetermined range may be different for each image pickup device 20.
- the recognition may be performed by using data interpolation (also called image hallucination) together with the area recognition.
- data interpolation also called image hallucination
- the area of an object in the foreground of a screen such as a vehicle is specified by performing area recognition on an image taken by a camera mounted on the vehicle, and further data interpolation is performed.
- the distant identification unit 122 may recognize the area of the road by using the technique disclosed in Non-Patent Document 1. In this way, by using data interpolation together with area recognition, the distant identification unit 122 can accurately cover the distant part of the road even when the road area in the image is temporarily concealed by a vehicle, a person, an obstacle, or the like. Can be identified.
- the distant identification unit 122 when the distant identification unit 122 processes an image (moving image) of a plurality of frames continuously in time, the distant identification unit 122 may omit the processing of some frames. good.
- the distant identification unit 122 divides a plurality of frames of images (moving images) into time units of 5 seconds, performs specific processing of the distant portion only for one frame in the time unit of 5 seconds, and performs other processing. Processing may be omitted for frames.
- the present invention is not limited to this, and the remote identification unit 122 may omit the processing of an arbitrary number of frames for each of other fixed or variable time units. In this case, the distant identification unit 122 may temporarily store the specified distant portion when the frame is processed.
- the distant identification unit 122 may specify the temporarily stored distant portion (in the previous frame in time) as the distant portion in the frame. .. By omitting the processing of a part of the frames in this way, the processing load in the distant specific unit 122 can be reduced.
- the distant identification unit 122 After specifying the distant portion as described above, the distant identification unit 122 generates distant identification information expressing the distant portion in a predetermined format.
- the distant identification information is, for example, the coordinates of each point of the rectangle in the image.
- the present invention is not limited to this, and any format can be used as the distant specific information according to the shape of the distant portion and the like.
- the second processing unit 123 of the image processing device 10 makes a first image with respect to the distant portion of the image acquired by the image acquisition unit 110 based on the distant identification information generated by the distant identification unit 122. A predetermined second image processing different from the processing is performed, and the second processing result is generated (step S140 in FIG. 3).
- FIG. 12 is a diagram showing the operation of the first example of the second image processing.
- the second processing unit 123 enlarges a distant portion of the image, performs a predetermined process on the enlarged image, and reduces the processing result of the predetermined process. ..
- the second processing unit 123 cuts out a distant portion of the image from the image acquired by the image acquisition unit 110 by using the distant identification information generated by the distant identification unit 122. Then, the second processing unit 123 enlarges the cut-out image to a predetermined size (step S141 in FIG. 12).
- the predetermined size is, for example, the same size as the size of the acquired image.
- the present invention is not limited to this, and the predetermined size may be any other size.
- the enlargement of the image can be performed by using the nearest neighbor interpolation method, the bicubic interpolation method, the bicubic interpolation method, or other known methods.
- the second processing unit 123 performs predetermined processing on the enlarged image and generates a processing result (step S142 in FIG. 12). For example, when the first processing unit 121 performs area recognition as the first image processing, the second processing unit 123 performs area recognition on the enlarged image and generates a processing result. For example, when the distant portion is the partial FR of FIG. 10, the second processing unit 123 enlarges the partial FR, performs region recognition on the enlarged image, and generates a processing result as shown in FIG. ..
- the second processing unit 123 reduces the processing result for the enlarged image (step S143 in FIG. 12).
- FIG. 14 is a diagram schematically showing the reduction of the processing result when the predetermined processing performed in step S142 is area recognition.
- the area recognition processing result is, for example, an image having the same resolution as the input image, and is expressed in the form of an image in which the subject type ID to which the corresponding pixel of the input image belongs is stored in each pixel.
- the processing result ER of FIG. 14 is an example of the processing result of area recognition for the enlarged image expressed in such a format.
- the resolution of the processing result ER is the same as that of the enlarged image.
- the processing result RR in FIG. 14 is an example of the processing result obtained by reducing the processing result ER.
- the resolution of the processing result RR is the same as that of the image before being enlarged.
- the reduction of the processing result means to determine the subject type ID to be stored in each pixel of the processing result RR based on the processing result ER.
- the second processing unit 123 sequentially selects the pixels of the processing result RR to be determined.
- Pixel RP1 is an example of pixels selected in this way.
- the second processing unit 123 extracts the pixels on the processing result ER corresponding to the positions of the selected pixels on the processing result RR.
- the second processing unit 123 may extract a single pixel on the corresponding processing result ER, or may extract a plurality of pixels including the periphery.
- Pixels EP1, EP2, EP3, and EP4 are examples of pixels extracted in this way. In the example of FIG. 14, four pixels are extracted, but any number of other pixels may be extracted.
- the second processing unit 123 determines the subject type ID to be stored in the pixels of the processing result RR based on the extracted pixels.
- Various methods can be used for this determination.
- the second processing unit 123 may use the methods shown in the following (A) to (D) as the method for determining the above.
- (A) Use the most frequent subject type ID
- (B) Determine the subject type ID based on the priority of the predetermined subject type
- (C) Prioritize the subject type from the processing result of the area recognition for the distant part Determine the degree and determine the subject type ID based on the priority
- D Determine the priority of the subject type from the comparison of the processing results between the area recognition for the distant part and the area recognition for the acquired image, and determine the priority.
- the subject type ID is determined based on the above.
- (A) to (D) will be described in detail.
- the second processing unit 123 determines the most frequent subject type ID included in the extracted pixels as the stored subject type ID.
- the extracted pixels are pixels EP1, EP2, EP3, and EP4, and the subject type IDs are EP1: 5 (road), EP2: 1 (person), and EP3: 5 (road), respectively.
- EP4: 5 (road) the second processing unit 123 determines the subject type ID stored in the pixel RP1 to be 5 (road).
- the second processing unit 123 determines the subject type ID to be stored based on the priority of the predetermined subject type.
- the extracted pixels are pixels EP1, EP2, EP3, and EP4, and the subject type IDs are EP1: 5 (road), EP2: 1 (person), and EP3: 5 (road), respectively.
- EP4: 5 (road) when "person" is defined as the priority subject type, the second processing unit 123 determines the subject type ID to be stored in the pixel RP1 to be 1 (person). can do.
- the format of priority is not limited to the above example, and various types can be used.
- the priority format may be expressed as a weighting coefficient for each type of subject.
- the method for determining the subject type ID based on the priority is not limited to the above example, and various methods can be used.
- the second processing unit 123 calculates the number of pixels for each type of subject for the extracted pixels, and calculates the evaluation value by multiplying the number of pixels and the above weighting coefficient for each type of subject.
- the subject type ID to be stored may be determined by comparing the calculated evaluation values.
- the second processing unit 123 by determining the subject type ID to be stored based on the priority of the subject type, the second processing unit 123 appropriately processes the subject to be recognized with priority. Can be included in the results.
- the second processing unit 123 determines the priority of the subject type from the processing result of the area recognition for the enlarged image corresponding to the distant portion of the image, and based on the priority, the second processing unit 123 determines the priority. Determine the subject type ID to be stored.
- the second processing unit 123 calculates the number of pixels for each type of subject from the processing result ER, and determines the priority of the type of subject according to the ratio of the calculated number of pixels. For example, when the ratio of the number of pixels recognized as "human” is small, "human” may be defined as the priority subject type. This makes it possible, for example, to prevent a small "human” area from being absorbed and lost by a surrounding subject (eg, "road”) area due to reduction. For example, in an image in which a person appears large, the priority subject type may be a surrounding subject (for example, "road”) instead of "person", but even in this case, it is usually "person". No space is lost.
- a surrounding subject for example, "road
- the second processing unit 123 may determine the priority of the subject type by any other method or format.
- the second processing unit 123 After determining the priority of the subject type, the second processing unit 123 determines the subject type ID to be stored based on the priority of the subject type by using the same method as in (B) above.
- the second processing unit 123 can be, for example, by determining the priority of the subject type from the processing result of the area recognition for the enlarged image corresponding to the distant portion of the image. Even a subject that occupies a small proportion (rare) in the distant part of the image can be appropriately included in the processing result.
- the second processing unit 123 includes a processing result of area recognition for the enlarged image corresponding to a distant portion of the image and a processing result of area recognition for the acquired image (first processing). By comparing with the result), the priority of the subject type is determined, and the subject type ID to be stored is determined based on the priority.
- the second processing unit 123 specifies the type of the subject included in the processing result ER.
- the second processing unit 123 specifies the type of the subject included in the first processing result generated by the first processing unit 121.
- the second processing unit 123 is a type of subject that is included in the processing result ER and is not included in the first processing result, or a type of subject whose ratio in the first processing result is equal to or less than a predetermined reference. Is specified, and the type of the subject is determined as the type of the subject to be prioritized.
- the processing result ER includes "person" as the subject type and the first processing result does not include "person”
- the second processing unit 123 sets "person" as the priority subject type. Can be determined.
- the second processing unit 123 may determine the priority of the subject type by any other method or format. For example, the second processing unit 123 identifies a subject that is included in the processing result ER and whose ratio in the first processing result is larger than a predetermined standard, and determines a low priority for the type of the subject. You may.
- the second processing unit 123 After determining the priority of the subject type, the second processing unit 123 determines the subject type ID to be stored based on the priority of the subject type by using the same method as in (B) above.
- the processing result of the area recognition for the enlarged image corresponding to the distant portion of the image and the processing result of the area recognition for the acquired image (first processing result).
- the second processing unit 123 is not sufficiently recognized by the area recognition for the acquired image, for example, but is recognized by the area recognition for the enlarged image.
- the subject can be appropriately included in the processing result.
- the second processing unit 123 reduces the processing result as described above and generates the second processing result which is the reduced processing result.
- the second processing unit 123 cuts out a distant portion of the image from the image acquired by the image acquisition unit 110 by using the distant identification information generated by the distant identification unit 122. Then, the second processing unit 123 performs image processing on the cut out image by applying settings different from those of the first image processing.
- the second processing unit 123 can use a setting that can recognize the type of the subject even with a smaller number of pixels as a different setting.
- the first processing unit 121 performs area recognition in the first image processing using a setting in which a group of eight pixels is the smallest unit for recognizing the type of the subject
- the second processing unit 123 may perform area recognition.
- area recognition may be performed using a setting in which a group of four pixels is the minimum unit for recognizing the type of the subject.
- the second processing unit 123 can use a setting that can recognize different types of subjects as different settings.
- a setting that can recognize six types of subjects human, automobile, road, other, building, and sky.
- the second processing unit 123 performs area recognition using a setting that can recognize eight types of subjects, such as people, automobiles, roads, others, buildings, and the sky, as well as signs and traffic lights. You may go.
- such a setting is useful when there is a high need to recognize a distant sign or traffic light as compared to a nearby sign or traffic light that is likely to be already visible.
- an image processing engine a processing component that performs substantive processing on an image.
- the image processing engine includes a recognition model generated by learning or the like.
- the image processing engine may be implemented by software or hardware.
- the application of different settings can be done, for example, by having the image processing engine read the settings at runtime.
- the image processing device 10 causes the image processing engine included in the image processing device 10 to read the settings as described above at the time of starting the image processing device 10 or at an arbitrary timing during execution. Then, the second processing unit 123 of the image processing device 10 performs the second image processing by using the image processing engine from which the settings are read in this way.
- the application of different settings can be performed, for example, at the time of creating an image processing engine.
- the creator of the image processing engine designs and creates the image processing engine so that the settings as described above are used.
- the second processing unit 123 of the image processing device 10 performs the second image processing using the image processing engine created in this way.
- the second processing unit 123 applies different settings as described above to execute the image processing and generate the second processing result.
- the compositing unit 124 of the image processing apparatus 10 synthesizes the first processing result generated by the first processing unit 121 and the second processing result generated by the second processing unit 123, and synthesizes the first processing result.
- the resulting synthesis processing result is generated (step S150 in FIG. 3).
- the synthesizing unit 124 replaces the subject type ID of each pixel corresponding to the distant portion of the image among the pixels of the first processing result with the subject type ID of each pixel of the second processing result.
- the synthesis unit 124 is described above. By the substitution described, the synthesis processing result as shown in FIG. 15 is generated.
- the portion corresponding to the second processing result (the portion corresponding to the distant portion of the image) in the composition processing result is shown by a rectangle. It is arbitrary whether or not the synthesis unit 124 includes information representing a portion corresponding to the second processing result, such as this rectangle, in the synthesis processing result. For example, when it is not necessary to display the composition processing result, the composition unit 124 does not have to include the information representing the portion corresponding to the second processing result such as the above rectangle in the composition processing result. ..
- the compositing unit 124 integrates the processing result corresponding to the distant portion of the image and the second processing result in the first processing result.
- the compositing unit 124 may use the methods shown in the following (E) to (F) as the above-mentioned integration method (E) integration based on the priority of the subject type (F) of the subject type. Integration based on reliability The following describes (E) to (F) in detail.
- the synthesis unit 124 integrates the subject type IDs based on the priority of the subject types. For example, in the first processing result, the subject type ID stored in a certain pixel corresponding to the distant portion of the image is 5 (road), and the subject type ID of the corresponding pixel of the second processing result is 1 (). (People), and if "person" is defined as the type of subject to be prioritized, the compositing unit 124 can determine the subject type ID of the pixel to be 1 (person). On the contrary, when "road” is defined as the priority subject type, the compositing unit 124 can determine the subject type ID of the pixel to 5 (road). By doing so, the compositing unit 124 can appropriately select and provide the required subject type from the subject types recognized by the first image processing and the second image processing. It becomes.
- the integration of the subject type ID based on the priority of the subject type can be performed by any other method.
- the compositing unit 124 statically or dynamically determines the priority of the subject type by any other method or format, and the priority is given.
- Subject IDs may be integrated based on the degree.
- the first processing result and the second processing result include the image in which the subject type ID is stored in each pixel and the reliability (for example, minimum 0.0 to maximum 1.0) for each pixel. What to do.
- the subject type ID stored in a certain pixel corresponding to the distant portion of the image is 5 (road), and the reliability thereof is 0.4, which corresponds to the second processing result.
- the compositing unit 124 determines the subject type ID of the pixel to be 1 (person) having a higher reliability.
- the compositing unit 124 integrates the subject type ID based on the reliability of the subject type and the predetermined weights of the first processing result and the second processing result.
- the subject type ID stored in a certain pixel corresponding to the distant portion of the image is 5 (road), and the reliability thereof is 0.8, which corresponds to the second processing result.
- the weighting coefficient of the first processing result is 0.5
- the weighting coefficient of the second processing result is 0.5.
- the synthesizing unit 124 calculates an evaluation value by multiplying each of the pixel of the first processing result and the pixel of the second processing result by the reliability and the weighting coefficient. In the above case, the synthesis unit 124 calculates 0.4 as the evaluation value of the pixel of the first processing result and 0.7 as the evaluation value of the pixel of the second processing result. Next, the synthesis unit 124 compares the comparison of the calculated evaluation values and determines the subject type ID of the pixel. In the above case, 1 (person) of the second processing result having a larger evaluation value is determined as the subject identification ID of the pixel.
- the synthesis unit 124 uses other reliability formats, an evaluation value calculation method, a subject identification ID, and the like, and uses a subject based on the reliability of the subject type.
- the type ID can be integrated.
- the scene recognition unit 130 of the image processing device 10 is generated by the first processing unit 121, the second processing result generated by the second processing unit 123, and the synthesis unit 124.
- Scene recognition is performed based on at least one of the combined processing results, and a scene recognition result is generated (step S160 in FIG. 3).
- the scene recognition is a process of recognizing the meaning of the scene represented in the image.
- various things can be used.
- the scene recognition unit 130 can perform a process of recognizing the road condition as scene recognition.
- a specific example of the process of recognizing such a road condition will be described with reference to FIGS. 16 and 17.
- the scene recognition unit 130 uses a process for determining whether or not a scene type such as "with a pedestrian crossing in front” or "with a pedestrian in front” is true or false as scene recognition. be able to.
- FIG. 16 is a diagram showing an example of a scene recognition result by such scene recognition.
- the scene recognition unit 130 can use a process for determining the numerical value of the scene type such as "distance to the intersection ahead" and "number of lanes in the traveling direction" for scene recognition.
- FIG. 17 is a diagram showing an example of a scene recognition result by such scene recognition.
- the scene recognition unit 130 may determine the position of the subject in the image (for example, the position of the pedestrian crossing or the position of the intersection) together with the truth value and the numerical value as described above, and may include it in the scene recognition result.
- the scene recognition unit 130 can perform the above-mentioned scene recognition by using a predetermined recognition model.
- the scene recognition unit 130 may use a recognition model created by any method.
- the scene recognition unit 130 uses a recognition model created by performing deep learning or learning by other known methods using teacher data in which the correct answer label of the scene is associated with the synthesis processing result. You may perform scene recognition.
- the output unit 140 of the image processing device 10 is generated by the first processing unit 121, the second processing result generated by the second processing unit 123, and the synthesis unit 124.
- a predetermined output is performed based on at least one of the combined processing result and the scene recognition result generated by the scene recognition unit 130. (Step S170 in FIG. 3).
- the output unit 140 can output using one or more of the following specific examples.
- the image acquired by the image acquisition unit 110 is an image including a road. Further, it is assumed that the image processing device 10 and the vehicle traveling on the above road are communicably connected to each other. In the first example of the output, the output unit 140 provides information as an output to the occupants of the vehicle traveling on the road.
- the output unit 140 can give an instruction to display as an output on a display device installed in the vehicle traveling on the road.
- the output unit 140 is instructed to display the image as shown in FIG. 15 on the above display device. May be done. Further, the output unit 140 shows a subject (for example, a person or a car) to be watched on the acquired image as shown in FIG. 4 based on the processing result of the area recognition as shown in FIG. You may perform processing to emphasize the area and give an instruction to display the processed image on the above display device. Further, the output unit 140 may instruct the display device to display the character information of the subject type displayed on the image together with the image.
- a subject for example, a person or a car
- the output unit 140 has the character information "there is a pedestrian in front”. May be instructed to be displayed on the above display device.
- the output unit 140 can give an instruction to announce as an output through a voice output device installed in the vehicle traveling on the road.
- the output unit 140 outputs the character information "there is a pedestrian in front".
- a voice an instruction to announce through the above-mentioned voice output device may be given.
- the output unit 140 can provide information to the occupants of the vehicle traveling on the road in any other manner.
- the image acquired by the image acquisition unit 110 is an image including a road. Further, it is assumed that the image processing device 10 and the vehicle traveling on the above road are communicably connected to each other. In the second example of the output, the output unit 140 gives an instruction of driving control to the vehicle traveling on the road as an output.
- the output unit 140 can instruct the vehicle traveling on the road to brake, steer the steering wheel, and turn on or off the light.
- the output unit 140 brakes the vehicle traveling on the above road. You may give the instruction of.
- the output unit 140 can give an instruction for driving control to the vehicle traveling on the road in any other manner.
- the vehicle traveling on the above road may be an automatically driven vehicle or a manually driven vehicle.
- the output unit 140 provides information to the manager.
- the manager includes various persons such as a vehicle manager, a road manager and a guard, a manager and a guard of other facilities, and the like. It is assumed that the manager uses the terminal device, and that the image processing device 10 and the terminal device are communicably connected to each other.
- the terminal device used by the manager may be installed close to the image processing device 10, may be installed remotely from the image processing device 10, or may be a portable terminal device. There may be.
- the output unit 140 can instruct the terminal device of the manager to present the first processing result, the second processing result, the synthesis processing result, and the scene recognition result as an output.
- the output unit 140 gives an instruction or a voice to display the first processing result, the second processing result, the synthesis processing result, and the scene recognition result in the same manner as the mode described in the first example of the output. You may give an instruction to announce as.
- the output unit 140 transmits information to an external device (not shown).
- the external device includes various devices such as a display device, a storage device, and an analysis device. It is assumed that the image processing device 10 and the external device are communicably connected to each other.
- the output unit 140 transmits, for example, information such as a first processing result, a second processing result, a synthesis processing result, and a scene recognition result to such an external device.
- the external device can perform various processes such as displaying the received information on the screen, accumulating the received information, and further analyzing based on the received information.
- the order of the processes shown in FIGS. 4, 6 and 12 is an example, and the order may be changed or some processes may be performed in parallel as long as the process results do not change.
- the image processing apparatus 10 may change the order of the processing of step S120 in FIG. 4 and the series of processing of steps S130 and S140, or may perform some processing in parallel.
- ⁇ Explanation of effect> when image processing is performed on a captured image, a distant subject can be recognized with high accuracy.
- the reason is that the distant identification unit 122 identifies the distant portion of the captured image, and the second processing unit performs a predetermined second image processing on the distant portion of the identified image.
- the distant identification unit 122 identifies the distant portion of the captured image, and the second processing unit performs a predetermined second image processing on the distant portion of the identified image. This is because it is not necessary to perform the second image processing up to the range excluding the distant portion.
- FIG. 18 is a diagram showing a functional block of the image processing device 10 in the second embodiment.
- the image processing apparatus 10 in the second embodiment is different from the first embodiment in that it includes a depth data acquisition unit 150.
- the other components of the second embodiment are the same as those of the first embodiment.
- the same reference numerals as those in FIGS. 1 and 2 are used, and detailed description thereof will be omitted.
- the depth data acquisition unit 150 functions as a means for acquiring depth data.
- the depth data is data representing the depth with respect to the object.
- Depth data is represented, for example, in the form of an image (called a depth image) in which the distance to an object is stored in each pixel.
- a depth image an image in which the distance to an object is stored in each pixel.
- the depth data acquisition unit 150 may acquire the measured depth data from the measuring device by communication or the like.
- measuring devices include LIDAR (Light Detection and Ringing, Laser Imaging Detection and Ringing), millimeter-wave radar, stereo camera, and ToF (Time of Flyht) camera.
- the depth data acquisition unit 150 may generate depth data using the image acquired by the image acquisition unit 110.
- a method called depth estimation which estimates the depth from a two-dimensional image using deep learning or the like, is known, and the depth data acquisition unit 150 can generate depth data using such depth estimation. can.
- the depth data acquisition unit 150 may acquire the depth data generated from the processing device by communication or the like.
- FIG. 19 is a flowchart showing the operation of the image processing device 10 in the second embodiment.
- the same operations as those in the first embodiment are designated by the same reference numerals as those in FIG. 3, and detailed description thereof will be omitted.
- the image acquisition unit 110 of the image processing device 10 acquires a captured image from the image pickup device 20 (step S110 in FIG. 19).
- the first processing unit of the image processing device 10 performs predetermined first image processing on the image acquired by the image acquisition unit 110 to generate a first processing result (step S120 in FIG. 19).
- the depth data acquisition unit 150 of the image processing apparatus 10 acquires depth data by using the method as described above (step S180 in FIG. 19).
- the depth data acquisition unit 150 acquires a depth image as shown in FIG. 20 as depth data.
- the depth is represented by shading. The darker the part, the smaller the depth (near), and the brighter the part, the larger the depth (far).
- the distant identification unit 122 of the image processing device 10 identifies the distant portion based on the depth data acquired by the depth data acquisition unit 150, and generates distant identification information (step S190 in FIG. 19).
- the distant identification unit 122 can identify the distant portion based on the depth data. A specific example will be described below.
- FIG. 21 is a flowchart showing the operation of the distant identification unit 122 based on the depth data.
- the distant identification unit 122 identifies a distant portion in the coordinate system of the depth data based on the depth data acquired by the depth data acquisition unit 150 (step S191 in FIG. 21).
- the distant identification unit 122 extracts the point having the largest depth from the points included in the depth data, identifies a predetermined portion including the point, and sets this as the distant portion in the coordinate system of the depth data. Can be done. Further, for example, the distant identification unit 122 extracts a point cloud whose depth is equal to or higher than a predetermined threshold value among the points included in the depth data, identifies a predetermined portion including the extracted point cloud, and determines the depth. It can be a distant part of the data coordinate system.
- the shape of the distant portion in the coordinate system of the depth data may be, for example, a rectangle, a polygon other than the rectangle, a circle, an ellipse, or any other shape.
- the distant identification unit 122 converts the distant portion in the coordinate system of the specified depth data into the distant portion in the coordinate system of the image (step S192 in FIG. 21).
- the process of obtaining the conversion formula between the coordinate system of the depth data and the coordinate system of the image is called calibration.
- various methods are known, for example, the above conversion formula is obtained based on a small number of points in the coordinate system of depth data and a small number of points in the coordinate system of the corresponding image. Has been done.
- the distant identification unit 122 converts the distant part in the coordinate system of the depth data into the distant part in the coordinate system of the image by using, for example, the conversion formulas obtained by such various methods.
- the distant identification unit 122 corrects the distant portion in the coordinate system of the image obtained by the transformation. (Step S193 in FIG. 21).
- the distant identification unit 122 may make a correction to approximate the shape to a rectangle.
- the present invention is not limited to this, and the distant identification portion 122 may be corrected to approximate a polygon, a circle, an ellipse, or another shape other than a rectangle, or may be corrected to various other shapes.
- the distant identification unit 122 may omit the correction process when it is not necessary to correct the distant part in the coordinate system of the image obtained by the conversion.
- FIG. 22 is a flowchart showing the operation of the distant identification unit 122 based on the depth data.
- the distant identification unit 122 generates a plurality of candidates for the distant portion based on the depth data acquired by the depth data acquisition unit 150 (step S196 in FIG. 22).
- the distant identification unit 122 identifies a plurality of candidates for the distant portion in the coordinate system of the depth data based on the depth data acquired by the depth data acquisition unit 150.
- the distant identification unit 122 first extracts a point cloud whose depth is equal to or higher than a predetermined threshold value among the points included in the depth data. Next, the distant identification unit 122 divides the extracted point cloud into a group of points having a short distance in the coordinate system of the depth data. Subsequently, the distant identification unit 122 identifies a portion including a point cloud included in the group for each of the plurality of divided groups. The distant identification unit 122 can make these plurality of parts a plurality of candidates for the distant part in the coordinate system of the depth data.
- the distant identification unit 122 converts each of the plurality of candidates for the distant portion in the coordinate system of the depth data specified above into the distant portion in the coordinate system of the image.
- various methods as described in the first example of distant identification based on depth data can be used.
- the distant identification unit 122 corrects each of the plurality of candidates for the distant portion in the coordinate system of the image obtained by the above conversion.
- various methods as described in the first example of distant identification based on depth data can be used.
- the distant identification unit 122 may omit the correction process when it is not necessary to correct the candidate of the distant part in the coordinate system of the image obtained by the conversion.
- the distant identification unit 122 can obtain a plurality of candidates for the distant portion based on the depth data acquired by the depth data acquisition unit 150.
- the distant identification unit 122 identifies the distant portion from the plurality of candidates for the distant portion generated by the above based on the image acquired by the image acquisition unit 110 (step S197 in FIG. 22).
- the distant identification unit 122 performs image recognition on the image acquired by the image acquisition unit 110.
- the image recognition performed by the distant identification unit 122 is, for example, area recognition.
- the present invention is not limited to this, and the image recognition may be object recognition or other image recognition.
- the remote identification unit 122 may use simple area recognition as described in the first example of vanishing point estimation in the first embodiment.
- the distant identification unit 122 calculates an evaluation value for each of the plurality of candidates in the distant portion based on the above-mentioned image recognition processing result. For example, when the distant identification unit 122 includes a larger range recognized as a predetermined subject type (for example, a road) as a result of image recognition processing, a larger evaluation value is given to the candidate in the distant portion. Can be given.
- a predetermined subject type for example, a road
- the distant identification unit 122 identifies the candidate having the largest calculated evaluation value as the distant portion.
- the distant identification unit 122 After specifying the distant portion as described above, the distant identification unit 122 generates distant identification information expressing the distant portion in a predetermined format.
- the second processing unit 123 of the image processing device 10 makes a first image with respect to the distant portion of the image acquired by the image acquisition unit 110 based on the distant identification information generated by the distant identification unit 122. A predetermined second image processing different from the processing is performed, and the second processing result is generated (step S140 in FIG. 19).
- the compositing unit 124 of the image processing apparatus 10 synthesizes the first processing result generated by the first processing unit 121 and the second processing result generated by the second processing unit 123, and synthesizes the first processing result.
- the resulting synthesis processing result is generated (step S150 in FIG. 19).
- the scene recognition unit 130 of the image processing device 10 is generated by the first processing unit 121, the second processing result generated by the second processing unit 123, and the synthesis unit 124.
- Scene recognition is performed based on at least one of the combined processing results, and a scene recognition result is generated (step S160 in FIG. 19).
- the output unit 140 of the image processing device 10 is generated by the first processing unit 121, the second processing result generated by the second processing unit 123, and the synthesis unit 124.
- a predetermined output is performed based on at least one of the combined processing result and the scene recognition result generated by the scene recognition unit 130. (Step S170 in FIG. 19).
- the order of the processes shown in FIGS. 19, 21 and 22 is an example, and the order may be changed or some processes may be performed in parallel as long as the process results do not change.
- the image processing apparatus 10 may change the order of the processing of step S120 in FIG. 19 and the series of processing of steps S180, S190, and S140, or may perform some processing in parallel.
- the depth data acquisition unit 150 acquires the depth data
- the distant identification unit 122 identifies the distant portion of the captured image based on the acquired depth data.
- the configuration of the third embodiment will be described.
- the configuration of the third embodiment is the minimum configuration in each embodiment.
- FIG. 23 is a diagram showing a functional block of the image processing device 10 in the third embodiment.
- the image processing apparatus 10 includes an image acquisition unit 110, an image processing unit 120, and an output unit 140.
- the image processing unit 120 further includes a first processing unit 121, a distant identification unit 122, and a second processing unit 123.
- Each component of the image processing device 10 functions as a means similar to the corresponding component in the image processing device 10 or the like of the first embodiment.
- FIG. 24 is a flowchart showing the operation of the image processing device 10 in the third embodiment.
- the same operations as those in the first embodiment are designated by the same reference numerals as those in FIG. 3, and detailed description thereof will be omitted.
- the image acquisition unit 110 of the image processing device 10 acquires a captured image from the image pickup device 20 (step S110 in FIG. 24).
- the first processing unit of the image processing device 10 performs predetermined first image processing on the image acquired by the image acquisition unit 110 to generate a first processing result (step S120 in FIG. 24).
- the distant identification unit 122 of the image processing device 10 identifies the distant portion based on the image acquired by the image acquisition unit 110 (step S130 in FIG. 24).
- the second processing unit 123 of the image processing device 10 is different from the first image processing with respect to the distant portion of the image acquired by the image acquisition unit 110 based on the distant identification information generated by the distant identification unit 122.
- a predetermined second image processing is performed to generate a second processing result (step S140 in FIG. 24).
- the output unit 140 of the image processing unit 10 outputs a predetermined output based on the first processing result generated by the first processing unit 121 and the second processing result generated by the second processing unit 123 (FIG. FIG. Step 3 S170).
- the order of the processes shown in FIG. 24 is an example, and the order may be changed or some processes may be performed in parallel as long as the process results do not change.
- the image processing apparatus 10 may change the order of the processing of step S120 in FIG. 24 and the series of processing of steps S130 and S140, or may perform some processing in parallel.
- the processing unit 121 and the second processing unit 123 may perform image recognition other than area recognition as image processing.
- image recognition object recognition for estimating a shape (for example, a rectangle) surrounding a subject included in an image and the type of the subject is known, and the first processing unit 121 and the second processing unit 123 are images. As a process, this object recognition may be performed.
- the first processing unit 121 and the second processing unit 123 of the image processing apparatus 10 may perform image processing other than image recognition, for example, image conversion / processing, as image processing.
- image conversion / processing super-resolution that generates a high-definition image by using pixel interpolation or the like is known, and the first processing unit 121 and the second processing unit 123 perform this as image processing. Super-resolution may be performed. As a result, high-quality processing results can be obtained for the distant portion.
- the image processing apparatus 10 when the first processing result satisfies a predetermined condition, the image processing apparatus 10 omits the specific processing of the distant portion by the distant identification unit 122 and the second image processing by the second processing unit 123. You may.
- the first processing unit 121 performs image recognition such as area recognition and object recognition as the first image processing, and generates the reliability together with the image recognition result as the first processing result, the image processing apparatus. If the reliability is sufficiently high, the 10 may omit the specific processing of the distant portion by the distant identification unit 122 and the second image processing by the second processing unit 123. By omitting the processing in this way, the processing load in the distant identification unit 122 and the second processing unit 123 can be reduced.
- the image processing device 10 has been described as acquiring an image from the image pickup device 20, but the image processing device 10 is not limited to this, and the image processing device 10 may be stored in a storage device or a recording medium in advance. The recorded image may be acquired and image processing may be performed on the image (offline image processing). As an example of such offline image processing, the image processing device 10 may perform image processing on pre-recorded images before and after the occurrence of an accident on the road. As a result of such image processing, the image processing apparatus 10 may provide information useful for detailed analysis of an accident after the fact, for example, a person passing through a distant pedestrian crossing before the accident occurred. can.
- the image processing device 10 has been described as specifying a distant portion of the captured image and performing a predetermined second image processing on the distant portion.
- the part to be watched may be specified by using a criterion other than "far".
- the distant identification unit 122 identifies a portion of the captured image to be watched using a reference other than "far”, and the second processing unit 123 is assigned to the portion to be watched.
- a predetermined second image processing may be performed.
- each component of the image processing apparatus 10 represents a functional block. A part or all of each component of the image processing apparatus 10 may be realized by any combination of the computer 1000 and the program.
- FIG. 25 is a block diagram showing an example of the hardware configuration of the computer 1000.
- the computer 1000 may include, for example, a processor 1001, a ROM (Read Only Memory) 1002, a RAM (Random Access Memory) 1003, a program 1004, a storage device 1005, a drive device 1007, a communication interface 1008, an input device 1009, and the like. It includes an output device 1010, an input / output interface 1011 and a bus 1012.
- Program 1004 includes instructions for realizing each function of each device.
- the program 1004 is stored in the ROM 1002, the RAM 1003, and the storage device 1005 in advance.
- the processor 1001 realizes each function of each device by executing the instruction included in the program 1004.
- the processor 1001 of the image processing device 10 realizes the functions of the image acquisition unit 110, the image processing unit 120, and the like by executing the instructions included in the program 1004.
- the drive device 1007 reads and writes the recording medium 1006.
- Communication interface 1008 provides an interface with a communication network.
- the input device 1009 is, for example, a mouse, a keyboard, or the like, and receives input of information from an operator or the like.
- the output device 1010 is, for example, a display, and outputs (displays) information to an operator or the like.
- the input / output interface 1011 provides an interface with peripheral devices. Bus 1012 connects each component of these hardware.
- the program 1004 may be supplied to the processor 1001 via a communication network, or may be stored in the recording medium 1006 in advance, read by the drive device 1007, and supplied to the processor 1001.
- FIG. 25 is an example, and components other than these may be added, or some components may not be included.
- the image processing device 10 may be realized by any combination of a computer and a program that are different for each component. Further, a plurality of components included in each device may be realized by any combination of one computer and a program.
- each component of each device may be realized by a general-purpose or dedicated circuit or a combination thereof. These circuits may be composed of a single chip or a plurality of chips connected via a bus. A part or all of each component of each device may be realized by the combination of the circuit or the like and the program described above.
- each component of each device when a part or all of each component of each device is realized by a plurality of computers, circuits, etc., the plurality of computers, circuits, etc. may be centrally arranged or distributed.
- Appendix 1 An image acquisition means for acquiring an image taken by an image pickup device, and A first processing means for performing a first image processing on the image, A distant identification means for specifying a distant part of the image, and A second processing means that performs a second image processing different from the first image processing on a distant portion of the image, An image processing apparatus including an output means for outputting a processing result of the first image processing and an output based on the processing result of the second image processing.
- Appendix 2 Further comprising a synthesizing means for synthesizing the processing result of the first image processing and the processing result of the second image processing.
- the image processing apparatus according to Appendix 1.
- the second processing means As the second image processing, the second processing means enlarges a distant portion of the image, performs a predetermined process on the enlarged image, and reduces the processing result of the predetermined process. 2.
- the image processing apparatus according to 2. (Appendix 4)
- the first processing means performs area recognition as the first image processing, and performs area recognition.
- a compositing means for synthesizing the processing result of the first image processing and the processing result of the second image processing is provided.
- the synthesizing means is based on the type of the subject estimated in the first image processing, the type of the subject estimated in the second image processing, and the priority of the predetermined type of the subject. Determining the type of subject in each region of the distant portion of the image.
- the image processing apparatus according to Appendix 4. As the second image processing, the second processing means performs image processing on a distant portion of the image by applying settings different from those of the first image processing.
- the image is an image including a road, and the distant identification means estimates the vanishing point of the road in the image, and specifies a predetermined portion based on the vanishing point of the road as a distant portion of the image.
- the image processing apparatus according to any one of Supplementary note 1 to 6.
- Appendix 8 The image processing apparatus according to Appendix 7, wherein the distant identification means estimates a vanishing point of a road in the image based on a region determined to be a road in the processing result of area recognition for the image.
- Appendix 9 Further equipped with a depth data acquisition means to acquire depth data, The distant identification means identifies a distant portion in the image based on the depth data.
- the image processing apparatus according to any one of Supplementary note 1 to 6.
- Appendix 10 A scene recognition means for recognizing a road condition based on the processing result of the first image processing and the processing result of the second image processing is further provided.
- the image processing apparatus according to any one of Supplementary note 1 to 9.
- the output means outputs a predetermined output to a vehicle traveling on a road based on the processing result of the first image processing and the processing result of the second image processing.
- the image processing apparatus according to any one of Supplementary note 1 to 10.
- the image processing device Acquire the image taken by the image pickup device and The first image processing is performed on the image, and the image is processed. Identify the distant part of the image and A second image process different from the first image process is performed on the distant portion of the image. Output based on the processing result of the first image processing and the processing result of the second image processing.
- Image processing method On the computer Acquire the image taken by the image pickup device and The first image processing is performed on the image, and the image is processed.
- Image processing device 20 Image pickup device 110 Image acquisition unit 120 Image processing unit 121 First processing unit 122 Remote identification unit 123 Second processing unit 124 Synthesis unit 130 Scene recognition unit 140 Output unit 150 Depth data acquisition unit 1000 Computer 1001 Processor 1002 ROM 1003 RAM 1004 Program 1005 Storage device 1006 Recording medium 1007 Drive device 1008 Communication interface 1009 Input device 1010 Output device 1011 Input / output interface 1012 Bus
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Multimedia (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Mechanical Engineering (AREA)
- Signal Processing (AREA)
- Image Processing (AREA)
- Image Analysis (AREA)
- Traffic Control Systems (AREA)
Abstract
Description
第1の実施形態について説明する。
まず、第1の実施形態の構成について説明する。
次に、第1の実施形態の動作について説明する。
遠方特定の第1の例について説明する。図6は、遠方特定部122の動作を示すフローチャートである。なお、遠方特定の第1の例において、画像取得部110によって取得された画像は、道路を含む画像である。
消失点推定の第1の例について説明する。
消失点推定の第2の例について説明する。
遠方部分決定の第1の例について説明する。
遠方部分決定の第2の例について説明する。
遠方特定の第2の例について説明する。
遠方特定の第3の例について説明する。
遠方特定のその他の例について説明する。
第2の画像処理の第1の例について説明する。
図12は、第2の画像処理の第1の例の動作を示す図である。第2の画像処理の第1の例では、第2処理部123は、画像の遠方部分を拡大し、拡大された画像に対して所定の処理を行い、当該所定の処理の処理結果を縮小する。
(A)最頻の被写体種別IDを用いる
(B)予め定められた被写体の種別の優先度に基づいて、被写体種別IDを決定
(C)遠方部分に対する領域認識の処理結果から被写体の種別の優先度を定め、当該優先度に基づいて被写体種別IDを決定
(D)遠方部分に対する領域認識と取得された画像に対する領域認識との処理結果の比較から被写体の種別の優先度を定め、当該優先度に基づいて被写体種別IDを決定
以下に、(A)~(D)について、詳細に説明する。
第2の画像処理の第2の例について説明する。
合成の第1の例について説明する。合成の第1の例においては、合成部124は、第1の処理結果のうち、画像の遠方部分に相当する処理結果を、第2の処理結果で置換する。
合成の第2の例について説明する。合成の第2の例においては、合成部124は、第1の処理結果のうち、画像の遠方部分に相当する処理結果と、第2の処理結果とを統合する。
(E)被写体の種別の優先度に基づいて統合
(F)被写体の種別の信頼度に基づいて統合
以下に、(E)~(F)について詳細に説明する。
出力の第1の例について説明する。出力の第1の例において、画像取得部110によって取得された画像は、道路を含む画像である。また、画像処理装置10と上記の道路を走行する車両は通信可能に接続されているものとする。出力の第1の例において、出力部140は、出力として、上記の道路を走行する車両の乗員に対する情報の提供を行う。
出力の第2の例について説明する。出力の第2の例において、画像取得部110によって取得された画像は、道路を含む画像である。また、画像処理装置10と上記の道路を走行する車両は通信可能に接続されているものとする。出力の第2の例において、出力部140は、出力として、上記の道路を走行する車両に対する運転制御の指示を行う。
出力の第3の例について説明する。出力の第3の例において、出力部140は、管理員に対する情報の提供を行う。ここで、管理員とは、車両の管理者、道路の管理者や監視員、その他の施設の管理者や監視員等、種々の人物を含む。管理員は端末装置を使用するものとし、画像処理装置10と当該端末装置は、通信可能に接続されているものとする。上記の管理員の使用する端末装置は、画像処理装置10に近接して設置されていてもよいし、画像処理装置10の遠隔に設置されていてもよいし、また、携帯可能な端末装置であってもよい。
出力の第4の例について説明する。
第1の実施形態によれば、撮影された画像に対して画像処理を行う場合において、遠方の被写体を精度よく認識することができる。その理由は、遠方特定部122が、撮影された画像の遠方部分を特定し、第2処理部が、特定された画像の遠方部分に対して所定の第2の画像処理を行うためである。
第2の実施形態について説明する。
まず、第2の実施形態の構成について説明する。
次に、第2の実施形態の動作について説明する。
深度データに基づく遠方特定の第1の例について説明する。図21は、深度データに基づく、遠方特定部122の動作を示すフローチャートである。
例えば、遠方特定部122は、深度データに含まれる点のうち、深度の最も大きい点を抽出し、その点を含む所定の部分を特定し、これを深度データの座標系における遠方部分とすることができる。また、例えば、遠方特定部122は、深度データに含まれる点のうち、深度が所定の閾値以上である点群を抽出し、抽出された点群を含む所定の部分を特定し、これを深度データの座標系における遠方部分とすることができる。なお、深度データの座標系における遠方部分の形状は、例えば、矩形であってもよいし、矩形以外の多角形、円、楕円、その他の形状であってもよい。
深度データに基づく遠方特定の第2の例について説明する。深度データに基づく遠方特定の第2の例では、深度データに加えて、画像取得部110によって取得された画像に基づいて、遠方部分を特定する。図22は、深度データに基づく、遠方特定部122の動作を示すフローチャートである。
第2の実施形態によれば、撮影された画像に対して画像処理を行う場合において、精度よく遠方部分を特定することができる。その理由は、深度データ取得部150が、深度データを取得し、遠方特定部122が、取得された深度データに基づいて、撮影された画像の遠方部分を特定するためである。
第3の実施形態について説明する。
第3の実施形態の構成について説明する。第3の実施形態の構成は、各実施形態における最小構成である。
第3の実施形態の動作について説明する。
第3の実施形態によれば、撮影された画像に対して画像処理を行う場合において、遠方の被写体を精度よく認識することができる。
以上、本発明の各実施形態を説明したが、本発明は、上記の各実施形態に限定されるものではなく、本発明の基本的な技術的思想を逸脱しない範囲で、更なる変形・置換・調整を加えることができる。
上記で説明した各実施形態において、画像処理装置10の各構成要素は、機能ブロックを示している。画像処理装置10の各構成要素の一部又は全部は、コンピュータ1000とプログラムとの任意の組み合わせにより実現されてもよい。
(付記1)
撮像装置で撮影された画像を取得する画像取得手段と、
前記画像に対して第1の画像処理を行う第1処理手段と、
前記画像の遠方部分を特定する遠方特定手段と、
前記画像の遠方部分に対して、第1の画像処理と異なる第2の画像処理を行う第2処理手段と、
前記第1の画像処理の処理結果及び前記第2の画像処理の処理結果に基づく出力を行う出力手段と、を備える
画像処理装置。
(付記2)
前記第1の画像処理の処理結果と前記第2の画像処理の処理結果とを合成する合成手段をさらに備える、
付記1に記載の画像処理装置。
(付記3)
前記第2処理手段は、前記第2の画像処理として、前記画像の遠方部分を拡大し、拡大された画像に対して所定の処理を行い、当該所定の処理の処理結果を縮小する
付記1又は2に記載の画像処理装置。
(付記4)
前記第1処理手段は、前記第1の画像処理として領域認識を行い、
前記第2処理手段は、前記第2の画像処理における前記所定の処理として領域認識を行う
付記3に記載の画像処理装置。
(付記5)
前記第1の画像処理の処理結果と前記第2の画像処理の処理結果とを合成する合成手段を備え、
前記合成手段は、前記第1の画像処理において推定された被写体の種別と、前記第2の画像処理において推定された被写体の種別と、予め定められた被写体の種別の優先度とに基づいて、前記画像の遠方部分の各領域における被写体の種別を決定する、
付記4に記載の画像処理装置。
(付記6)
前記第2処理手段は、前記第2の画像処理として、前記画像の遠方部分に対して、第1の画像処理とは異なる設定を適用した画像処理を行う、
付記1又は2に記載の画像処理装置。
(付記7)
前記画像は道路を含む画像であって
前記遠方特定手段は、前記画像における道路の消失点を推定し、当該道路の消失点を基準する所定の部分を、前記画像の遠方部分として特定する、
付記1乃至6のいずれか一項に記載の画像処理装置。
(付記8)
前記遠方特定手段は、前記画像に対する領域認識の処理結果において、道路であると判定された領域に基づいて、前記画像における道路の消失点を推定する
付記7に記載の画像処理装置。
(付記9)
深度データを取得する深度データ取得手段をさらに備え、
前記遠方特定手段は、前記深度データに基づいて、前記画像における遠方部分を特定する、
付記1乃至6のいずれか一項に記載の画像処理装置。
(付記10)
前記第1の画像処理の処理結果及び前記第2の画像処理の処理結果に基づいて、道路の状況を認識するシーン認識手段をさらに備える、
付記1乃至9のいずれか一項に記載の画像処理装置。
(付記11)
前記出力手段は、前記第1の画像処理の処理結果及び前記第2の画像処理の処理結果に基づいて、道路を走行する車両に対して所定の出力を行う、
付記1乃至10のいずれか一項に記載の画像処理装置。
(付記12)
画像処理装置が、
撮像装置で撮影された画像を取得し、
前記画像に対して第1の画像処理を行い、
前記画像の遠方部分を特定し、
前記画像の遠方部分に対して、第1の画像処理と異なる第2の画像処理を行い、
前記第1の画像処理の処理結果及び前記第2の画像処理の処理結果に基づく出力を行う、
画像処理方法。
(付記13)
コンピュータに、
撮像装置で撮影された画像を取得し、
前記画像に対して第1の画像処理を行い、
前記画像の遠方部分を特定し、
前記画像の遠方部分に対して、第1の画像処理と異なる第2の画像処理を行い、
前記第1の画像処理の処理結果及び前記第2の画像処理の処理結果に基づく出力を行う処理を実行させる、
プログラムの記録媒体。
20 撮像装置
110 画像取得部
120 画像処理部
121 第1処理部
122 遠方特定部
123 第2処理部
124 合成部
130 シーン認識部
140 出力部
150 深度データ取得部
1000 コンピュータ
1001 プロセッサ
1002 ROM
1003 RAM
1004 プログラム
1005 記憶装置
1006 記録媒体
1007 ドライブ装置
1008 通信インタフェース
1009 入力装置
1010 出力装置
1011 入出力インタフェース
1012 バス
Claims (13)
- 撮像装置で撮影された画像を取得する画像取得手段と、
前記画像に対して第1の画像処理を行う第1処理手段と、
前記画像の遠方部分を特定する遠方特定手段と、
前記画像の遠方部分に対して、第1の画像処理と異なる第2の画像処理を行う第2処理手段と、
前記第1の画像処理の処理結果及び前記第2の画像処理の処理結果に基づく出力を行う出力手段と、を備える
画像処理装置。 - 前記第1の画像処理の処理結果と前記第2の画像処理の処理結果とを合成する合成手段をさらに備える、
請求項1に記載の画像処理装置。 - 前記第2処理手段は、前記第2の画像処理として、前記画像の遠方部分を拡大し、拡大された画像に対して所定の処理を行い、当該所定の処理の処理結果を縮小する
請求項1又は2に記載の画像処理装置。 - 前記第1処理手段は、前記第1の画像処理として領域認識を行い、
前記第2処理手段は、前記第2の画像処理における前記所定の処理として領域認識を行う
請求項3に記載の画像処理装置。 - 前記第1の画像処理の処理結果と前記第2の画像処理の処理結果とを合成する合成手段を備え、
前記合成手段は、前記第1の画像処理において推定された被写体の種別と、前記第2の画像処理において推定された被写体の種別と、予め定められた被写体の種別の優先度とに基づいて、前記画像の遠方部分の各領域における被写体の種別を決定する、
請求項4に記載の画像処理装置。 - 前記第2処理手段は、前記第2の画像処理として、前記画像の遠方部分に対して、第1の画像処理とは異なる設定を適用した画像処理を行う、
請求項1又は2に記載の画像処理装置。 - 前記画像は道路を含む画像であって
前記遠方特定手段は、前記画像における道路の消失点を推定し、当該道路の消失点を基準する所定の部分を、前記画像の遠方部分として特定する、
請求項1乃至6のいずれか一項に記載の画像処理装置。 - 前記遠方特定手段は、前記画像に対する領域認識の処理結果において、道路であると判定された領域に基づいて、前記画像における道路の消失点を推定する
請求項7に記載の画像処理装置。 - 深度データを取得する深度データ取得手段をさらに備え、
前記遠方特定手段は、前記深度データに基づいて、前記画像における遠方部分を特定する、
請求項1乃至6のいずれか一項に記載の画像処理装置。 - 前記第1の画像処理の処理結果及び前記第2の画像処理の処理結果に基づいて、道路の状況を認識するシーン認識手段をさらに備える、
請求項1乃至9のいずれか一項に記載の画像処理装置。 - 前記出力手段は、前記第1の画像処理の処理結果及び前記第2の画像処理の処理結果に基づいて、道路を走行する車両に対して所定の出力を行う、
請求項1乃至10のいずれか一項に記載の画像処理装置。 - 画像処理装置が、
撮像装置で撮影された画像を取得し、
前記画像に対して第1の画像処理を行い、
前記画像の遠方部分を特定し、
前記画像の遠方部分に対して、第1の画像処理と異なる第2の画像処理を行い、
前記第1の画像処理の処理結果及び前記第2の画像処理の処理結果に基づく出力を行う、
画像処理方法。 - コンピュータに、
撮像装置で撮影された画像を取得し、
前記画像に対して第1の画像処理を行い、
前記画像の遠方部分を特定し、
前記画像の遠方部分に対して、第1の画像処理と異なる第2の画像処理を行い、
前記第1の画像処理の処理結果及び前記第2の画像処理の処理結果に基づく出力を行う処理を実行させる、
プログラムの記録媒体。
Priority Applications (4)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
PCT/JP2021/000031 WO2022145054A1 (ja) | 2021-01-04 | 2021-01-04 | 画像処理装置、画像処理方法、及び記録媒体 |
EP21914916.8A EP4254944A4 (en) | 2021-01-04 | 2021-01-04 | IMAGE PROCESSING APPARATUS, IMAGE PROCESSING METHOD AND RECORDING MEDIUM |
US18/270,034 US20240071104A1 (en) | 2021-01-04 | 2021-01-04 | Image processing device, image processing method, and recording medium |
JP2022572879A JP7505596B2 (ja) | 2021-01-04 | 2021-01-04 | 画像処理装置、画像処理方法、及び画像処理プログラム |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
PCT/JP2021/000031 WO2022145054A1 (ja) | 2021-01-04 | 2021-01-04 | 画像処理装置、画像処理方法、及び記録媒体 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2022145054A1 true WO2022145054A1 (ja) | 2022-07-07 |
Family
ID=82259184
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/JP2021/000031 WO2022145054A1 (ja) | 2021-01-04 | 2021-01-04 | 画像処理装置、画像処理方法、及び記録媒体 |
Country Status (4)
Country | Link |
---|---|
US (1) | US20240071104A1 (ja) |
EP (1) | EP4254944A4 (ja) |
JP (1) | JP7505596B2 (ja) |
WO (1) | WO2022145054A1 (ja) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2024009519A1 (ja) * | 2022-07-08 | 2024-01-11 | 日本電信電話株式会社 | 認識装置、認識方法及び認識プログラム |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2004194169A (ja) * | 2002-12-13 | 2004-07-08 | Auto Network Gijutsu Kenkyusho:Kk | 視覚補助装置及び視覚補助方法 |
WO2012172713A1 (ja) * | 2011-06-13 | 2012-12-20 | 日産自動車株式会社 | 道路形状判定装置、車載用画像認識装置、撮像軸調整装置およびレーン認識方法 |
JP2013142972A (ja) * | 2012-01-10 | 2013-07-22 | Toyota Motor Corp | 走行路認識装置 |
JP2015149028A (ja) * | 2014-02-07 | 2015-08-20 | トヨタ自動車株式会社 | 区画線検出システム及び区画線検出方法 |
JP2015179368A (ja) * | 2014-03-19 | 2015-10-08 | 株式会社日本自動車部品総合研究所 | 走行区画線認識装置及び走行区画線認識プログラム |
JP2019014407A (ja) * | 2017-07-10 | 2019-01-31 | 株式会社Soken | 走路認識装置 |
Family Cites Families (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2004288100A (ja) * | 2003-03-25 | 2004-10-14 | Minolta Co Ltd | 撮像装置及び移動体カメラ |
WO2020027233A1 (ja) * | 2018-07-31 | 2020-02-06 | ソニーセミコンダクタソリューションズ株式会社 | 撮像装置及び車両制御システム |
-
2021
- 2021-01-04 EP EP21914916.8A patent/EP4254944A4/en active Pending
- 2021-01-04 JP JP2022572879A patent/JP7505596B2/ja active Active
- 2021-01-04 WO PCT/JP2021/000031 patent/WO2022145054A1/ja active Application Filing
- 2021-01-04 US US18/270,034 patent/US20240071104A1/en active Pending
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2004194169A (ja) * | 2002-12-13 | 2004-07-08 | Auto Network Gijutsu Kenkyusho:Kk | 視覚補助装置及び視覚補助方法 |
WO2012172713A1 (ja) * | 2011-06-13 | 2012-12-20 | 日産自動車株式会社 | 道路形状判定装置、車載用画像認識装置、撮像軸調整装置およびレーン認識方法 |
JP2013142972A (ja) * | 2012-01-10 | 2013-07-22 | Toyota Motor Corp | 走行路認識装置 |
JP2015149028A (ja) * | 2014-02-07 | 2015-08-20 | トヨタ自動車株式会社 | 区画線検出システム及び区画線検出方法 |
JP2015179368A (ja) * | 2014-03-19 | 2015-10-08 | 株式会社日本自動車部品総合研究所 | 走行区画線認識装置及び走行区画線認識プログラム |
JP2019014407A (ja) * | 2017-07-10 | 2019-01-31 | 株式会社Soken | 走路認識装置 |
Non-Patent Citations (2)
Title |
---|
S. SCHULTER ET AL.: "Learning to Look around Objects for Top-ViewRepresentations of Outdoor Scenes", ECCV, 2018, pages 787 - 802 |
See also references of EP4254944A4 |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2024009519A1 (ja) * | 2022-07-08 | 2024-01-11 | 日本電信電話株式会社 | 認識装置、認識方法及び認識プログラム |
Also Published As
Publication number | Publication date |
---|---|
JPWO2022145054A1 (ja) | 2022-07-07 |
EP4254944A4 (en) | 2024-01-31 |
JP7505596B2 (ja) | 2024-06-25 |
US20240071104A1 (en) | 2024-02-29 |
EP4254944A1 (en) | 2023-10-04 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
JP6741107B2 (ja) | 交通状況のリアルタイム検出 | |
JP7332726B2 (ja) | ヒートマップを使用したドライバーの注目の検出 | |
JP6548691B2 (ja) | 画像生成システム、プログラム及び方法並びにシミュレーションシステム、プログラム及び方法 | |
US20190385336A1 (en) | Vehicle Localization Using Cameras | |
US7957559B2 (en) | Apparatus and system for recognizing environment surrounding vehicle | |
US11338807B2 (en) | Dynamic distance estimation output generation based on monocular video | |
CN106647776B (zh) | 车辆变道趋势的判断方法、判断装置和计算机存储介质 | |
JP6082802B2 (ja) | 物体検知装置 | |
GB2560625A (en) | Detecting vehicles in low light conditions | |
US11574462B1 (en) | Data augmentation for detour path configuring | |
CN112215306A (zh) | 一种基于单目视觉与毫米波雷达融合的目标检测方法 | |
WO2020154990A1 (zh) | 目标物体运动状态检测方法、设备及存储介质 | |
JP6653361B2 (ja) | 路面標示画像処理装置、路面標示画像処理方法及び路面標示画像処理プログラム | |
JP3849505B2 (ja) | 障害物監視装置及びプログラム | |
US8170284B2 (en) | Apparatus and method for displaying image of view in front of vehicle | |
US20240005642A1 (en) | Data Augmentation for Vehicle Control | |
Kühnl et al. | Visual ego-vehicle lane assignment using spatial ray features | |
CN114750696A (zh) | 车辆视觉呈现的方法、车载设备及车辆 | |
JP2024019588A (ja) | 地図データ生成装置 | |
WO2022145054A1 (ja) | 画像処理装置、画像処理方法、及び記録媒体 | |
US20240193957A1 (en) | Advanced driver assist system and method of detecting object in the same | |
KR20190134303A (ko) | 영상 인식 장치 및 그 방법 | |
US11702011B1 (en) | Data augmentation for driver monitoring | |
Franke et al. | Steps towards an intelligent vision system for driver assistance in urban traffic | |
JP4469980B2 (ja) | 移動体追跡のための画像処理方法 |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 21914916 Country of ref document: EP Kind code of ref document: A1 |
|
WWE | Wipo information: entry into national phase |
Ref document number: 18270034 Country of ref document: US |
|
ENP | Entry into the national phase |
Ref document number: 2022572879 Country of ref document: JP Kind code of ref document: A |
|
ENP | Entry into the national phase |
Ref document number: 2021914916 Country of ref document: EP Effective date: 20230629 |
|
NENP | Non-entry into the national phase |
Ref country code: DE |