JP6676885B2 - Recognition device, recognition method and program - Google Patents

Description

  The present invention relates to a recognition device, a recognition method, and a program.

  2. Description of the Related Art Techniques for assisting a driver or storing an image at the time of an accident using an in-vehicle camera are conventionally known. For example, a driver assisting technology using an in-vehicle camera has an automatic braking function for avoiding an obstacle or reducing an impact at the time of a collision, and an alarm function for warning a maintenance of an inter-vehicle distance with a preceding vehicle.

  For example, Patent Literature 1 discloses an invention that recognizes a signal of a traffic signal from an image with high accuracy by separately recognizing the color of the signal of the traffic signal and the shape of the signal from two images of the same subject. It has been disclosed.

  However, in the related art, if the recognition process is performed with high accuracy, the recognition process may take a long time, and a warning based on the result of the recognition process may not be able to keep up.

  The present invention has been made in view of the above, and an object of the present invention is to provide a recognition device, a recognition method, and a program capable of accurately recognizing a signal region indicating a signal of a traffic light in a shorter time.

In order to solve the above-described problem and achieve the object, the present invention provides an acquisition unit that acquires an image in a (R, G, B) color space showing the periphery of a vehicle, and an image capturing unit that acquires the (R, G, B) color. The image of the space is converted into an image of the (Y, U, V) color space, and the color of the area included in the image of the (Y, U, V) color space is converted to a signal of the traffic light as a first-stage color recognition process . based on whether within the scope of color indicating a signal, a first color recognition section that performs processing for recognizing the first recognition target region, as color recognition process of the second stage, the signal of the traffic A second color recognition for performing a process of recognizing a second recognition target area from the first recognition target area using a recognition dictionary learned from the signal pixels shown and the non-signal pixels that are not the signal pixels; And a shape recognition unit configured to recognize a signal shape region indicating a shape of a signal of the traffic signal from the second recognition target region. It comprises a part, a.

  ADVANTAGE OF THE INVENTION According to this invention, there exists an effect that a signal area can be correctly recognized in a shorter time.

FIG. 1 is a diagram illustrating an example of a vehicle on which the recognition device of the embodiment is mounted. FIG. 2 is a diagram illustrating an example 1 of an image including a blue light region. FIG. 3 is a diagram showing Example 2 of an image including a blue light region. FIG. 4 is a diagram illustrating an example of the configuration of the recognition device according to the embodiment. FIG. 5 is a diagram illustrating an example of a (U, V) distribution of red signal pixels. FIG. 6 is a diagram illustrating an example of the (U, V) distribution of the green signal pixel. FIG. 7 is a diagram illustrating an example of a (U, V) distribution of yellow signal pixels. FIG. 8 is a diagram illustrating an example of a method for learning a recognition dictionary for green signal pixels according to the embodiment. FIG. 9 is a diagram illustrating an example of first and second recognition target areas according to the embodiment. FIG. 10 is a diagram illustrating an example of an expansion region according to the embodiment. FIG. 11 is a diagram illustrating an example of a circular pixel region recognized by the Hough transform according to the embodiment. FIG. 12 is a diagram illustrating an example of a rectangular area according to the embodiment. FIG. 13 is a diagram illustrating an example of the recognition result information of the green signal region according to the embodiment. FIG. 14 is a flowchart illustrating an example of a recognition method according to the embodiment. FIG. 15 is a diagram illustrating an example of a hardware configuration of an acquisition unit according to the embodiment.

  Hereinafter, embodiments of a recognition device, a recognition method, and a program will be described in detail with reference to the accompanying drawings.

  FIG. 1 is a diagram illustrating an example of a vehicle 200 on which the recognition device 100 of the embodiment is mounted. The recognition device 100 according to the embodiment is installed near the rearview mirror of the windshield of the vehicle 200. The recognition device 100 according to the embodiment recognizes a signal region indicating a signal of the traffic signal 300 from an image in which the periphery of the vehicle 200 is captured. Here, as an example of the signal region, a blue signal region indicating a blue signal will be described.

  FIG. 2 is a diagram illustrating an example 1 of an image including the blue signal region 101. The example of FIG. 2 shows the blue signal region 101 of the traffic signal 300a in which the signal arrangement is horizontal. For example, in Japan, a horizontal traffic light 300a is common.

  FIG. 3 is a diagram illustrating an example 2 of an image including the blue signal area 401. The example of FIG. 3 shows the blue signal area 401 of the traffic signal 300b in which the signal arrangement is vertical. For example, in the United States, a vertical traffic light 300b is common. In Japan, a vertical traffic signal 300b is installed in an area with a large amount of snowfall. Hereinafter, when the traffic light 300a and the traffic light 300b are not distinguished, they are simply referred to as the traffic light 300.

  Next, an example of the configuration of the recognition device 100 according to the embodiment will be described.

  FIG. 4 is a diagram illustrating an example of a configuration of the recognition device 100 according to the embodiment. The recognition device 100 according to the embodiment includes an acquisition unit 10 and a signal processing unit 20. The signal processing unit 20 includes an interface unit 21, a first color recognition unit 22, a second color recognition unit 23, a shape recognition unit 24, and an output unit 25.

  The acquisition unit 10 acquires an image around the vehicle. The acquisition unit 10 of the present embodiment is a camera. The acquisition unit 10 captures an image of the periphery of the vehicle 200, and acquires a color image represented by the (R, G, B) color space. The acquisition unit 10 inputs the image to the interface unit 21. Note that the acquisition unit 10 may acquire an image around the vehicle from another device.

  When receiving the image from the acquisition unit 10, the interface unit 21 converts the image into a data format that the first color recognition unit 22 can accept as a time-series continuous image frame. The interface unit 21 inputs the image whose data format has been converted to the first color recognition unit 22.

  When the first color recognizing unit 22 receives an image from the interface unit 21, the first color recognizing unit 22 performs a threshold determination process of determining whether a color of an area included in the image is included in a color range indicating a signal of the traffic light 300. , The first recognition target area is recognized. Specifically, first, the first color recognizing unit 22 converts an image in the (R, G, B) color space into an image in the (Y, U, V) color space using the following equation (1). Convert.

  Next, the first color recognition unit 22 determines a first recognition target area by a threshold determination process using a threshold for determining a red signal pixel, a threshold for determining a blue signal pixel, and a threshold for determining a yellow signal pixel. recognize. Here, the red signal pixel is a pixel indicating the color of the red signal. The blue signal pixel is a pixel indicating the color of the blue signal. The yellow signal pixel is a pixel indicating the color of the yellow signal.

  The threshold for determining the red light pixel is a threshold (Ur-min and Ur-max) indicating the range of U and a threshold (Vr-min and Vr-max) indicating the range of V. That is, the first color recognition unit 22 recognizes (Y, U, V) satisfying Ur-min ≦ U ≦ Ur-max and Vr-min ≦ V ≦ Vr-max as the first recognition target area. I do. This threshold is determined based on the (U, V) distribution of the red signal pixels.

  FIG. 5 is a diagram illustrating an example of a (U, V) distribution of red signal pixels. The example of FIG. 5 shows a (U, V) distribution of red signal pixels obtained from pixel information in a (R, G, B) color space of a plurality of image samples in which a red signal is captured. The specific values of the thresholds (Ur-min and Ur-max) indicating the range of U and the thresholds (Vr-min and Vr-max) indicating the range of V may be arbitrarily determined. However, if the range of the color determined by the threshold is too wide, the possibility of detecting a pixel that is not the color of the red signal increases. Therefore, the threshold is set so as not to overlap the (U, V) value of the pixel that is not the red signal. Set.

  The thresholds for determining the green pixel are a threshold (Ub-min and Ub-max) indicating the range of U and a threshold (Vb-min and Vb-max) indicating the range of V. That is, the first color recognition unit 22 recognizes (Y, U, V) satisfying Ub-min ≦ U ≦ Ub-max and Vb-min ≦ V ≦ Vb-max as the first recognition target area. I do. This threshold value is determined based on the (U, V) distribution of the green signal pixels.

  FIG. 6 is a diagram illustrating an example of the (U, V) distribution of the green signal pixel. The example of FIG. 6 illustrates a (U, V) distribution of blue signal pixels obtained from pixel information in the (R, G, B) color space of a plurality of image samples in which a blue signal is captured. The specific values of the thresholds (Ub-min and Ub-max) indicating the range of U and the thresholds (Vb-min and Vb-max) indicating the range of V may be arbitrarily determined. However, if the range of colors determined by the threshold is too wide, the possibility of detecting a non-blue color pixel increases, so the threshold is set so as not to overlap the (U, V) value of the non-blue color pixel. .

  The thresholds for determining a yellow signal pixel are a threshold (Uy-min and Uy-max) indicating a range of U and a threshold (Vy-min and Vy-max) indicating a range of V. That is, the first color recognition unit 22 recognizes (Y, U, V) satisfying Uy-min ≦ U ≦ Uy-max and Vy-min ≦ V ≦ Vy-max as the first recognition target area. I do. This threshold is determined based on the (U, V) distribution of the yellow signal pixels.

  FIG. 7 is a diagram illustrating an example of a (U, V) distribution of yellow signal pixels. The example of FIG. 7 illustrates a (U, V) distribution of yellow signal pixels obtained from pixel information in the (R, G, B) color space of a plurality of image samples in which yellow signals are captured. The specific values of the thresholds (Uy-min and Uy-max) indicating the range of U and the thresholds (Vy-min and Vy-max) indicating the range of V may be arbitrarily determined. However, if the range of the color determined by the threshold is too wide, the possibility of detecting a non-yellow color pixel increases. Therefore, the threshold is set so as not to overlap the (U, V) value of the non-yellow color pixel. Set.

  Since the recognition processing of the first recognition target area by the first color recognition unit 22 is performed by the above-described threshold determination processing, it can be executed at high speed.

  Returning to FIG. 4, when the first color recognizing unit 22 recognizes the first recognition target area, the information indicating the first recognition target area is input to the second color recognition unit 23.

  When the second color recognizing unit 23 receives the information indicating the first recognition target area from the first color recognizing unit 22, the second color recognizing unit 23 uses the recognition dictionary to convert the first recognition target area into the second recognition target area. Recognize the area. The recognition dictionary is a dictionary learned from signal pixels indicating signals of the traffic light 300 and non-signal pixels that are not signal pixels. Here, an example of a learning method of a recognition dictionary for recognizing a green signal pixel will be described.

  FIG. 8 is a diagram illustrating an example of a method for learning a recognition dictionary for green signal pixels according to the embodiment. The recognition dictionary of the green signal pixel is learned by a machine learning method such as SVM (Support Vector Machine). When learning by the SVM, the straight line 124 is determined so that the distance (margin) between the dotted line 123 and the straight line 124 and the distance between the straight line 124 and the dotted line 125 are maximized.

  The blue signal pixel 121 and the non-blue signal pixel 122 are classified by the straight line 124 determined in this way. Specifically, parameters a1, b1 and c1 of the evaluation function f of the following equation (2) are calculated as a recognition dictionary.

  f (U, V) = a1 × U + b1 × V + c1 (2)

  When f (U, V) is equal to or larger than the threshold value for each (U, V) of the first recognition target area in the (Y, U, V) color space, (U, V) is recognized as a green signal pixel, and if it is less than the threshold, the (U, V) is recognized as a non-green signal pixel. The second color recognizing unit 23 recognizes the area of the blue signal pixel recognized based on the evaluation function f as a second recognition target area.

  Note that in the description of the case where the red signal pixel is recognized as the second recognition target area and the case where the yellow signal pixel is recognized as the second recognition target area, the above-described blue signal pixel is recognized as the second recognition target area. The description is omitted because it is the same as the case described above.

  Note that the recognition processing by the recognition dictionary is not limited to images in the (U, V) color space. However, since the signal of the traffic signal 300 has clear color characteristics, recognition processing using a recognition dictionary for images in the (U, V) color space is more preferable.

  The above-described process of recognizing the second recognition target area by the second color recognizing unit 23 is performed not on all pixels but on the first recognition target area in a limited range. it can. Further, by further performing the processing of the second color recognition unit 23, the first recognition target area obtained by the first color recognition unit 22 is affected by the influence of the weather, the shooting time of the image, noise, and the like. When an error is included, the error can be reduced. Thereby, the recognition target area of the signal of the traffic signal 300 can be recognized with higher accuracy.

  Hereinafter, the case of recognizing the blue signal region 101 in FIG. 2 will be specifically described with reference to FIGS. The same applies to the case where the red signal area and the yellow signal area are recognized.

  FIG. 9 is a diagram illustrating an example of first and second recognition target areas according to the embodiment. First, the first color recognition unit 22 satisfies a threshold value (Ub-min and Ub-max) indicating a range of U for determining a green signal pixel and an area satisfying threshold values (Vb-min and Vb-max) indicating a range of V ( U, V) are recognized as the first recognition target area 102. Next, the second color recognition unit 23 determines that f (U, V) is equal to or larger than the threshold for each (U, V) of the first recognition target area 102 in the (Y, U, V) color space. , (U, V) is recognized as the second recognition target area 103.

  The size of the second recognition target area 103 in FIG. 9 is smaller than the size of the actual blue signal area 101 (see FIG. 2). That is, the example of FIG. 9 illustrates a case where an area that should be originally recognized as a blue signal area is not recognized as a green signal pixel due to the influence of a noise pixel. The noise pixel includes a noise pixel caused by a surrounding situation at the time of photographing, a noise pixel caused by characteristics of the image sensor, a noise pixel caused by dust attached to the surface of the image sensor, and the like. The noise pixel due to the surrounding state at the time of shooting is a pixel in a region of a blue signal reflected by light such as sunlight when the traffic light 300a is shot. The noise pixel due to the characteristics of the image sensor is, for example, a pixel affected by random noise.

  Referring back to FIG. 4, the second color recognizing unit 23 inputs information indicating the second recognition target area 103 to the shape recognizing unit 24.

  When receiving the information indicating the second recognition target region 103 from the second color recognition unit 23, the shape recognition unit 24 performs an expansion process for expanding the second recognition target region 103. Specifically, the shape recognition unit 24 expands the second recognition target area 103 into an expansion area by covering the pixel with a plurality of pixels for each pixel of the second recognition target area 103. The shape recognition unit 24 covers each pixel with, for example, n × n (n is an integer of 1 or more) block pixels. For example, when n = 7, each pixel of the second recognition target area 103 is expanded into an expansion area further including 48 (7 × 7-1) pixels around the pixel.

  FIG. 10 is a diagram illustrating an example of the expansion region 104 according to the embodiment. In the example of FIG. 10, the shape recognition unit 24 performs the expansion processing on the second recognition target region 103 recognized by the second color recognition unit 23, thereby indicating the size of the original blue signal region. This shows a case where an expanded area 104 including a green signal area 101 is obtained.

  Returning to FIG. 4, the shape recognizing unit 24 performs a shape recognizing process for recognizing a blue signal shape region indicating the shape of the green signal region 101. Specifically, the shape recognizing unit 24 recognizes a circular pixel region in the expanded region 104 by performing a Hough transform on the second recognition target region 103 recognized by the second color recognizing unit 23. Determine if you can. When a circular pixel area can be recognized, the shape recognizing unit 24 recognizes the circular pixel area as a green signal shape area. This green signal shape area is recognized as the green signal area 101 of the traffic light 300a.

  FIG. 11 is a diagram illustrating an example of a circular pixel region recognized by the Hough transform according to the embodiment. The example in FIG. 11 illustrates an example in which the blue signal region 101 is recognized as a circular pixel region by subjecting the second recognition target region 103 (see FIG. 9) to Hough transform.

  Referring back to FIG. 4, the shape recognition unit 24 recognizes a rectangular area including the blue signal area 101. The shape recognition unit 24 inputs recognition result information indicating a rectangular region including the blue signal region 101 to the output unit 25. When receiving the recognition result information from the shape recognition unit 24, the output unit 25 outputs the recognition result information. The recognition result information is used for processing such as detection of dangerous driving such as ignoring signals and sudden start by the driver.

  FIG. 12 is a diagram illustrating an example of the rectangular area 105 according to the embodiment. The example of FIG. 12 illustrates an example in which a rectangular area 105 including a blue signal area 101 recognized as a circular pixel area is recognized. Note that the rectangular area 105 may be recognized so as to circumscribe the blue signal area 101.

  FIG. 13 is a diagram illustrating an example of recognition result information of the green signal region 101 according to the embodiment. The example of FIG. 13 illustrates a case where the blue signal area 101 of the traffic light 300a included in the image illustrated in FIG. 2 is recognized as recognition result information indicating the rectangular area 105 including the blue signal area 101.

  Next, a recognition method according to the embodiment will be described.

  FIG. 14 is a flowchart illustrating an example of a recognition method according to the embodiment. Since the description of the case where the signal of the traffic light 300 is a blue signal, a yellow signal, and a red signal is the same, the description of the recognition method of the embodiment will be given by taking the case where the signal is a blue signal as an example. Is optional.

  First, the first color recognition unit 22 receives a color image represented by the (R, G, B) color space from the interface unit 21 (Step S1). Next, the first color recognizing unit 22 converts the image in the (R, G, B) color space into an image in the (Y, U, V) color space by using the above equation (1). (Step S2).

  Next, the first color recognition unit 22 recognizes the first recognition target area 102 by the above-described threshold determination processing using the threshold for determining the green signal pixel (step S3). Next, the second color recognizing unit 23 recognizes the area of the blue signal pixel recognized based on the evaluation function f of the above equation (2) as the second recognition target area 103 (Step S4).

  Next, the shape recognition unit 24 expands the second recognition target area 103 recognized in step S4 into an expansion area 104 (step S5).

  Next, the shape recognizing unit 24 recognizes a blue signal shape region in the expanded region 104 obtained by the process of step S5 (step S6). Specifically, the shape recognition unit 24 determines whether or not a circular pixel region can be recognized in the expanded region 104 by performing a Hough transform on the second recognition target region 103. When a circular pixel area can be recognized, the shape recognizing unit 24 recognizes the circular pixel area as a green signal shape area. This green signal shape area is recognized as the green signal area 101 of the traffic light 300.

  Next, the output unit 25 outputs recognition result information indicating the rectangular area 105 including the blue signal area 101 (Step S7).

  Finally, an example of a hardware configuration of the recognition device 100 according to the embodiment will be described.

  First, an example of a hardware configuration of the acquisition unit 10 will be described.

  FIG. 15 is a diagram illustrating an example of a hardware configuration of the acquisition unit 10 according to the embodiment. The acquisition unit 10 of the embodiment includes a photographing optical system 201, a mechanical shutter 202, a motor driver 203, a charge coupled device (CCD) 204, a correlated double sampling (CDS) circuit 205, an A / D converter 206, and a timing. Signal generator 207, image processing circuit 208, LCD (Liquid Crystal Display) 209, CPU (Central Processing Unit) 210, RAM (Random Access Memory) 211, ROM (Read Only Memory) 212, SDRAM (Synchronous Dynamics) 213, compression / expansion circuit 214, memory 215 An operation unit 216 and an output I / F 217.

  The image processing circuit 208, the CPU 210, the RAM 211, the ROM 212, the SDRAM 213, the compression / expansion circuit 214, the memory 215, the operation unit 216, and the output I / F 217 are connected via the bus 220.

  The imaging optical system 201 collects light reflected by the subject. By opening the mechanical shutter 202 for a predetermined time, the light collected by the photographing optical system 201 is made incident on the CCD 204. The motor driver 203 drives the photographing optical system 201 and the mechanical shutter 202 according to a control signal from the CPU 210.

  The CCD 204 forms light incident through the mechanical shutter 202 as an image of a subject, and inputs analog image information indicating the image of the subject to the CDS circuit 205. Upon receiving the analog image information from the CCD 204, the CDS circuit 205 removes a noise component of the image information and inputs the analog image information from which the noise component has been removed to the A / D converter 206. Upon receiving the analog image information from the CDS circuit 205, the A / D converter 206 converts the analog image information into digital image information. The A / D converter 206 inputs digital image information to the image processing circuit 208. The timing signal generator 207 transmits a timing signal to the CCD 204, the CDS circuit 205, and the A / D converter 206 according to a control signal from the CPU 210, so that the CCD 204, the CDS circuit 205, and the A / D converter 206 Control the timing of operation.

  Upon receiving digital image information from the A / D converter 206, the image processing circuit 208 performs image processing on the digital image information using the SDRAM 213. The image processing includes, for example, YUV (YCrCb) conversion processing, white balance control processing, contrast correction processing, edge enhancement processing, color conversion processing, and the like. The YUV (YCrCb) conversion process is a conversion process according to the above equation (1). That is, the conversion processing of the above equation (1) may be performed by the image processing circuit 208. The white balance control process is an image process for adjusting the color density of image information. The contrast correction processing is image processing for adjusting the contrast of image information. The edge enhancement process is a process for adjusting the sharpness of image information. The color conversion process is an image process for adjusting the hue of image information.

  The image processing circuit 208 inputs the image information subjected to the above-described image processing to the LCD 209 or the compression / expansion circuit 214. The LCD 209 is a liquid crystal display that displays image information received from the image processing circuit 208.

  The CPU 210 controls the operation of the acquisition unit 10 by executing a program. The RAM 211 is a work area when the CPU 210 executes a program, and a readable and writable storage area used for storing various information. The ROM 212 is a read-only storage area that stores programs executed by the CPU 210 and the like.

  The SDRAM 213 is a storage area for temporarily storing image information to be subjected to image processing when the image processing circuit 208 performs image processing.

  Upon receiving the image information from the image processing circuit 208, the compression / expansion circuit 214 compresses the image information. The compression / expansion circuit 214 stores the compressed image information in the memory 215. Upon receiving the image information from the memory 215, the compression / expansion circuit 214 expands the image information. The compression / expansion circuit 214 temporarily stores the expanded image information in the SDRAM 213. The memory 215 stores the compressed image information.

  The operation unit 216 receives an operation from the user of the acquisition unit 10. The operation unit 216 receives an operation of storing image information displayed on the LCD 209 in the memory 215, for example. The output I / F 217 is an interface for transmitting image information from the acquisition unit 10 to the signal processing unit 20.

  The interface unit 21, the first color recognition unit 22, the second color recognition unit 23, the shape recognition unit 24, and the output unit 25 of the signal processing unit 20 described with reference to FIG. ) May be realized by hardware, or may be realized by software (program) executed by the CPU 210 of the acquisition unit 10 and the CPU of another device. The interface unit 21, the first color recognition unit 22, the second color recognition unit 23, the shape recognition unit 24, and the output unit 25 may be realized by combining hardware and software.

  The program executed by the recognition device 100 (CPU 210) of the embodiment is read in a computer such as a CD-ROM, a memory card, a CD-R, and a DVD (Digital Versatile Disk) as an installable or executable file. It is recorded on a possible storage medium and provided as a computer program product.

  The program executed by the recognition device 100 of the embodiment may be stored on a computer connected to a network such as the Internet, and provided by being downloaded via the network. Further, the program executed by the recognition apparatus 100 of the embodiment may be provided via a network such as the Internet without downloading the program.

  Further, the program of the recognition device 100 of the embodiment may be provided by being incorporated in the ROM 212 or the like in advance.

  When the interface unit 21, the first color recognition unit 22, the second color recognition unit 23, the shape recognition unit 24, the output unit 25, and the like are realized by a program executed by the recognition device 100 according to the embodiment, the interface unit 21, the first color recognizing unit 22, the second color recognizing unit 23, the shape recognizing unit 24, the output unit 25, and the like are realized in the RAM 211 by, for example, the CPU 210 reading a program from the ROM 212 or the storage medium and executing the program. Is done.

  As described above, in the recognition device 100 according to the embodiment, the first color recognition unit 22 determines whether the color of the region included in the image is included in the color range indicating the signal of the traffic light 300 or not. , The first recognition target area is recognized. Next, the second color recognition unit 23 uses the recognition dictionary learned from the signal pixels indicating the signal of the traffic signal 300 and the non-signal pixels that are not the signal pixels, to perform the second recognition from the first recognition target area. Recognize the recognition target area. Then, the shape recognition unit 24 recognizes a signal shape region indicating the shape of the signal of the traffic signal 300 from the second recognition target region. Thus, according to the recognition device 100 of the embodiment, the signal region indicating the signal of the traffic light 300 can be accurately recognized in a shorter time.

  Note that the reason why the recognition device 100 of the embodiment does not recognize the shape of the entire traffic light 300 is that it is assumed that the image of the distant traffic light 300 cannot sufficiently recognize the shape of the traffic light 300 as a whole. It is. Therefore, the recognition device 100 of the embodiment recognizes the circular shape of the signal portion of the traffic light 300 that can relatively reliably recognize the shape. In addition, if only the recognition of the circular shape is performed or the recognition of the circular shape is performed in the first recognition target area, inconvenience occurs, for example, a circular pattern accidentally included in the signboard is recognized. Therefore, in the recognition device 100 of the embodiment, a two-stage color recognition process for recognizing the second recognition target region is further performed on the first recognition target region, so that a high-accuracy signal region can be generated more quickly. Make it recognizable.

Reference Signs List 10 acquisition unit 20 signal processing unit 21 interface unit 22 first color recognition unit 23 second color recognition unit 24 shape recognition unit 25 output unit 100 recognition device 101 blue signal area 102 first recognition target area 103 second recognition target Area 104 expansion area 105 rectangular area 200 vehicle 201 shooting optical system 202 mechanical shutter 203 motor driver 204 CCD
205 CDS circuit 206 A / D converter 207 Timing signal generator 208 Image processing circuit 209 LCD
210 CPU
211 RAM
212 ROM
213 SDRAM
214 compression / expansion circuit 215 memory 216 operation unit 217 output I / F
220 bus 300 traffic light

JP 2009-244946 A

Claims (5)

An acquisition unit configured to acquire an image in the (R, G, B) color space indicating the periphery of the vehicle;
The image in the (R, G, B) color space is converted into an image in the (Y, U, V) color space, and the image in the (Y, U, V) color space is converted as a first-stage color recognition process. A first color recognition unit that performs a process of recognizing a first recognition target region based on whether a color of the included region is included in a range of colors indicating a signal of a traffic signal;
As a second stage color recognition process, a second recognition process is performed from the first recognition target area using a recognition dictionary learned from signal pixels indicating the signal of the traffic light and non-signal pixels that are not the signal pixels. A second color recognition unit that performs a process of recognizing a recognition target area;
A shape recognition unit that recognizes a signal shape region indicating the shape of the signal of the traffic signal from the second recognition target region;
A recognition device comprising: The second color recognition unit recognizes the second recognition target area using an evaluation function f (U, V) calculated as the recognition dictionary.
The recognition device according to claim 1 . The shape recognition unit expands the second recognition target region into an expansion region by covering the second recognition target region with a plurality of pixels, and expands the second recognition target region in the expansion region. Recognize a region obtained by Hough transform as the signal shape region,
Recognition apparatus according to claim 1 or 2. A step in which the recognizing device acquires an image in an (R, G, B) color space showing the periphery of the vehicle;
A recognition device converts the image in the (R, G, B) color space into an image in the (Y, U, V) color space, and performs the (Y, U, V) color Performing a process of recognizing a first recognition target region based on whether a color of a region included in the image of the space is included in a color range indicating a signal of a traffic light;
The recognition device uses the recognition dictionary learned from the signal pixels indicating the signal of the traffic light and the non-signal pixels that are not the signal pixels as the second stage color recognition processing, and uses the recognition dictionary learned from the first recognition target area. Performing a process of recognizing a second recognition target region from
A recognition device for recognizing a signal shape region indicating a signal shape of the traffic signal from the second recognition target region;
Recognition method including. A recognition device including an acquisition unit that acquires an image in the (R, G, B) color space indicating the periphery of the vehicle,
The image in the (R, G, B) color space is converted into an image in the (Y, U, V) color space, and the image in the (Y, U, V) color space is converted as a first-stage color recognition process. A first color recognition unit that performs a process of recognizing a first recognition target region based on whether a color of the included region is included in a range of colors indicating a signal of a traffic signal;
As a second stage color recognition process, a second recognition process is performed from the first recognition target area using a recognition dictionary learned from signal pixels indicating the signal of the traffic light and non-signal pixels that are not the signal pixels. A second color recognition unit that performs a process of recognizing a recognition target area;
A shape recognition unit that recognizes a signal shape region indicating a shape of a signal of the traffic signal from the second recognition target region;
Program to function as