JP4995555B2 - Image processing device - Google Patents

Description

  The present invention relates to an image processing apparatus that binarizes a captured image around a vehicle captured by an in-vehicle camera and detects an edge in the binarized captured image.

  There is known an image sensor including a camera including a CMOS image sensor and an image processing apparatus that detects an edge (contour) from a captured image and recognizes a subject from the shape of the edge. In recent years, a vehicle safety control system for recognizing an object in front of the vehicle by using such an image sensor is spreading, and an example thereof is described in Patent Document 1.

  The image processing apparatus of such an image sensor first binarizes a captured image into, for example, “1” corresponding to black or “0” corresponding to white for each pixel corresponding to the resolution of the in-vehicle camera. The converted image data is stored in the memory. Next, the image processing apparatus individually accumulates the binarized image data in the vertical and horizontal directions (histogram) and obtains the distribution of the pixels to which black is assigned based on the accumulated value to obtain an edge in the image. Is detected. Then, the image processing apparatus performs image recognition by comparing the shape of the detected edge with the shape of the edge of the preceding vehicle or other obstacle stored in advance.

  In such image processing, a large amount of memory is required to store image data to be processed. For example, in order to store binarized image data binarized for each pixel, a memory having a capacity of at least the same number of bits as the number of pixels is required. In edge detection, histograms are separately formed in the vertical direction and the horizontal direction, respectively. Therefore, a memory having a capacity capable of storing binary image data for two images is required for edge detection of one image. Thus, the binarized image data contributes to the pressure on memory resources.

  Further, as the amount of binarized image data increases, the amount of edge detection processing increases. Therefore, the binarized image data also contributes to increase the processing load on the image processing apparatus.

  Therefore, various methods for reducing the amount of binarized image data have been proposed for the purpose of reducing the amount of memory and the load of edge detection processing. As an example, a method has been proposed in which pixels of a captured image are divided into pixel blocks made up of, for example, two vertical and horizontal pixels and binarized in units of pixel blocks. According to this method, the data amount of the binarized image data can be suppressed to a quarter while reducing the resolution of the image.

As another example, paying attention to the continuity of images captured at a frequency of, for example, 30 images per second by an in-vehicle camera, the image in the range near the edge detected in the previous captured image is binarized in units of one pixel. A method has also been proposed in which images in other ranges are binarized in units of pixel blocks. According to this method, edge detection can be performed without reducing the resolution, and the amount of binarized image data can be suppressed to some extent.
JP-A-2005-332120

  However, in the conventional technology as described above, the preceding vehicle, which is the main imaging target, moves at a high speed, and at the same time, the host vehicle also moves at a high speed. That is, since the amount of movement of the object in the image increases from one captured image to the next captured image, in order to detect the moved edge accurately, the binarized range in units of one pixel is set to the previous time. It must be set widely near the edge of the captured image. Then, the data amount of the binarized image data is not so small, and the effect of reducing the memory amount and the load of edge detection processing cannot be obtained sufficiently.

  Accordingly, an object of the present invention is to provide an image processing apparatus capable of reducing the amount of memory and the load of edge detection processing while performing accurate edge detection even when the amount of movement of the photographing object is large. is there.

  In order to achieve the above object, according to a first aspect of the present invention, there is provided an image processing apparatus that binarizes captured images around a vehicle that are sequentially captured by an in-vehicle camera and detects an edge in the binarized captured image. The second position of the edge in the second photographed image after the first photographed image based on the first position of the edge detected in the first photographed image and the movement information of the edge In the second captured image, the range around the second position is binarized in units of the first number of pixels, and the other range is in units of the second number of pixels larger than the first number of pixels. It is characterized by binarizing with.

  According to the above aspect, the preceding position is predicted by predicting the second position of the edge in the second photographed image based on the first position of the edge detected in the first photographed image and the movement information of the edge. Even when an object such as a vehicle and the host vehicle move at high speed, the minimum range around the predicted edge position is binarized in units of the first number of pixels, and the other range is converted into the first range. Binarization can be performed in units of a second number of pixels larger than the number of pixels. In other words, the binarization range in units of the first number of pixels is high in resolution but consumes a large amount of memory. Therefore, binarization is performed without reducing the accuracy of edge detection by minimizing the range. The amount of image data can be reduced. Therefore, it is possible to reduce the amount of memory for storing binarized image data and the load of edge detection processing.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings. However, the technical scope of the present invention is not limited to these embodiments, but extends to the matters described in the claims and equivalents thereof.

  FIG. 1 is a diagram illustrating an example in which the image processing apparatus according to the present embodiment is mounted on a vehicle. The vehicle 1 is equipped with an image sensor 2 that captures the front of the vehicle and recognizes a captured image. The image sensor 2 is configured by integrating an in-vehicle camera 2a and an image processing device 2b. The in-vehicle camera 2a is a fixed camera, and includes, for example, a 640 × 480 pixel CMOS image sensor. The in-vehicle camera 2a captures 30 images per second in a certain range 2c in front of the vehicle, and sends the captured image data to the image processing device 2b. input. The image processing device 2b processes the captured image data, binarizes the captured image, detects an edge, and recognizes the edge shape, and outputs recognition result data.

  Further, the vehicle 1 includes a scan-type in-vehicle radar 3 that detects a target state such as a relative position, a distance, and a speed of a target such as a preceding vehicle in a detection range 3a at a predetermined angle in front of the vehicle, Various sensors such as a vehicle speed sensor 4 for detecting speed, a G sensor 5 for detecting acceleration of the vehicle 1, and a steering sensor 6 for detecting a steering angle by a driver of the vehicle 1 are mounted. The image recognition result data by the image sensor 2, the target state data detected by the in-vehicle radar 3, and the vehicle state data such as the vehicle speed, acceleration, and steering direction detected by the various sensors 4-6 are controlled by a control ECU (Electronic Control). Unit) 8. Based on these input data, the control ECU 8 controls an actuator such as the brake 7 to control the traveling of the vehicle 1 or causes an alarm device (not shown) to give a warning to the driver. Thus, a vehicle safety control system is configured by the control ECU 8 and various devices connected thereto.

  FIG. 2 is a diagram illustrating the configuration of the image processing apparatus 2b in the present embodiment. First, the binarization processing unit 21 is input with the captured image data 2d of 30 images per second from the in-vehicle camera 2a, and binarizes the image according to the density gradation value of the pixels constituting each captured image ("0" or “1”), and binarized image data 21 a of the binarized captured image is stored in the first memory 22. Next, the histogram processing unit 23 accumulates the values in the vertical and horizontal directions of the image based on the binarized image data 21 a read from the first memory, forms a histogram, and stores the histogram data 23 a in the second memory 24. To do.

  The edge detection unit 25 reads the histogram data 23a from the second memory, obtains the distribution of pixels assigned “1” in the image, performs edge detection, and obtains edge position data 25a composed of the coordinates of the detected edge. 3 Store in memory 26. Here, “edge” refers to a line that forms the contour of an object in the image, or a line that forms the shape or pattern of an area inside the contour and the contour, depending on the change point of brightness of the captured image. In the following description, lines constituting the contour of an object will be described as an example.

  The CPU 27 reads out the edge position data 25a stored in the third memory 26, compares the detected edge shape with the edge shapes of various objects stored in the fifth memory 30 in advance, and performs image recognition of the subject. The image recognition result data 27a is output to the control ECU 8.

  Here, the binarization processing of the captured image by the binarization processing unit 21 will be described with reference to FIG. First, FIG. 3A shows a case where the captured image data 2d input from the in-vehicle camera 2a is processed to binarize the captured image in the resolution of the in-vehicle camera 2a, that is, in units of one pixel of 640 × 480 pixels. . For example, with regard to captured image data of 256 gradations (8 bits) per pixel, the binarization processing unit 21 applies “1” to pixels having a gradation value equal to or higher than the intermediate value (128), and pixels having gradation values less than that. “0” is assigned to binarize the captured image to generate binarized image data 21a. As a result, 640 × 480-bit binary image data 21a is generated. Further, since the cumulative total calculation is performed in the vertical and horizontal directions by the histogram processing unit 23, the first memory 22 needs to have a capacity of 640 × 480 × 2 bits in order to store the binary image data 21a for two images. Is done.

  Next, FIG. 3B shows a case where a captured image is divided into pixel blocks each composed of 4 pixels of 2 pixels vertically and horizontally and binarized in units of pixel blocks. In this case, the total value of the gradation values for the four pixels included in the pixel block is a value from 0 to 1020. Accordingly, the binarization processing unit 21 assigns “1” to the pixel block having the total value higher than the intermediate value (510) and “0” to the pixel block having the lower total value, and sets 2 for each pixel block. The valued image data 21a is generated. The data amount of the binarized image data 21a generated as a result is a bit number that is a quarter of the number of pixels. Therefore, even if binary image data 21a for two images is stored in the first memory 22 for histogram processing, the required memory amount is 640 × 480 × 2/4 bits.

  Comparing the binarization processing of FIGS. 3A and 3B described above, the method of FIG. 3A has a high resolution although the resolution of the binarized photographed image is high, and FIG. On the other hand, the resolution is low, but the amount of data is kept small. In the present embodiment, a binarization process combining the methods of FIGS. 3A and 3B is performed as follows.

  Returning to FIG. 2, the edge position prediction unit 28 reads out the edge position data 25 a detected for each captured image from the third memory 26 and stores the captured images continuously input from the in-vehicle camera 2 a. 29. The fourth memory 29 has a capacity for storing edge position data for several images. When the latest edge position data 25a is input, old data is overwritten. Then, the edge position prediction unit 28 predicts the edge position of the current captured image from the edge position of the previous captured image read from the fourth memory 29 and the edge movement information described later, and inputs the prediction result data 28 a to the CPU 27. .

  Then, the CPU 27 calculates a certain range in the vicinity of the predicted edge position of the photographed image as an edge detection range in which binarization is performed for each pixel as shown in FIG. The edge detection range data 27b is input to the processing unit 21. Then, the binarization processing unit 21 binarizes the captured image in the edge detection range set by the input in units of one pixel, and binarizes the other range in units of pixel blocks composed of four pixels. The valued image data 21 a is stored in the first memory 22.

  In this way, the edge detection range near the edge can be limited and set by predicting the edge position. Therefore, it is possible to reduce the data amount of the entire binarized image data 21a without reducing the resolution near the edge of the binarized captured image. Next, an example of edge position prediction in this embodiment will be described for each type of edge position movement information.

  FIG. 4 is a diagram for explaining a first example of edge position prediction in the present embodiment. FIGS. 4A to 4D schematically show images taken by the in-vehicle camera 2a. As described above, captured image data for 30 images per second is sequentially input from the in-vehicle camera 2a to the binarization processing unit 21. Therefore, FIG. 4A shows an edge E1 (displayed by a one-dot chain line) detected in the previous captured image and an edge E2 (displayed by a solid line) detected in the previous captured image, which are indicated by an arrow D. The moving direction and moving amount from the edge E1 to E2 shown can be obtained. Therefore, in the first example, the movement position and movement amount from the edge E1 to E2 are used as the “movement information” of the edge to predict the edge position in the current captured image.

  Specifically, the edge position prediction unit 28 stores the position data of the previous edge E1 and the position data of the previous edge E2 in the fourth memory 29, and from these data, the edge movement direction and the movement amount are stored. The movement information of the edge is obtained. Then, as shown in FIG. 4B, the edge position prediction unit 28 determines the position of the edge E3 (displayed by a dotted line) of the current captured image as the previous edge E2 and the edge movement information indicated by the arrow D. Predict based on.

  Then, the CPU 27 sets, in the binarized processing unit 21, the edge detection range A1 (range indicated by hatching) in the vicinity of the edge position E3 predicted as shown in FIG. . Then, the binarization processing unit 21 binarizes the edge detection range A1 in units of one pixel, and the other range A2 binarizes in units of four pixel blocks, thereby binarizing image data of the entire captured image. Is stored in the first memory 22.

  Here, the prior art and the binarization processing in the present embodiment will be compared. As shown in FIG. 4D, in the prior art, the edge detection range A3 near the previous edge E2 is binarized in units of one pixel, and the other ranges are binarized in units of pixel blocks. However, since the edge position is not predicted in the prior art, the edge detection is performed so that the edge can be detected with high accuracy by binarizing the image area where the edge exists even if the edge E2 moves in any direction. It is necessary to set the range A3 wide. Then, in the edge detection range A3, since the binarized image data has a data amount of the same number of bits as the number of pixels, the data amount of the entire image is not so small. On the other hand, in the present embodiment, as shown in FIG. 4C, the edge detection range A1 can be set more limited by predicting the position of the edge E3, so the data amount of the binarized image data 21a is made smaller. Can be suppressed. Further, since the range where the edge E3 exists with high accuracy is set as the edge detection range, the image in the range can be binarized in units of one pixel with high resolution, and the edge detection can be performed with high accuracy.

  FIG. 5 is a diagram for explaining second and third examples of edge position prediction in the present embodiment. 5A and 5B schematically show images taken by the in-vehicle camera 2a. First, a second example will be described. As shown in FIG. 5A, the edge position predicting unit 28 determines the position of the edge E21 (indicated by a solid line) of the previous captured image and the vehicle state input from the vehicle speed sensor 4, the G sensor 5, and the steering sensor 6. Based on edge movement information including data, the position of the edge E22 (displayed with a dotted line) of the current captured image is predicted.

  Specifically, as shown in FIG. 5C, for example, the preceding vehicle 100 is present in the imageable range 2c of the in-vehicle camera 2a of the host vehicle 1, and the preceding vehicle 100 is detected as the edge E21 in the captured image. To do. Assuming that the host vehicle 1 starts to move to the left as indicated by the arrow L by the driving operation of the driver, the vehicle speed sensor 5 indicates the moving speed, the steering sensor 6 indicates the steering direction of the steering wheel, and the G sensor indicates the lateral acceleration. The vehicle state data consisting of the detection results is input to the edge position prediction unit 28. Then, the edge position prediction unit 28 determines that the preceding vehicle 100 represented by the edge E21 moves to the right relative to the in-vehicle camera 2a of the own vehicle 1 based on the vehicle state data, and according to the moving speed of the vehicle 1 Then, the position where the edge E21 has moved to the right side by the movement amount represented by the arrow D1 is predicted as the position of the edge E22.

  Then, the CPU 27 sets an edge detection range A11 (range indicated by oblique lines) near the predicted edge E22 as shown in FIG. Then, the binarization processing unit 21 binarizes the edge detection range A11 in units of one pixel, binarizes the other range A21 in units of four pixel blocks, and binarized image data of the entire captured image. Is stored in the first memory 22. As described above, the image processing device 2b also achieves the same effects as the first example described above by predicting the edge position of the current captured image based on the previous edge position and the vehicle state of the vehicle 1.

  Further, as a third example, as shown in FIG. 5C, the current photographing is performed based on target state data such as the position, distance, and speed of the preceding vehicle 100 existing within the detection range 3a of the in-vehicle radar 3. The edge position E31 in the image can be predicted. That is, when it is detected by the in-vehicle radar 3 that the angle direction of the preceding vehicle 100 is moving to the right as indicated by the arrow R, the target state data is converted to the edge position prediction unit 28 of the image processing device 2b. Is input. Then, the edge position prediction unit 28 determines that the preceding vehicle 100 represented by the edge E21 moves to the right relative to the in-vehicle camera 2a of the host vehicle 1 as shown in FIG. The position where the edge E21 has moved to the right side by the corresponding amount (arrow D1) is predicted as the position of the edge E22. Then, the CPU 27 and the binarization processing unit 21 perform the same processing as in the second example described above. Therefore, also in this case, the same effect as the second example is achieved.

  It is also possible to predict the edge position by combining the second and third examples. That is, by predicting the edge position based on the vehicle state data of the vehicle 1 and the target state data of the other vehicle, a more accurate prediction can be performed. Therefore, the edge detection range A11 that is binarized in units of one pixel can be set narrower, and the data amount of the binarized image data 21a can be further reduced.

  Furthermore, the edge position can be predicted by combining the first example described above and one or both of the second and third examples. That is, the edge movement information is obtained from the edge position in the previous captured image, and the edge position is predicted based on the edge movement information and the vehicle state data of the vehicle 1 or the target state data of another vehicle. , More accurate prediction is possible. Therefore, in this case as well, the edge detection range that is binarized on a pixel-by-pixel basis can be set narrower than when the first to third examples are used alone, and the data amount of the binarized image data 21a is further reduced. It becomes possible.

  FIG. 6 is a flowchart for explaining the operation procedure of the image processing apparatus 2b in the present embodiment. First, the image processing apparatus 2b performs processing of the first captured image captured by the in-vehicle camera 2a immediately after the power is turned off (step S100). That is, the entire captured image is binarized in units of four pixel blocks without setting an edge detection range that is binarized in units of one pixel (S1), and edge detection is performed (S2). Since accurate edge detection is not necessary before the vehicle starts traveling, the amount of binarized image data can be reduced by binarizing the image in units of four pixels. Then, image recognition is performed based on the detected edge, and recognition result data of the first captured image is transmitted to the control ECU 8 (S3).

  Next, the image processing apparatus 2b performs image processing of the second captured image (S200). That is, the movement information of the edge including the vehicle state of the vehicle 1 is acquired (S4), and the second captured image is obtained based on the edge position in the first captured image detected in step S2 and the acquired edge movement information. The edge position is predicted (S5). Then, an edge detection range to be binarized in units of one pixel around the predicted edge position is set (S6), and the captured image is binarized to detect an edge (S7). Then, image recognition is performed based on the detected edge, and recognition result data of the second captured image is transmitted to the control ECU 8 (S8).

  Then, the image processing device 2b performs image processing after the third captured image (S300). That is, a new edge position is predicted based on the position of the edge detected in the previous captured image and the difference between the edge position and the edge position detected in the previous captured image (S9), and the vehicle state of the vehicle 1 Alternatively, edge movement information including the target state by another vehicle is acquired (S10), and the edge position is predicted again from the prediction result predicted in step S9 and the edge position movement information (S11). Then, an edge detection range around the predicted edge position is set (S12), and the captured image is binarized to detect an edge (S13). Then, image recognition is performed based on the detected edge, and recognition result data is transmitted to the control ECU 8 (S14). While the vehicle 1 is traveling, the procedure S300 including the procedures S9 to S14 is repeatedly executed.

  The procedure 200 shown above corresponds to the above-described second example, and the procedure 300 corresponds to a combination of the above-described first example and the second or third example. However, the procedure S10 and S11 in the procedure 300 may be omitted, and the procedure corresponding to the first example may be performed, and the procedure S9 may be omitted and the procedure corresponding to the second or third example may be performed. It is.

  FIG. 7 is a sequence diagram for explaining image processing in the related art and the present embodiment. FIG. 7 (1) shows the operation of the conventional in-vehicle camera 2a and image processing apparatus 2b, and FIG. 7 (2) shows the in-vehicle camera 2a, image processing apparatus 2b, various sensors 4-6, in-vehicle radar 3 and the like of this embodiment. An image processing sequence by the operation of is shown. The operation of the image processing apparatus 2b is denoted by the same reference numeral as the procedure shown in the flowchart of FIG. 7, and the repeated procedure S300 has a subscript “−n” (n is a natural number) corresponding to the number of iterations. ) Is attached.

  As shown in FIG. 7A, conventionally, the vehicle-mounted camera 2a sequentially captures 30 images per second, that is, three images every 100 milliseconds (P1, P2,..., Pn). However, since the data amount of the binarized image data 21a is large, the processing load of the edge detection by the image processing device 2b is large, and the binarization process, the edge detection, and the output of the image recognition result data based on the edge detection result are the first photographing. After the image processing (S100), it was performed only once every three images (S200, S300-1, ..., S300-n). On the other hand, as shown in FIG. 7B, in the present embodiment, the data amount of the binarized image data 21a can be reduced, so that the processing load of edge detection processing is reduced, and binarization processing is performed for each captured image. Edge detection is possible, and further, image recognition based on the detected edge and the recognition result can be communicated to the control ECU 8 (S200, S300-1,..., S300-n). Therefore, it is possible to improve time accuracy in image recognition.

  Further, as shown in FIG. 7 (3), in image processing S300-n for one captured image, edge position prediction (S9, S11), edge detection range setting (S12), binarization processing and edge detection ( S13), and image recognition and transmission of recognition result data to the control ECU 8 (S14) are sequentially performed. When the edge position is predicted, vehicle state data of the vehicle 1 from the various sensors 4 to 6 and other vehicles from the in-vehicle radar 3 The target state data is input (S10). Here, the timing at which these data are input can be any timing before the end of the edge position prediction. Further, the transmission of the recognition result data to the control ECU 8 (S14) and the prediction of the edge position in the image processing sequence of the next photographed image can be processed in parallel. By doing so, it is possible to further improve the time accuracy.

  FIG. 8 is a diagram illustrating a fourth example of binarization processing in the present embodiment. In the fourth example, an edge position prediction unit based on target state data representing the relative position, distance, and speed of a target detected outside the imaging range of the in-vehicle camera 2a, which is input from the in-vehicle radar 3. 28 predicts an edge position in a captured image to be binarized. For example, as shown in FIG. 8A, the preceding vehicle 100 in the adjacent lane that exists in the detection range 3a of the in-vehicle radar 3 moves in the direction of the photographing range 2c of the in-vehicle camera 2 as indicated by the arrow L1. The target state data is input from the in-vehicle radar 3 to the edge position prediction unit 28. Then, the edge position prediction unit 28 predicts that the preceding vehicle 100 enters the image capturing possible range 2c of the in-vehicle camera 2, and the edge E32 appears by capturing the preceding vehicle 100 as shown in FIG. 8B. Predict. Then, the CPU 27 sets an edge detection range A31 that is binarized in units of one pixel with respect to the right end region of the image where the edge E32 is located. Then, the binarization processing unit 21 performs binarization processing.

  In this way, even when the edge is not detected in the previous captured image, the edge position in the current captured image can be predicted, and the data of the binarized image data 21a without reducing the edge detection accuracy. The amount can be reduced. Then, once the edge E32 is detected, the image processing can be performed by the second and third examples described above, and further by the first example after processing two or more images.

  In the present embodiment, the vicinity of the predicted edge position is binarized in units of one pixel, and the other ranges are binarized in units of four pixel blocks. However, the number of pixels for binarization is not limited to this, and the unit for the number of pixels for binarizing the predicted edge detection range near the edge may be smaller than the unit for the number of pixels for binarizing the other ranges. For example, the data amount of the binarized image data as the entire captured image can be reduced without reducing the resolution in the vicinity of the edge from the resolution in other ranges.

  As described above, according to the present invention, it is possible to reduce the memory amount and the load of edge detection processing while performing accurate edge detection even when the amount of movement of the photographing object is large.

It is a figure explaining the example by which the image processing apparatus in this embodiment is mounted in a vehicle. It is a figure explaining the structure of the image processing apparatus 2b in this embodiment. It is a figure explaining the binarization process of the picked-up image by the binarization process part. It is a figure explaining the 1st example of edge position prediction in this embodiment. It is a figure explaining the 2nd and 3rd example of edge position prediction in this embodiment. It is a flowchart figure explaining the operation | movement procedure of the image processing apparatus 2b in this embodiment. It is a sequence diagram explaining the image processing in a prior art and this embodiment. It is a figure explaining the 4th example of the binarization process in this embodiment.

Explanation of symbols

2b: image processing device, 3: on-vehicle radar, 4: vehicle speed sensor, 5: G sensor,
6: steering sensor, 21: binarization processing unit, 22: first memory,
23: Histogram processing unit, 25: Edge detection unit, 27: CPU, 28: Edge position prediction unit

Claims (5)

A binarization processing unit that binarizes captured images around the vehicle that the in-vehicle camera sequentially captures,
An edge detection unit for detecting an edge in the binarized captured image;
An edge for predicting the second position of the edge in the second photographed image after the first photographed image based on the first position of the edge detected in the first photographed image and the movement information of the edge A position prediction unit,
The binarization processing unit binarizes a range around the second position in units of the first number of pixels in the second photographed image, and a second range larger than the first number of pixels. characterized by binarizing the number of pixels, the image processing apparatus. In claim 1,
Movement information of the edge is determined by the edge position prediction unit, having a difference between the third third position and said first position of said edge that definitive a captured image in front of the first captured image An image processing apparatus. In claim 1 or 2,
Movement information before Symbol edge, the input to the edge position prediction unit, a vehicle speed of the vehicle, acceleration, an image processing apparatus which comprises a vehicle condition such as a steering direction. In any one of Claims 1 thru | or 3,
Movement information before Symbol edge, the input to the edge position prediction unit, the relative position of the vehicle radar device of target object detected by the shooting range of the vehicle camera the vehicle, distance, speed, etc. target An image processing apparatus including a state . A binarization processing unit that binarizes captured images around the vehicle that the in-vehicle camera sequentially captures,
An edge detection unit for detecting an edge in the binarized captured image;
Target state data representing a target state such as a relative position, a distance, and a speed of a target detected in a detection range including a range outside the imaging range of the in-vehicle camera is input from the in-vehicle radar device of the vehicle, An edge position prediction unit that predicts the position of the edge in the captured image based on the target state ;
The binarization processing unit binarizes a range around the predicted position in the captured image in units of a first number of pixels, and sets the other range to a second number of pixels larger than the first number of pixels. characterized by binarizing a unit, the image processing apparatus.

Images