JPS6199803A - Recognizing device for vehicle driver's position - Google Patents

Abstract

PURPOSE:To detect a doze at the wheel by detecting the specific part candidate position of a driver's face part on the basis of the standard pattern of the driver's face part from image data, and detecting a specific part on the basis of the standard pattern of the whole specific part. CONSTITUTION:A driver's position detecting device is equipped with a specific part candidate position detecting means M4 which detects the specific part candidate position of the driver's face part, a specific part detecting means M5 which detects a specific part on the basis of the standard pattern of the whole specific part from image data on the periphery of the specific part candidate position, and a position recognizing means M6 which recognizes the three- dimensional position of a driver on the basis of the specific part detected by the specific part detecting means. Then, the specific part candidate part of the driver's face part is detected from the image data on the basis of the pattern of part of the specific part of the driver's face part and the specific part is detected from the image data on the basis of the standard pattern of the whole specific part to recognize the three-dimensional position of the vehicle driver speedily with high precision, thereby detecting a doze at the wheel.

Description

[Detailed Description of the Invention] [Industrial Application Field 1] The present invention relates to a vehicle driver position recognition device that captures the riding condition of a vehicle driver as an image and recognizes the vehicle driver's position in the vehicle interior. The present invention relates to a vehicle driver position recognition device configured to quickly and accurately recognize a predetermined part of a vehicle driver.

1. Prior Art] With the rapid development of 31i electronic equipment, research is underway on new devices aimed at improving the operability, comfort, accident avoidance, etc. of vehicles. The driver's riding position, such as the position of the driver's eyes and head (f/m), is recognized, and the rearview mirror, headrest, air outlet
Alternatively, it is being considered to automatically adjust the position and angle of the steering wheel, etc., or to detect drowsy driving based on periodic changes in the position of the driver's eyes or head and issue a warning to the driver. The riding position recognition device for recognizing the riding position of the vehicle driver, which is necessary to execute such control, is a device that images the riding state of the vehicle driver,
A double-peak detection unit consisting of a camera, etc. that detects the image, and the detected image is analyzed using so-called image processing, and the position of the driver's eyes and head in the image is determined in advance. <'
It is being considered that the driver's position could be recognized using an image processing unit such as

Problem 1 to be solved by the invention E: Automatic adjustment of the positions and angles of the above-mentioned rearview mirrors, headrests, air-conditioned air outlets, steering wheels, etc., and periodic changes in the position of the driver's eyes and head. In order to detect drowsy driving and issue a warning to the driver, it is necessary to accurately recognize predetermined parts of the driver, such as the positions of the eyes and nose.

However, it takes too much processing time to accurately recognize a predetermined part of the driver.

An object of the present invention is to provide a vehicle driver position recognition device that recognizes a predetermined part of the driver quickly and accurately.

[Means for Solving the Problems] As shown in FIG. 1, the configuration of the present invention includes a light emitting means M2 that irradiates the vehicle driver M1, and light emitted from the driver M1 by the light emitting means M2. Image detection means M that receives reflected light and detects the driver as image data
3, a predetermined part candidate position detection means M4 for detecting the position of the predetermined part candidate f17 of the driver's facial region based on the standard pattern of a part of the predetermined region of the driver's facial region from the image data; -1- Detecting a predetermined part based on the standard pattern of the entire predetermined part from image data near the candidate position of the predetermined part fsl 16 Predetermined part detection means M5; 1- Predetermined part detected by the predetermined part detection means A vehicle driver position recognition device comprising: a position recognition device M6 that recognizes the three-dimensional position of the driver based on the location of the driver.

[Function] In the present invention, the image detection means M3 detects one clear image.
The vehicle driver has a light emitting means as a light source other than natural light to illuminate the vehicle driver, and the light emitting means M2 illuminates the vehicle driver. In this case, it is preferable to irradiate infrared light because the driver will not feel dazzled by the light emitting means.

The image detection means M3 receives reflected light from the driver due to irradiation from the light emitting means, and detects information necessary for position recognition as image data of the driver. In this case, it is preferable to arrange the image detection means M3 on the passenger seat side so as to detect a side image of the driver so that characteristic parts of the face such as the nose and chin can be easily recognized. Further, it is preferable to use a two-dimensional solid-state image element as the image detecting means because it requires less space and has excellent impact resistance.

The predetermined part candidate position detection means M4 checks the degree of matching based on a standard pattern of a very small part representing the characteristics of a predetermined part of the driver's face from the image data detected by the image detection means M3, and detects the predetermined part candidates. It detects the position.

The predetermined part candidate position detecting means M4 can consider, for example, the difference between two points in a predetermined direction in the image data as the minimum pattern as a part of the predetermined part, and calculates the difference and calculates a difference that is ten or more than a predetermined value. The candidate position of a predetermined part can be detected by searching for a position where there is a position. Here, if the predetermined direction is horizontal, it is convenient for knowing the outline of the edge of the face and for detecting candidate positions such as the tip of the nose. The predetermined direction may be determined so that the shape of the facial part with respect to the direction has a good characteristic of the predetermined portion.

The predetermined part detection means M5 is a predetermined part candidate position detection means M.
Based on the standard pattern of the entire predetermined region, the degree of matching is checked from the image data in the vicinity of the predetermined region candidate position detected in step 4, and the predetermined region is detected. Since the predetermined part detection means M5 is based on the standard pattern of the entire predetermined part, it is working precisely to recognize the position of the vehicle driver.

Further, the predetermined part candidate position detecting means M4 functions to quickly recognize the position of the vehicle driver.

This is because candidate positions are detected based on a standard pattern of a very small part of a predetermined part, and the pixels in the image data used for the search are compared to when the entire predetermined part is immediately searched. In addition, the predetermined part candidate position detecting means first detects the predetermined part candidate position based on a standard pattern of a part of the predetermined part, and then detects the predetermined part candidate position in the vicinity of the candidate position. Even if a two-step pattern search is performed to detect the predetermined part candidate position based on a standard pattern of a part of the predetermined part that is larger than the standard pattern of the step, the search can be performed more quickly. .

Note that the process of detecting the predetermined part or its candidate position based on the camphor pattern of the whole or part of the predetermined part is a process performed using a so-called spatial filter or the like.

A spatial filter is a filter that obtains new image data by filtering image data under specific conditions.For example, it calculates the degree of matching with a specific pattern, and if the degree of matching is high, extracts the image data of the part A. or two adjacent
-8 It is also possible to find the difference between one point and extract image data that is greater than a predetermined value.

[Example] A first embodiment of the present invention will be described below with reference to the drawings.

FIG. 2 is a perspective view showing the instrument panel part and its surroundings of a vehicle in which the occupant position recognition device of this embodiment is installed, in which 1 represents the driver, 2 represents the driver's seat, and 3 represents the instrument panel. ing. A light emitting unit 5 that illuminates the upper body of the driver 1 from the left side is located diagonally forward to the left of the driver's seat 2 of the instrument panel 3, that is, in front of the passenger seat side. An image detection unit 6 that can generate a two-dimensional image is provided.

Here, the light emitting unit 5 corresponds to the above-mentioned light emitting means (2), and an infrared strobe that emits infrared light is used so as not to dazzle the driver 1 during irradiation, as shown in FIG.
An infrared light emitter 5a, a non-zero 5b for irradiating a wide range of infrared light emitted by the infrared light emitter 5a to the driver, and an infrared filter 5C that transmits infrared light but blocks visible light. and a case 5d for storing the infrared emitter 5a, lens 5b, and infrared filter 5C and attaching it to a predetermined circle 1a of the instrument panel 3.

Further, the image detection section 6 corresponds to the above-mentioned image detection means (2), and as shown in FIG. A lens 6i with a focal length of, for example, 12.5111 to form an image on the image plane of the lens, and a liquid crystal aperture element 6C to adjust the amount of light.
and a MOS comprising a photo diode A-door array and a switching circuit, which extracts the image on the weave edge surface formed by passing through the infrared filter 6a, lens 6b, and liquid crystal aperture element 6C as an electrical signal by switching scanning. A non-interlaced solid-state camera is used, which is comprised of a solid-state image sensor 6e in the form of a solid-state image sensor 6e, and a case 6t for storing the above-mentioned parts and attaching them to the instrument panel 3.

Next, the overall configuration of the riding position recognition device of this embodiment will be explained based on the block diagram shown in FIG. 5.

As shown in the figure, in addition to the light emitting unit 5 and the image detecting unit 6, this recognition device processes () the image of the driver 1 captured by the image detecting unit 6, and processes the driver's position (in this example An image processing unit 8 is provided to recognize the position of the nose. The image processing M8 controls the image detection section 6, converts the image signal jqed by the image detection section 6 into a digital signal, and stores the converted digital signal at a glance as image data. An image input section 10 that can output light, and an irradiation signal output section 12 that outputs an irradiation signal for causing the light emitting section 5 to emit light, etc. Detects the edge of a predetermined part facing the face, performs image processing, and executes a series of position recognition processes such as recognizing the edge image data f) 1 device on driver 1's nose a central processing unit (CPU) 14 to perform position recognition processing, and a read memory (ROM>16) in which control programs and data for executing position recognition processing on the CPUI/l are stored in advance;
Random access memory (RAM) 18 and 1. Random access memory (RAM) 18 temporarily read and write data used for execution of arithmetic processing. It is composed of a path line 20 that connects each of the above sections and serves as a path for image signals and control signals, and a power supply circuit 22 that supplies power to each of the above sections. The image input section 10 also includes a synchronization signal generation circuit 10a that generates a vertical synchronization signal and a horizontal synchronization signal for the image detection section 6, and an image signal obtained by the image detection section 6, for example, at 256 pixels horizontally per frame.
, A/l') conversion circuit 10b that converts the image data into image data consisting of 1J digital signals of 240 vertical pixels, and is capable of storing the converted image data for the past two frames. It is composed of an image data storage circuit 10c.

Next, the operation of the position recognition process executed by the image processing section 8 will be explained in detail along the line 70 shown in FIG. Note that this process is executed when a request for recognizing the position of the driver 1 is inputted by a signal of a certain period or a signal from a switch, etc., and as described above, in this embodiment, the position of the driver 1 is determined as the position of the driver 1. It helps that the position of the nose is recognized.

As shown in the figure, when the process is started, step 101 is first executed, and the image detection section 6 outputs an irradiation signal to the light emitting section 5 so that an image is detected when the light emitting section 5 emits light. An image data reading process is executed in which the detected image is stored in the image input section 10 as image data G consisting of a digital signal.

Here, the image input section 10 includes synchronization signal generation circuits 10a and A.
/D transformation circuit 10b is equipped with an image data storage circuit 10c, and controls the image detection section 6 according to the vertical synchronization signal and horizontal synchronization signal output from the synchronization signal generation circuit 10aJ. The image signals that are sequentially taken in and converted into digital signals by the A/D transformation circuit 10b can be stored in the storage circuit 10c as image data G. This can be easily carried out by simply operating the input section 10 and causing the light emitting section 5 to emit light in synchronization with the vertical synchronization signal output from the synchronization signal generation circuit 10a. In other words, as shown in FIG. 7, first, after a predetermined time Δt has elapsed from the rise time to of the vertical synchronization signal, the irradiation signal S is output to the light emitting unit 5, thereby causing the light emitting unit 5 to emit light in coincidence with the vertical retrace time Δt1. The image obtained by the light emission is read as image data G.

The image data G converted and stored in the image input unit 10 represents the brightness of the pixels that divide the image obtained by the image detection unit 6 into 256.1!240 pixels horizontally, as described above. As shown in FIG. 8, the brightness G (X, V) of each pixel is stored.

In this way, in step 101, the light emitting section 5, the image detecting section 6
And the image input unit 10 is braked. Step 102 that follows when the image data G is read in the image input unit 10
is executed, a predetermined area set in advance in the image data G, that is, the area 30 shown in FIG. Edge detection processing is performed to detect an edge as a predetermined region candidate position. Incidentally, the difference between two points in the horizontal direction of image data can be considered as a standard pattern of a part of a predetermined part, and the difference is calculated and a position where the difference in A is greater than a predetermined value is searched. The searched difference is binarized and image processed to obtain a binary image 31 shown in FIG. 9(D), so that edges can be easily distinguished.

In other words, the main edge detection process starts from the upper left pixel of the nose presence 1430 in FIG. 9(a),
Find the difference in brightness between adjacent pixels in the horizontal direction from left to right until the right end is reached. However, in this embodiment, the pixels used for the difference calculation are every three pixels in the vertical direction and every other pixel in the horizontal direction, as shown by the shaded pixels in FIG. 10 in the nose existing region 30. In other words, only 1/8th the number of pixels is used, which greatly reduces the amount of calculation needed to calculate the difference and speeds up the calculation time. Sufficient pixels can be maintained. Now, let the brightness of the pixel used to calculate the difference as a standard pattern of a part of a predetermined part be GK(i,,i), and the upper left corner of the pixel used for difference calculation is (0) as shown in Figure 10. ,0) and i
If the coordinates are taken such that G increases to the right in the horizontal direction and j increases to the bottom in the vertical direction, the brightness difference GK1 (i, j) can be found by the following formula. However, GK (i, , i) takes values in 16 stages and is expressed as an integer value from 0 to 15.

GKI(i, j> =16x rGK (i+1. j ) +GK (
i.

j ) -GK (i-1,j) -GK (i -
2, j)]-1-16 The above formula is a calculation formula for calculating the difference that makes it easy to detect the experimentally determined edge. The above calculation is performed for each pixel. Next, choose the largest one among the calculated differences,
If the value is greater than a certain value (48 or greater in this embodiment), the search is made to determine that the edge is at that point. Subsequently, a binarization process is performed to obtain a binary image 31 as an edge image. However, as shown in FIG. 11, when there are a plurality of pixels that take the maximum value, the rightmost pixel may be taken as the edge, and in this case, the portion 1-16 becomes the maximum difference. The immediate means of the binarization process is to set the brightness GK2(i) of one section on the binary image to pixels that can be considered as edges.

, j) is set to 1, and GK2(
+, , i) is O. In addition, in the above binary image, a collection of 8 pixels of the original image data (one unit used when calculating the difference) is considered as one section, and all pixels in that section have a constant brightness. This gives a brightness of 1 or a brightness KtO that does not have a constant brightness, and an image as shown in FIG. 9 (b) is obtained.

When the edge is detected in step 102, the next step 103 is executed.

In step 103, a process is performed to select a portion of the edge detected in step 102 that may be a nose, and a nose candidate is detected. In this embodiment, in the binary image obtained in step 102 as shown in FIG. 9(b), a portion having a convex shape to the left is searched, and the pixel at the apex of the convex portion is detected as a nasal tip candidate point. The convex shape to the left above is shown in Figure 12 (a) in the binary image.
Alternatively, it is considered that an image is taken as shown in (b), and in this case, the following logical formula is used.

to GK2 (i, j) (GK2 (i, , i
-1) VGK2 (i +1.j-1)) to (GK2 (i +1., i -2)VGK2 (i
+2. j-2>) In the following step 104, it is determined whether or not the longevity candidate was detected in step 103.

In step 103, 21
In some cases, such as when there is no convex portion to the left on the i-image, the nose candidate cannot be detected, and the process branches to execute a detection failure warning in the subsequent step 107.

The following step 105 is executed when it is determined in step 104 that the nose candidate has been detected = 18
- In step 103, the nose candidate point detected in step 103 is examined to determine the nose position. In this embodiment, all the nasal tip candidate points detected in step 103 are located in the vicinity of the corresponding points in the image data obtained in step 101, and are located in the vicinity of the corresponding points in the image data obtained in step 101, and in the vicinity of the corresponding points as shown in FIG. Examine the standard pattern of the nose as an lli quasi-pattern for the entire region. Examine the degree of phase 1, and consider the position of the image where the correlation is maximum as the position where the nose exists.
This determines the position of the nose.

In other words, the 14th
As shown in the figure, a total of 121 pixels, 5 pixels each on the top and bottom, left and right sides, are applied to the standard pattern of the nose shown in Figure 13.
The correlation value between the image data and the standard pattern is calculated as the nose center position ya (0, 0>), and the pixel position where the value is the largest is recognized as the nose position G (hx, hy). The center position yp (0, O) in the standard nose pattern in Figure 13 is set in advance as a part corresponding to the nose, and the correlation value is given by the following formula (1)%) You can ask for more. The standard pattern of the nose is configured as data with a spread of 16 above, 7 below, 1 on the left, and 6 on the right with respect to the center va (0, 0). Since there is, it becomes 1 (-1)
(+6>1. An integer value between (-16) and (+7) may be used for i.

Also, the correlation value is the center position Im in the standard pattern of the nose.
(0, 0), there are 5
Since the calculation is performed by applying a total of 121 pixels, the values of x and Y in the above equation are as follows:
, Y becomes an integer value between V±5, and in the end, a maximum of 121 types of correlation values are calculated by combinations of x and Y. Note that there may be multiple nasal tip candidate points found in step 103, and in this case, it is necessary to calculate the above-mentioned correlation values for the pixels of all nasal tip candidate points, and the maximum value among all the correlation values is The image data l:ノ1ifii*G
(X, Y) Nose (1) position G (hx, h
y) It is recognized as

In the following step 106, it is determined whether the position of the nose was recognized in step 105.

In step 105, if the nose position G (hx, hy) cannot be recognized, such as when the degree of correlation between the standard pattern of the nose and the image data is too small and there is no part that can be recognized as a nose, it is determined that the nose position G (hx, hy) cannot be recognized in the following step 107. Processing branches to execute the warning.

In step 107, a detection failure warning is issued, and the driver is notified of this through a display or audio display. When the driver receives the warning, it is advisable to take measures such as correcting his or her posture and executing the vehicle driver position recognition process again.

In the following step 108, ]: Step 106
The position of the nose on the pixel data G determined by a < hx
. In other words, assuming that there is no movement of the driver 1 in the left-right (Z) direction shown in FIG. The position of the nose is recognized by determining the three-dimensional position HM of the nose as coordinates (X, Y, O), but since the image detection unit 6 is fixed, this process requires image data Store in advance the digit 1fG (px, py) of the standard point P above,
Nose position G with respect to this reference position NG (px, py)
(hx, hV) (7) By calculating t it, the task of recognizing the position of the vehicle driver's nose in this embodiment, which can be easily executed, is completed.

Note that step 101 corresponds to processing as the image detection means M3, steps 102 to 103 correspond to processing as the predetermined region candidate position detection means M4, and step 105 corresponds to processing as the predetermined region detection means M5. However, step 108 corresponds to processing as the position 121H recognition means M6.

As explained above, in the vehicle driver position recognition device of the present embodiment, when recognizing a predetermined portion of the edge, the minimum pattern as a part of the predetermined portion is used as the minimum pattern between two points in a predetermined direction in image data. By considering the difference and finding the difference, it becomes possible to clearly detect the edges, and if the second image is 1q, it becomes possible to recognize the line parts even more clearly.

In addition, we decided to perform a two-step process to search for the nose tip IQ, which is a convex part at the edge, along with the above-mentioned difference start time.
: A predetermined part (limiting the position where the O detection means is activated, and
Since it is configured to recognize the entire predetermined part using a standard pattern, the position of the predetermined part is not erroneously recognized, and therefore the riding position can be recognized quickly and accurately. Become.

Next, a second embodiment of the present invention will be described with reference to the drawings.

In this embodiment, the three-dimensional position of the vehicle driver's 1 eyes is determined based on the position of the vehicle driver's 1 nose recognized in the first embodiment.

Regarding the operation of the position recognition process executed in this embodiment,
This will be explained in detail along the flowchart shown in FIG. Note that steps 201, 202, 203, 204, 205, 206, and 207 in this embodiment are the same as steps 101, 102, 103, 104, 105, and 106 in the first embodiment. and step 107, so the explanation will be omitted.

By sequentially executing steps 201, 202, 203, 204, 205, 206 and 207, the position of the nose
Once (hx, hy) is determined, the following step 208
, and now this nose position G (hx, hy)
Based on this, the position of the eyes is determined. To find the position of the eyes, first calculate the position of the nose G (hx, hy
), find the pixel position G (hx+12, tty-13) 12 pixels to the right and 13 pixels above, and use the standard eye pattern as shown in Fig. 17, which is set in advance near this position. The degree of correlation with the image data is examined, and the eye position Q(n+x) is determined.

my). Note that this determination method can be performed in the same manner as the nose position determination in step 205 above.
Explanation will be omitted.

In the following step 209 - the above step 208
The eye position G (ax,
n+y), the three-dimensional position of the eyes of the driver 1 in the vehicle interior is determined, assuming that the driver 1 does not move in the left-right direction. Note that the operation is the same as in step 108 of the first embodiment if the position of the nose is considered as the position of the eyes, so the explanation will be omitted.

As explained above, this embodiment has the same effect as the first embodiment, and is also easier to detect on the image data instead of directly determining the position of the driver's eyes from the image data. The position of the prominent nose is recognized, and the position of the eyes is then recognized from the correlation between the nose and the eyes, making it easy to adjust the riding position.
(the position of the left eye) can be recognized. In addition, when calculating the position of the eye, it is configured to use the S+ single pattern, so the three-dimensional position of the eye may be mistakenly recognized m'! It is possible to recognize the riding position more accurately.

Another method for calculating the eye position is to first recognize the position of the nose by performing edge detection processing in the same way as in this embodiment, and then predict the position of the chin from the position of the nose and detect the position of the chin in the vicinity. The position of the chin is calculated using a standard pattern, and the specific relationship between the two positions and the position of the eyes (average data created by taking images of multiple vehicle drivers in advance) is used to calculate the position of the vehicle driver. The position of the eyes may be calculated. This method makes it possible to correctly recognize the position of the eyes even when the eyes cannot be detected, such as when the vehicle driver 1 is wearing glasses or the like.

Furthermore, in this embodiment, the position of the eyes of the driver 1 (more precisely, the position of the left eye) is recognized as the riding position of the vehicle driver, but the position of other parts such as the nose and chin is recognized by the vehicle. The three-dimensional position may be recognized as the driver's position, and the processing after recognizing the vehicle driver's position, such as automatically adjusting the rearview mirror and automatically adjusting the steering angle, etc. J: The most appropriate riding position may be recognized according to the situation. Incidentally, when the position of the eyes of the driver 1 is recognized as in this embodiment, it can be used to adjust the angle of the rearview mirror or to detect drowsy driving.

In addition, in the first and second embodiments, the pixels used to calculate the difference as the minimum pattern as a part of a predetermined region in the edge detection process of the nose region are set every three pixels in the vertical direction. Instead of every other pixel in the horizontal direction, you can change it to two pixels horizontally and change the spacing between the pixels used. In this case, although the horizontal interval is increased, the pixel interval calculated when calculating the horizontal difference becomes rougher, so the edge image has a rougher grain in the horizontal direction than in the first and second embodiments. However, since the vertical spacing is small, in the vertical direction,
A finer-grained edge image is created. Also, the first
In the embodiment and the second embodiment, 1 out of 8 pixels of the original image is used as the pixel used to calculate the difference as the minimum pattern as a part of a predetermined region in the edge detection process.
Instead of using pixel division, the sum of every 8 pixels of the original image may be calculated, and the difference may be found between the sums of each 8 pixels. If one of the eight pixels happens to have a large brightness due to the influence of external light, there is a possibility that an incorrect difference will be obtained in the first and second embodiments, but in this case, the sum of the eight pixels Errors are averaged to obtain a more accurate difference, which results in more accurate edge detection. Alternatively, the difference may be obtained by summing two pixels in the horizontal direction among the eight pixels, thereby making it easy to start calculation and detecting accurate edges.

Furthermore, in the first embodiment and the second embodiment, the logical formula for nose candidate detection in the nose candidate detection process is GK2 (i,
j ) to (GK2 (i , , i −1) GK2
(+ +1. j −1> ) to (GK2 (i +1
, j −2) VGK2 (i +2., i 2)
), but the shape of the nose is as shown in Figure 18 (a).

If you want to take any binary image of (b), (c), (d), (e), or (f), you can use the following logical formula to convert GK2 (i, j) to (GK2 ( i +1.
j −1) VGK2 (“i +2. j 1 )
) Other logical formulas assuming other shapes that can be considered as noses may be used. Alternatively, a nose candidate portion may be detected by comparing the detected binary image with a standard pattern of the edge of the nose 211C1. The comparison means checks the degree of correlation in the same way as in the position recognition process of the ten noses, and is performed at the position of the image where the correlation is the maximum value and has a value greater than a certain value.

Further, in the first embodiment and the second embodiment, the predetermined part candidate position detection means M4 calculates the difference as a minimum pattern as a part of the predetermined part of the image data detected by the image detection means M3. The pattern search consists of a two-step pattern search in which the edge is detected, and then the edge is searched using a logical formula representing the shape to detect the candidate position of the nasal tip. The position of the tip of the nose as a candidate position of the predetermined part may be detected in one step by searching the image data for a position corresponding to the pattern based on a standard pattern of the nose tip of a person.

Furthermore, when the edge is detected by the predetermined part candidate position detecting means, the position of the nose is detected by completing the pattern search in one step and directly executing the predetermined detecting means, instead of subsequently finding the tip of the nose. You may also do so.

Further, in the first embodiment and the second embodiment, the light emitting unit M2 described above includes the light emitting unit 5, the image detecting unit M3 includes the image detecting unit 6, and the predetermined part candidate position detecting unit M4J the predetermined part detecting unit. M5 and IIg recognition means M6 include an image processing section 8.
In addition to the image detecting unit 6 made of a MOS type solid-state image sensor as in the above embodiment, the image detecting means M3 may be an image sensor used in a so-called solid-state camera such as a COD, for example. An interlaced type may also be used.

In addition, in the first embodiment and the second embodiment, there is one uniformity detection unit 6, and the J-uniformity detection unit 6 recognizes the riding position of the driver using image data captured from one direction. The driver's movement in the left and right direction is not indicated by the blue line, but the image detection unit is different from 11'? If the configuration is configured to recognize the driver's riding position using the so-called triangulation principle, it is possible to detect the driver's movement in the left and right directions, and the driver's three-dimensional position can be detected. J: Be able to recognize more accurately.

In addition, in the first embodiment and the second embodiment, the light emitting means M2
Although an infrared strobe is used as a flashlight, a continuous light such as an incandescent bulb may also be used.

[Effects of the Invention] As detailed above, the vehicle driver position recognition device of the present invention detects the driver's face based on a standard pattern of a predetermined part of the driver's face from image data obtained by the image detection means. The vehicle driver is constructed by detecting the candidate position of the predetermined part of the body, and detecting the predetermined part by using the image data in the vicinity of the candidate position of the predetermined part. It automatically adjusts the position and angle of mirrors, headrests, air-conditioned air outlets, steering wheels, etc., or detects drowsiness while driving based on periodic changes in the position of the driver's eyes and head and issues a warning to the driver. It is useful for such things.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing the configuration of the present invention, FIGS. 2 to 18 show embodiments of the present invention, and FIG. 2 shows an instrument panel of a vehicle and its surroundings to which the embodiment of the present invention is applied. FIG. 3 is an explanatory diagram of the light emitting unit 5,
FIG. 4 is an explanatory diagram of the image detection section 6, FIG. 5 is an overall configuration diagram of the riding position recognition device, and FIG. 6 is an illustration of the image processing section 8 of the first embodiment.
7 is a flowchart showing the operation of image data G.
FIG. 0 is an explanatory diagram showing the irradiation image data G stored in the image input unit 10, FIG. 9(A) is an image diagram of the image data G, FIG.
b) is an edge image diagram after performing edge detection of the nose existing region; Figure 10 is an explanatory diagram of pixels used for difference calculation;
The figure is an explanatory diagram explaining how to select the maximum value of the obtained difference, and Figures 12 (a) and (b) are explanatory diagrams for detecting nose candidates.
Figure 13 is a data diagram representing the standard pattern of the nose, Figure 14
15 is an explanatory diagram of the nose position detection process, and FIG. 15 is an explanatory diagram of the three-dimensional position of the nose. 16 is a flowchart showing the operation of the image processing unit 8 of the second embodiment, FIG. 17- is a data diagram showing the standard eye pattern, and FIG.
Figures (A), (B), (C), (D), (E), and (F) are explanatory diagrams for detecting nose candidates of other shapes. M2... Light emitting means M3... Image detecting means M4... Edge detecting means M5... Position recognition means 5... Light emitting section 6... Image detecting section 8... Image processing section 10. ...Image input section 14...CPU

Claims (1)

[Scope of Claims] 1. A light emitting means for illuminating a vehicle driver; an image detecting means for receiving reflected light from the driver caused by the illumination of the light emitting means and detecting the driver as image data; and an image detecting means for detecting the driver as image data. a predetermined part candidate position detection means for detecting a predetermined part candidate position of the driver's face based on a standard pattern of a part of the predetermined part of the driver's face; and the entire predetermined part from image data in the vicinity of the predetermined part candidate position. a predetermined part detection means for detecting a predetermined part based on a standard pattern of the driver; and a position recognition means for recognizing the three-dimensional position of the driver based on the predetermined part detected by the predetermined part detection means. Characteristic vehicle driver position recognition device. 2. The predetermined part candidate position detection means determines the difference between two points in the image data as a standard pattern of a part of the predetermined part and searches for a position where the difference is greater than or equal to a predetermined value, thereby detecting the edge as the predetermined part candidate position. A vehicle driver position recognition device according to claim 1, which detects the position of a vehicle driver. 3. The predetermined part candidate position detecting means searches for a pattern corresponding position from image data based on a standard pattern of the driver's nose tip as a standard pattern of a part of the predetermined part, thereby determining the position of the nose tip as a predetermined part candidate position. , and the predetermined part detection means detects the position of the nose as the predetermined part by searching for a position corresponding to the pattern from image data in the vicinity of the predetermined part candidate position based on the standard pattern of the nose as a standard pattern of the entire predetermined part. A vehicle driver position recognition device according to claim 1.