KR100686564B1 - Apparatus for detecting occupant's posture - Google Patents

Abstract

An object of the present invention is to simplify and speed up the attitude determination process of a rider.

A occupant sensor 1 which forms an image of the occupant 2 with at least one pair of linear light sensor arrays composed of a plurality of optical sensor elements, for example, on the ceiling of the vehicle 3, and at least one of the occupant images A distance measurement arithmetic processing unit 101 for measuring a distance distribution in a linear visual field and a rider discrimination processing device 102 for determining the presence or absence of a rider by comparing the distance distribution with a standard distance distribution are provided. By focusing on the characteristics (similarity) of the intra-view distance distribution by the apparatus 102, the object of pattern matching is compressed, and the discrimination process is simplified and speeded up especially.

Description

Rider's posture determination device {APPARATUS FOR DETECTING OCCUPANT'S POSTURE}

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a device for determining the presence or absence of a rider, in particular an automobile rider, using a linear optical sensor array comprising a plurality of optical sensor elements. In particular, the present invention detects a vehicle impact and deploys an airbag installed near a seat. The present invention relates to an occupant's attitude determination device suitable for use in control of time.

Airbag installation rates have been increasing in recent years, and airbags are becoming standard equipment regardless of vehicle type. However, airbags have been reported to cause death if the child is standing in front of the seat or a small female is sitting. For this reason, various deployment control of airbags for preventing such a death accident has been proposed. Among them, a conventional method of installing a sensor at a ceiling position in a vehicle to measure a distance from the sensor to a rider and detecting the presence or absence of a rider from the distance distribution (US patent application filed on June 20, 1997) Reference is made to the embodiment disclosed in 08 / 879,821).

FIG. 21 is a schematic view for explaining an embodiment of US Patent Application No. 08 / 879,821 of this reference document, where 1 is a rider sensor, 2 is a rider, and 3 is a car.

This forms an image of the occupant 2 in the occupant sensor 1. Here, for example, four linear viewing regions R1, R2, R3, and R4 are set for the occupant 2, and a plurality of portions are provided for each region. Obtain the occupant sensor output, calculate the output with a processing device not shown, measure the distance of each part of the occupant to obtain a distance distribution for each inspection area, and from this distance distribution to the processing device, The attitude is also judged. The principle of the distance measurement will be described later.

By the way, various attitudes such as a rider to be discriminated can be considered. However, in FIGS. 22 (a) to (c) and (a) 'to (c)' to FIG. Relationship is shown, and the distance measurement example by a rider sensor is shown to FIG. 25 (a)-(c)-FIG. 27 (a) and (b).

Thus, FIG. 22A is a perspective view of an unmanned seat, (a) is a side view thereof, (b) is a normal attitude perspective view of a car rider, (b) is a side view thereof, and (c) is a front view of a car rider. (C) is a side view thereof, FIG. 23 (a) is a perspective view when the front child seat is mounted, (a) is its side view, and (b) is a perspective view when the rear child seat is mounted. (b) is the side view, (c) is the case of the standing child in which the child stands in the car, and (c) is the side view, and FIG. 24 shows the sitting position sideways, respectively. In addition, the side view of FIG. 24 is abbreviate | omitted since it is the same as the side view of a normal posture, ie, FIG. 22 (b) '.

25A to 25C show distance measurement examples corresponding to FIGS. 22A to 22C, and FIGS. 26A to 26C show FIGS. 23A to 23C. A corresponding distance measurement example and FIG. 27A shows a distance measurement example corresponding to FIG. 24, respectively. In addition, FIG. 27 (b) has shown the distance measurement example at the time of unattended by mounting a forward child seat. 25 (a) to 25 (c) to 27 (a) and (b), the vertical axis represents the distance value from the sensor and the horizontal axis represents the position in the field of view, respectively. Esau is not specifically defined, but the same.

That is, it turns out that the distance value of each linear viewing area | region R1, R2, R3, R4 is obtained as a discrete value, respectively. The concept of obtaining such discrete distance distribution will be described later.

By the way, it is common to determine the presence or absence of a vehicle rider and the attitude by so-called pattern matching which compares the distance distribution for each field of vision obtained as described above with a standard pattern. However, since the standard patterns vary depending on the presence or absence of the vehicle rider, the attitude, and the distance from the reference position of the car seat or the inclination, the standard patterns require a lot of processing time in comparison with the standard patterns. have.

Accordingly, an object of the present invention is to reduce the processing time for the presence or absence and the posture determination, in particular the processing time for pattern matching.

In order to solve this problem, in the present invention, a sensor having at least one pair of linear optical sensor arrays composed of a plurality of optical sensor elements to form an image appearing at another position of the occupant, and the occupant formed by the sensor A distance measurement arithmetic processing unit for measuring a distance distribution in each linear viewing area monitored by the at least one pair of linear light sensor arrays from an image, extracting an abstract pattern from the measured distance distribution, and determining the occupant A plurality of model patterns stored in advance in the processing apparatus are compared with the abstract patterns to provide a rider discrimination processing apparatus that determines a specific posture of the rider or the subject.

Further, in the present invention, the at least one linear viewing region is set to extend in a substantially horizontal direction with respect to the occupant, and the occupant discrimination processing apparatus abstracts by checking whether the distance distribution in each of the linear viewing regions is symmetrical. After extracting the pattern and comparing the plurality of preselected model patterns with the abstract pattern according to whether the distance distribution has symmetry, the model pattern in the narrow range is selected, and then the specific posture of the rider or the object is determined.

Further, in the present invention, the left and right symmetry of the distance distribution inverts the distance distribution pattern in each of the linear viewing areas with respect to the center portion thereof, thereby inverting the distance distribution pattern before being inverted in each of the linear viewing areas. The distance difference at the corresponding position with the subsequent distance distribution pattern is obtained, and it is checked by determining whether each said distance difference is not larger than a predetermined value or whether the integrated value of the distance difference is not larger than a predetermined value.

In addition, in this invention, the said linear viewing area | region is set so that it may extend in substantially horizontal direction with respect to a rider, and the said rider discrimination processing apparatus correlates between each visual field regarding the change of the distance distribution pattern between adjacent areas of the viewing area. The abstract pattern is extracted by checking the figure, and the model pattern of the narrow range is selected by comparing a plurality of model patterns previously selected with the abstract pattern according to the correlation, and then the specific attitude of the rider or the subject is determined.

In the present invention, the correlation between adjacent viewing areas in the linear viewing area is obtained by calculating a difference of changes in the distance distribution pattern in each viewing area at a corresponding position of adjacent viewing areas in the viewing area. It is checked by determining whether the difference is not larger than a predetermined value or whether the integrated value of the difference is not larger than a predetermined value.

Further, in the present invention, the linear viewing region is set to extend in a substantially horizontal direction with respect to the occupant, and the concave or convex pattern of the distance distribution between the first viewing position and the second viewing position of each of the viewing regions is set. By checking, the abstract pattern is extracted, and the model pattern in the narrow range is selected by comparing a plurality of preselected model patterns with the abstract patterns according to whether the abstract pattern and the distance distribution pattern are concave or convex, and then the passenger or The specific posture of the object is determined.

Further, in the present invention, the concave or convex shape of the distance distribution pattern is an average value of the distance values at the first viewing position and the second viewing position in each of the linear viewing regions, and the first and second positions. It discriminates by comparing the distance value in the intermediate position with.

Further, in the present invention, the linear viewing region is set to extend in a substantially horizontal direction with respect to the occupant, and the occupant discrimination processing apparatus measures at a predetermined position of each linear viewing region in comparison with the distance value in the adjacent viewing region. Extracting an abstract pattern by checking a trend of increasing or decreasing the distance, selecting a model pattern in a narrow range by comparing the abstract pattern with a plurality of pre-selected model patterns according to the trend of increasing or decreasing the distance, and then specifying the passenger or the target Determine your posture.

In addition, in the present invention, the predetermined position may be the center position of the rider sitting in the normal posture when viewed in the lateral direction.

Further, in the present invention, the linear viewing region is set to extend in a substantially horizontal direction with respect to the occupant, and the occupant discrimination processing apparatus is configured based on a comprehensive judgment regarding the symmetry of the distance distribution pattern in each of the linear viewing regions. Correlation between the occupant's posture, the concave or convex shape of the distance distribution pattern, and the variation of the distance distribution pattern, and the correlation between adjacent viewing areas in the at least one linear viewing area, and measured at a predetermined position of the linear viewing area. The trend of increasing and decreasing the distance is determined.

Further, in the present invention, the linear viewing region is set to extend in a substantially vertical direction with respect to the occupant, and the occupant discrimination processing apparatus is a distance distribution pattern of a pair of viewing regions at positions substantially symmetric with respect to the deployment axis of the airbag. The abstract pattern is extracted by checking the degree of correlation between them, and the specific posture of the occupant or the object is determined by comparing the abstract pattern with a plurality of preselected model patterns according to the degree of correlation.

In addition, in the present invention, the correlation between the distance distribution patterns of the pair of viewing areas may be compared with a predetermined value or the integrated value of each difference measured at a position corresponding to the pair of viewing areas. It is determined by comparing with politics.

Further, in the present invention, the rider discrimination processing apparatus usually measures only the distance distribution in the field of view related to the deployment axis of the airbag to determine a rough posture, and at the time of a collision, the rider's posture including the coarse posture discrimination. The discrimination process is given priority over other urgent processes.

In the present invention, at least one pair of occupant attitude determination apparatuses for use in an airbag system having an airbag housed in a receiving portion provided near a seat of a vehicle, and a sensor for detecting an impact of the vehicle and deploying the airbag. A distance sensor for detecting a distance distribution in a plurality of portions of a rider or a subject existing in a space between the airbag accommodation portion and the seat that accommodates the airbag, and a pattern and distance distribution of the distance distribution; By comparing a plurality of model patterns, the posture of the rider or the shape of the object is determined, and according to the determination result, the deployment of the airbag is controlled or whether the airbag should be deployed.

That is, according to the present invention, since the standard distance distribution is previously classified into a plurality of groups by its symmetry, the distance distribution in the visual field is classified according to the grouping, and then the pattern is compared with the group of the standard distance distribution. The object for matching can be compressed and the processing time for attitude determination can be shortened.

1 is a block diagram showing the entire apparatus to which the present invention is applied.

That is, the occupant sensor 1 is provided in the ceiling center part of the motor vehicle 3, and the distance measurement arithmetic unit 101 and the occupant discrimination apparatus 102 are connected to the output of this occupant sensor 1, and is comprised. Reference numeral 2 denotes a rider.

That is, the image of the occupant 2 is imaged on the occupant sensor 1, but as shown in FIG. 21, four linear viewing regions R1, R2, R3, and R4 are set for the occupant 2 as shown in FIG. The occupant sensor output of a plurality of areas is obtained for each area, the output is calculated by the distance measuring arithmetic processing unit 101, and then compared with a standard pattern stored in advance in the occupant discrimination processing apparatus 102, and the like. Not only the presence but also the attitude to judge.

The rider's attitude signal determined by the rider discrimination processing apparatus 102 is input to the central processing unit 201 simultaneously with the output signal of the seat sensor 204, for example. In the central processing unit 201, when the shock detection sensor 202 detects an impact, the deployment command of the airbag is determined based on the previous passenger attitude information and the seating information of the passenger from the seat sensor 204. The command is instructed to an inflator 203 for activating the airbag in the accommodation portion provided near the seat.

FIG. 2 shows an example of a occupant sensor used in FIG. 1.

That is, in this embodiment, the occupant sensor 1 is configured by integrating the multi-stage light receiving IC 10 and the imaging lenses 21 and 22 formed by providing four pairs of optical sensor arrays. In addition, although at least one pair of optical sensor array pairs may be sufficient, in this embodiment, the example which set it in four stages in order to set several linear viewing area | regions R1, R2, R3, R4 is shown. 2 denotes an auxiliary light source, and V denotes a viewing center direction of the sensor.

3 shows a specific example of the distance measurement arithmetic unit 101.

That is, it is comprised of the registers 1011 and 1012 as memory, the partial data extraction part 1013, the correlation detection / distance calculation part 1014, the controller 1015 which controls these parts, etc. Hereinafter, the distance measuring method by the distance measurement arithmetic processing apparatus 101 is demonstrated. First, the principle will be described with reference to FIG.

As shown in Fig. 4, the horizontal axis X and the vertical axis Y are set with the center of the imaging lenses 21 and 22 as the origin 0, and the imaging positions by the respective lenses are set to (L ₁ , R ₁ ). In addition, the coordinates of the center O _L of the imaging lens 22 is (-B / 2, 0) coordinates of the center point O of the _R, and the imaging lens 21 (B / 2, 0). The coordinate of the point M of the object (subject) is (-x, y), and the position where the straight line parallel to the straight line M0 intersects with the optical sensor array 12 through the point O _L is through L ₀ , the point O _R. The position where the straight line parallel to the straight line M0 intersects with the optical sensor array 11 is called R ₀ .

Here, a _L represents the distance between the point L ₀ and the point L ₁ , and a _R represents the distance between the point R ₀ and the point R ₁ . L ₀ and R ₀ are reference positions when a _L and a _R are obtained. At this time, Δ M0 _L 0 and Δ0 _L L ₁ L ₀ , ΔM0 _R 0 and Δ0 _R R ₁ R ₀ are similar to each other,

[Equation 1]

(-x + B / 2) f = a L · y

[Equation 2]

(x + B / 2) f = a R · y

Is established. Eliminating x from Equations 1 and 2,

[Equation 3]

y = B · f / (a _L + a _R )

Becomes If the sensor pitch is p and the number of sensors is x, a _L + a _R = p.x.

[Equation 4]

y = Bf / (px)

It is also expressed as In other words, knowing the imaging position L ₁ and point L ₀ between the distance a _L and a image forming position R ₁ and the distance a _R of the point R ₀ of the optical sensor array 11 to the right of the optical sensor array 12 to the left, the The distance y from the equation 3 to the object can be obtained. In addition, this distance measurement principle itself is known.

In order to determine the relative positions of the two images as described above, the above-described correlation calculation by the ranging measurement processing apparatus 101, for example, is performed.

The output of each element of the optical sensor arrays 11 and 12 on both sides is, for example, converted into an 8-bit digital signal, and then stored in registers 1011 and 1012 as a memory shown in FIG. Are calculated by the partial data extracting unit 1013. As shown in Fig. 5, the calculation regions 111 and 112 are composed of n elements, and each of the calculation regions 111 and 112 is composed of quantized data of (A ₁ to A _n ) and (B ₁ to B _n ).

Herein, the quantization data is used to obtain the distance index a _L + a _R from the front of the sensor to an object at a predetermined angle (an angle formed by a straight line M0 connecting the Y axis, the object coordinate M, and the origin 0 in FIG. 4). The correlation detection / distance calculation unit 1014 shown in FIG. 3 sets windows W ₁ to W _{m + 1} of predetermined sizes as shown in FIG. 5, and alternately shifts by one sensor unit (1 bit). Considering the subset (C ₀ -C _m ) of the (m + 1) pair, the evaluation functions f (C ₀ )-f (C _m ) consisting of the sum of the absolute values of the differences of the quantization data for each subset are calculated, By finding the combination C _k at which the evaluation function value is the minimum, the state in which the left and right phases deviate from the value of the subscript k can be known, which is proportional to (a _L + a _R ) as expressed by Equation 3 above. It is known that distance indicators can be obtained (for example, Japanese Patent No. 2676985, large). See U.S. Patent No. 5,602,944 and corresponding German Patent No. 4121145 No.).

Thus, for example, as shown in Fig. 6, the calculation regions 111 and 112 are sequentially shifted by one bit (1 pitch: p), and the distance indices as described above for each calculation region 111 and 112 are obtained. By performing the processing, s discrete distance data can be obtained in the first to the sth calculation areas.

7 is a block diagram showing a specific example of the occupant discrimination processing apparatus 102. The occupant discrimination apparatus 102 of this specific example includes four discrimination units 103-106 for selecting a narrow model pattern to be compared with the distance distribution pattern viewed according to another discrimination algorithm, and this discrimination unit 103-106. The reference value memory 107 stores a plurality of model patterns each forming a target shape or a specific posture of the occupant for comparison with any combination of the output values AD of the reference value and the reference value memory 106 for pattern matching. And a comparator 108 for generating a result of pattern matching by comparing a combination of the stored data and the output values AD of the discriminating units 103-106. Here, the occupant discrimination processing apparatus 102 including such discriminating units 103-106 is used, and a plurality of linear viewing areas monitored by the optical sensor array 12 of the occupant sensor 1 are supplied to the occupant 2. It is generally set or formed in the horizontal direction with respect to it.

The determination unit 103 includes a distance distribution symmetry determining device 1031, a reference value memory 1033, and a symmetry / reference value comparator 1032.

The distance distribution symmetry discriminating apparatus 1031 receives information about the distances in each of the linear viewing areas R1, R2, R3, and R4, detects whether the left and right symmetry of the distance distribution is related to each of the linear viewing areas, The results relating to the left and right symmetry in the plurality of viewing areas R1, R2, R3, and R4 are output in the order of R1, R2, R3, and R4. For example, the result obtained by the discriminating device 1031 may be patterned or coded as "0110" (where "0" indicates symmetry and "1" indicates no symmetry) and the coded Output in the form On the other hand, the reference value memory 1033 stores in advance a plurality of model symmetry patterns relating to various postures of the passenger or various shapes of the object. The symmetry / reference value comparator 1032 compares the outputs of the distance distribution symmetry discrimination apparatus 1031 with a plurality of model symmetry patterns previously stored in the plurality of reference value memories 1033, and narrows the rider's attitude or the shape of the object. An output A indicating the selection result of the model pattern of the range is output to the comparator 108. The algorithm used by the discriminating unit 103 corresponds to the above-described "discerning about the left and right symmetry of the distance distribution (first embodiment)".

The discriminating unit 104 includes an interline correlation calculator 1041, a reference value memory 1043, and a correlation value / reference value comparator 1042. The inter-line correlation calculator 1041 receives information about the distance distribution in each of the linear viewing areas R1, R2, R3, and R4, and the adjacent linear viewing area by the change in the distance distribution in each viewing area, that is, The correlation is detected between (1) R1 and R2, (2) between R2 and R3, and (3) between R3 and R4, and the results are output in the order of (1), (2) and (3). For example, the result obtained by the discriminating apparatus 1041 is patterned or coded as "100" (where "0" indicates symmetry and "1" indicates no symmetry), and the coding form thereof. Is caused by. On the other hand, the reference value memory 1043 stores in advance a plurality of model correlation patterns relating to various postures or types of objects of the rider. The correlation value / reference value comparator 1042 compares the plurality of model correlation patterns stored in the reference value memory 1043 with the outputs of the line correlation discrimination apparatus 1041, and models a narrow range of models regarding the occupant attitude or the shape of the object. An output B indicating the result of the selection is output to the comparator 108. The algorithm used by the discriminating unit 104 corresponds to the " determination on correlation between adjacent lines (second embodiment) ".

The discriminating unit 105 includes a distance distribution unevenness determining unit 1051, a reference value memory 1053, and an unevenness / reference value comparator 1052. The distance distribution unevenness determining unit 1051 receives information about the distance in each of the linear viewing regions R1, R2, R3, and R4, and the first viewing position and the first viewing position in the viewing region opposite to the center position in the viewing region. Based on the distance distribution between the second viewing positions, any unevenness of the distance distribution is detected with respect to each linear viewing region, and the results regarding the unevenness of the plurality of viewing regions R1, R2, R3, and R4 are displayed. Outputs in order of R1, R2, R3, R4. For example, the results obtained through the discriminating device 1051 are patterned or coded as "0120", where "0" represents a flat shape, "1" represents a convex shape, and "2" represents a concave shape. Indicates. On the other hand, the reference value memory 1053 stores in advance a plurality of model uneven patterns related to the shape of each object and the attitude of each occupant. The unevenness / reference value comparator 1052 compares the output of the distance distribution unevenness determination device 1051 with a plurality of standard unevenness patterns previously stored in the reference value memory 1053, and narrows the range of the occupant posture or the shape of the object. The output C indicating the selection result of the model pattern is output to the comparator 108. The algorithm used by this discriminating unit 105 corresponds to the "unevenness determination (third embodiment)" described above.

The discriminating unit 106 includes an extractor 1061, a reference value memory 1063, and an intermediate distance / reference value comparator 1062 for extracting a distance at an intermediate position in the viewing area. The extractor 1061 receives information about the distance in each of the linear viewing areas R1, R2, R3, and R4, and detects the increase and decrease of the distance value at an intermediate position in each viewing area compared to that in the adjacent viewing area. The results of the distance increase and decrease in each of the linear viewing areas R1, R2, R3, and R4 are output in the order of R1, R2, R3, and R4. For example, the results obtained by extractor 1061 can be patterned or coded as "+,-,-,-", where "+" indicates an increase and "-" indicates a decrease. On the other hand, the reference value memory 1063 stores in advance a plurality of model increase / decrease patterns related to various postures or types of objects of the rider. The intermediate distance value / reference value comparator 1062 compares a plurality of model increase / decrease patterns previously stored in the reference value memory 1063 with the outputs of the extractor 1061, and a narrow range model pattern relating to the rider's posture or object shape. An output D indicating the result of the selection is output to the comparator 108. The algorithm used by the discriminating unit 106 corresponds to the above-described "decrease / decrease determination of the distance between lines (fourth embodiment)".

As described above, the comparator 108, which receives an output AD different from the discriminator 103-106, initially outputs the output D of the discriminator 106 as a result of the discrimination regarding the increase or decrease of the distance between the lines. Receive and compare the model pattern stored in the reference memory 107 with the characteristic pattern of the output D to generate a model pattern group that matches the characteristic pattern. This data is generated as attitude information if it can be specified from such a group of model patterns, and if a specific result cannot be obtained only from this output D and a plurality of model patterns correspond to the characteristic pattern, a corresponding model increase / decrease pattern Are classified into one group, and "discrimination about irregularities " is performed with respect to this model pattern group. That is, based on the output C of the discriminating unit 105 as a result of " discrimination concerning irregularities " Appear as Any model uneven pattern matching the uneven pattern is extracted from the selected model increase / decrease pattern group as described above.When a specific model uneven pattern is obtained, the pose of the rider or the object associated with this pattern is generated as the posture information. However, if a particular result still cannot be obtained by "determination regarding unevenness", the model unevenness pattern corresponding to the output (C) pattern is classified into 1 group, and one specific pattern is related to "correlation between lines". As a result of the discrimination ", it is extracted from said group 1 based on the output B of the discrimination unit 104. If necessary, the output A of the discrimination unit 103 as a result of the" discrimination concerning the left-right symmetry of the distance distribution ". ) Can also be used to determine the rider's posture.

FIG. 28 shows the discrimination results according to the four discrimination algorithms described above with respect to each of the rider's postures or object shapes. 29-34 show the difference in the determination result according to the change of a sheet position in detail. In the case of "adult (lean forward) / sheet position:" 0 ", for example, a narrow increase / decrease pattern result is" 9 "(the meaning of this number is described later), but this increase / decrease pattern is shown in FIG. In the case of the "child seat (rear position) / seat position:" 0 "shown and the" standing child / seat position: "0" shown in Fig. 34, "9" is also indicated. It will be understood that it cannot be determined only by the judging section 106 which performs "determination on the increase or decrease of the distance between lines." The result of the "unevenness determination" performed through the judging section 105 for these three cases is If considered, "0100" is obtained as a result of "unevenness determination" in the case of "adult (lean forward) / seat position: 0", whereas "0120" and "0102" are "child" in FIG. Seat (rear position) / seat position: "0" and the standing child / seat position: "0" of FIG. 34 It will be understood that these three cases can be clearly distinguished from each other on the other hand, on the other hand, the "unattended / seat position shown in Fig. 29: + 75mm and the" child seat "shown in Fig. Forward position) / sheet position: + 75mm means "determination on increasing / decreasing distance between lines" outputting output D, "unevenness determination" outputting output C, and "line outputting output B" The two cases cannot be mutually discriminated by the four discrimination algorithms, because the same result is obtained in the determination of the correlation between the " and the determination of the symmetry of the distance distribution of the output A ". This is an extremely limited case and if necessary the cases can be mutually discriminated by comparing the distance values in the linear viewing area R3.

In the above-described embodiment, the comparator 108 initially processes the output D as a result of "determination regarding increase or decrease of distance between lines", processes the output C as a result of "determination of unevenness", and the "line" Process the output B as a result of the "determination on correlation between" and output A in the order described, i.e. in the order of D, C, B, A Process. However, comparator 108 processes the outputs A-D in a different order, and in many cases the two outputs are particularly sufficient to specify a particular pattern, among others. What is important here is that the "determination regarding the increase or decrease of the distance between lines" outputting the output D is first performed, and this "determination regarding the increase or decrease of the distance between lines" and at least one of the other three discrimination algorithms must be combined together. will be. From the results of the experiments conducted so far, the inventors of the present invention have made the "determination regarding increase and decrease of the distance between lines" performed by the discriminating unit 106 is the easiest method for narrowing the range of the model pattern for pattern matching, and the attitude of the rider. Or, when determining or determining the shape of the object, it was confirmed that it is very effective in reducing the processing time for the determination.

Although the occupant sensor 1 is attached to the ceiling of the vehicle and the viewing area of the sensor 1 is set in the horizontal direction with respect to the automobile seat in the illustrated embodiment, the present invention is not limited to this apparatus. For example, the occupant sensor can be attached to a front mounted room mirror, which is visible to the driver, or can be attached to a dashboard mounted in front of the passenger.

Next, the content of the discrimination process of each of the discriminating units 103-106 is determined by the discriminating unit 104, " a discrimination on the left and right symmetry of the distance distribution (first embodiment) " "Determination on Correlation Between Lines (Second Embodiment)", "Determination on Unevenness (Third Embodiment)" Determined by 105, " Lines Performed by Determination Unit 106 Will be described in detail in the order of " discrimination (fourth embodiment) relating to the increase / decrease in the liver ".

In the present invention, the following processing is performed on the distance distribution data thus obtained.

That is, as can also be seen in FIG. 25, the airbag is the same as the normal posture of FIG. 25B, the forward inclined posture of FIG. 25C, the forward child seat of FIG. 26A, and the rearward child seat of FIG. 26B. It is roughly symmetrical from side to side with respect to the development axis (in the present embodiment, the center position of the viewing range), and asymmetrical with respect to the side sitting position as shown in Fig. 27A. Therefore, in the present invention, first, the attitude is determined by paying attention to the symmetry of the distance distribution data for each field of view.

In order to focus on the symmetry of the distance distribution data, for example, as shown in FIG. 8A, with respect to the distance distribution data for each field of view, the distribution as shown in FIG. 8B in which this is inverted in the field of view is obtained. The difference is obtained as shown in Fig. 8C. Similarly, as shown in FIG. 9A, the inversion distribution FIG. 9B and FIG. 9C, which is the difference between them, are obtained for the distance distribution data for each visual field.

As for the difference obtained in this way, the symmetry can be compared with a predetermined threshold TH, or an integrated value of the positive (+) side or the negative (-) side of the difference is S, and this is compared with the predetermined threshold THS. Determine. It is the determination unit 103 that performs this process.

(A)-(c) of FIG. 8 is a case where it is symmetrical, and the integrated value of the difference and each difference for every viewing position is simultaneously below a predetermined value. 9 (a) to 9 (c) show a side sitting position as shown in FIG. 24, which is asymmetrical, and the difference between each viewing position and the integrated value of the difference are equal to or larger than a predetermined threshold. In this way, the discriminator 1031 can determine the symmetry, and based on this, the object for pattern matching can be compressed, so that the processing time can be greatly reduced. This compression is performed by a comparator 1032, which compares the output of the discriminator 1031 with a group of standard distance distributions stored in the memory 1033, and outputs the result as A. FIG.

As a result, when symmetry is perceived, as a group having symmetry, for example, an unattended time (FIG. 22A), a normal posture time of the rider (FIG. 22B), and a forward tilted time of the rider (FIG. 22). The pattern of the distance distribution can be compressed with (c)), the forward child sheet (FIG. 23A), and the rear child sheet (FIG. 23B). On the other hand, when symmetry is not recognized, it can be compressed into various patterns in the posture of the occupant (FIG. 24).

Next, as a second method of compressing an object for pattern matching, there is a method of paying attention to the degree of correlation between each field of view with respect to the change in distance distribution in the field of view. The determination unit 104 of FIG. 7 performs the processing.

(A)-(c) of FIG. 10, (a)-(c) of FIG. 11 are explanatory drawing for demonstrating this concretely, FIG. 25 (a)-(c), FIG. 26 (a) Unlike (-), the result which extracted the part from which a distance changes is shown. As is also apparent from FIGS. 10A to 10C, the view of the unseat seat of FIG. 10A, for example, the views R1, R2, and R3 of FIG. Field of view (R1, R2) of the normal posture of (b), field of view (R2, R3) of the forward child seat of (a) of FIG. 11, and field of view (R1, R2, R3) of the standing child of (c) of FIG. ), It can be seen that there is a high correlation with each other. It is the calculator 1041 that detects the correlation.

Thus, the difference R1-R2 between the distance distribution of the field of view R1 and the distance distribution of the field of view R2 in FIGS. 10A to 10C, and FIGS. The difference R2-R3 between the distance distribution of R2 and the distance distribution of the field of view R3 and the difference R3-R4 between the distance distribution of the field of view R3 and the distance distribution of the field of view R4 are respectively shown in FIG. (a) to (c) and Fig. 13 (a) to (c), each difference is compared with the predetermined thresholds TH1 to TH6, or the integrated value S as the absolute value of each difference is respectively determined as the predetermined threshold ( The height of the correlation can be seen by comparing with THS1 to THS3). Therefore, using this result, if the comparator 1042 and the memory 1043 compress the target group for pattern matching, the processing time can be shortened.

14A to 14C are first explanatory diagrams for explaining a third embodiment of the present invention, and FIGS. 15A to 15C are second explanatory diagrams similarly.

This focuses on having a convex shape and a concave shape peculiar to each of the visual fields R1 to R4 according to various attitudes in the visual field distance distribution.

1) R1, R2, R3 of unmanned seat: upper convex, R4: forged (flat)

2) R1, R2 of normal posture: lower concave

3) R1, R2 in forward tilted position: lower concave

4) R1, R2, R3: convex lower, R4: forged (flat) of forward car seat

5) R1 of the rear car seat: upper convex, R2, R3, R4: lower convex

6) R1, R2, R3 of standing child: upper convex, R4: lower convex.

Therefore, the object can be compressed by checking the distance distribution state of the visual field with the discriminator 1051 and comparing it with the reference value from the memory 1053 in the comparator 1052.

14 (a) to 14 (c) and 15 (a) to 15 (c) are for describing this in detail, and only the visual field R1 is shown here.

That is, the average value LA of the distance value in the 1st in-view position PF1 of the visual field R1, and the distance value in the in-view position PF2 is calculated | required. This is compared with the distance LC at the intermediate position PFC of PF1 and PF2,

LC ＞ LA: Stomach convex

LC = LA: flat

LC <LA: lower convex

Judgment is as follows. Further, if the field of view is not limited to R1 but also for other fields of view, the object can be further compressed.

As another method (fourth embodiment) for compressing an object, a method of determining the outline attitude of the rider from at least one, preferably distance information of a specific in-view position of a plurality of views is considered.

This is explained with reference to Figs. 16A to 16F. This represents a distance value of a specific in-view position of the visual field R1 to R4, for example, its center position (center position in the left and right direction of the normal posture rider), and FIG. 16A shows an unmanned seat, and b shows a normal posture. (c) is an inclined forward position, (d) is an example of a tendency to increase toward the field of view R1 to R4 such as a forward child seat, and (e) is a field of view (R1 to R4) like a rear child seat. (F) may be classified as an example having a maximum value such as a standing child.

As shown in FIG. 35A, distance information in each viewing area is obtained by the occupant sensor 1, and the linear viewing areas R1, R2, R3, and R4 are generally set in the horizontal direction with respect to the occupant. Or formed. As shown in FIG. 35B, the distance information is received by the extractor 1061 of the discriminating unit 106, and from this information, the distance value at an intermediate position in each field of view is sequentially measured, and the field of view is determined. You will get a change in the distance between them. When the distance increase / decrease result in each of the viewing areas R1, R2, R3, and R4 is patterned in the order of R1, R2, R3, and R4, the result in the case of FIG. 35 (b) is "+, +, +, Patterned with "+", where "+" represents an increase and "-" represents a decrease. On the other hand, the reference memory 1063 stores the increase and decrease patterns of the 27 models in advance as shown in Figs. 36 to 38, and this pattern is related to the posture of the rider or the shape of the object. The intermediate distance value / reference value comparator 1063 compares the output patterns " +, +, +, + " of the extractor 1061 with the increase / decrease patterns of 27 models stored in the reference value memory 1063, and the output pattern " +, +, +, + "Is judged to be a" monostatic increase "represented by the model increase / decrease pattern 1, and this result is output to the comparator 108 as an output D.

It should be noted that the patterning performed by the extractor 1061 is based only on the increase and decrease of the distance between the viewing areas, and the absolute value of the distance measured at the intermediate position in each viewing area is not used for comparison or pattern matching. This means that the "child seat (rear) / seat position of FIG. 33 is different even if the increase / decrease pattern obtained in the case of" adult (lean forward) / sheet position: "0" in FIG. 31 differs in the absolute value of the distance of each viewing area. This is because it has a result " 9 " as in " 0 " and " standing child / sheet position: " 0 " Therefore, these three cases cannot be mutually discriminated only through " determination of increase / decrease in distance between lines &quot;

As described above, in the pattern matching model pattern range, the discrimination unit 103 performs "determination on the left and right symmetry of the distance distribution", and the determination unit 104 performs "determination on the correlation between lines", and the determination unit Since it can be narrowed by performing " unevenness determination " at 105 and by " discerning about distance increase / decrease between lines " in the discrimination unit 106, the attitude of the rider can be determined based on a relatively small amount of information. . The "unevenness determination" performed by the determining unit 105 is not limited to the illustrated embodiment, but may be executed in other ways. For example, the difference between the distance value at the intermediate position in the linear viewing area and all other measurement points in the linear viewing area is obtained, respectively, and the average of the difference is compared with the threshold. This pattern is convex (or protruding upwards) if the mean is greater than the threshold and concave (or protruding downwards) if the average is less than the threshold. If the mean is within a range of the threshold, the pattern can be determined to be flat.

Further, when a collision occurs, when the method of airbag deployment can be determined from the above determination result of the fourth embodiment, it may be deployed by that method. However, since the central processing unit 201 (ECU) may be used for other vehicle control or the like, detailed attitude information is obtained by using distance information of many positions other than a position within a specific field of view, and based on this information, the airbag If it is necessary to determine the development method, there is a possibility that the attitude determination processing may not be timely adjusted. However, if the attitude determination processing at the time of occurrence of collision is made to be prioritized over the processing that does not require other emergency, In this case, the processing can be performed in a short time, and it is possible to prevent the possibility of timely misalignment.

17 (a) and 17 (b) are explanatory diagrams for explaining the fifth embodiment of the present invention, and FIGS. 18 (a) to (b) and (a) 'to (b)' show the case; The first explanatory diagram for explaining the relationship between the seat, the occupant, and the surveillance field of view, (a) to (b) and (a) 'to (b)' of Fig. 19 are similarly the second explanatory diagram of Fig. 20 ( a) to (d) are the angles of (a) to (b) and (a) 'to (b)' of FIG. 18, and (a) to (b) and (a) 'to (b)' of FIG. The distance distribution example in the case is shown, respectively.

Fig. 18A is a perspective view of a seat when unmanned, (a) is a side view thereof, (b) is a normal posture perspective view of a car rider, (b) is a side view thereof, and Fig. 19A is a car (A) is a side view, (b) is a perspective view of the side sitting position, (b) 'shows the side view, respectively. In other words, the distance is measured by setting the field of view almost horizontally with respect to the seat and the rider, but this embodiment is characterized in that the field of view is set almost perpendicular to the seat and the rider.

20 (a) is an unmanned seat, (b) is a normal position of the car rider, (c) is a forward position of the car rider, and (d) is a side sitting position of the car rider. Each distance distribution data is shown, and (R1-R5) has shown each visual field. Therefore, in the distance distribution data of FIGS. 20A to 20D, the vertical axis represents the position in the visual field, the horizontal axis represents the distance from the sensor, respectively, and FIGS. 20B and 20C show the distance from the sensor. The dashed-dotted line of the back shows the reference pattern (distance value) at the time of an unmanned seat.

From (a)-(d) of FIG. 20, in each posture

(1) Field of view R1 and R5, R2 and R4 for unattended seating

(2) Field of view R1 and R5 in normal posture, R2 and R4

As with high posture,

(3) sitting sideways view R1 and R5, R2 and R4

As can be seen, the posture is low in correlation. In addition, the postures having a high correlation include the view R1 and R5, R2 and R4 of the child seat in addition to the above.

Therefore, in the present invention, the object to be discriminated is compressed by the degree of correlation between the left and right corresponding surveillance fields, and the posture discrimination time can be shortened.

17 (a) and (b) show specific examples of the correlation determination, and in FIG. 17 (a), the distribution of the difference obtained by subtracting the distance distribution of the field of view R5 from the distance distribution of the field of view R1 in the normal posture [ R1-R5] and the distribution [R2-R4] of the difference obtained by subtracting the distance distribution of the field of view R4 from the distance distribution of the field of view R2, and the same process was also performed to the side sitting position in Fig. 17B, respectively. It is shown.

In FIG. 17A, the correlation between the visual fields is determined by setting predetermined thresholds TH1 to TH4 for the above difference. Here, the case where the difference is high within all the thresholds is shown. Similarly, the integrated value of the car can also be considered, and in FIG. 17B, the predetermined thresholds THS1, S2, S2 (see the hatched portion in FIG. 17B) are respectively regarded. THS2) is used to determine the correlation between the visual fields. In this case, the integrated values S1 and S2 exceed the thresholds THS1 and THS2, and the correlation is low.

That is, in the posture with high correlation, the visual fields R1, R5 and the visual fields R2, R4 in the case of the unmanned seat as shown in FIG. 20A, and the visual fields R1, R5 and the visual field in the normal posture as shown in FIG. There are R2, R4, and the like. In a posture with low correlation, there are visual fields R1, R5 and visual fields R2, R4 in the case of sitting sideways as shown in FIG. 20 (d). Therefore, with respect to the measured distance distribution, when the visual fields R1, R5 and the visual fields R2, R4 have a high correlation with each other, in the standard pattern prepared in advance, a standard pattern having a high correlation such as an unmanned seat or a normal posture is selected. The posture determination is performed by comparison with this, and when the correlation is low, the posture determination is performed by comparison with a standard pattern having a low correlation, such as a sitting posture, to shorten the time for pattern matching.

However, since the deployment of the airbag needs to be performed after the collision has occurred and there is no large movement of the occupant, the attitude of the occupant needs to be performed in a short time from the collision to the generation of the airbag deployment signal. There is. Here, since it takes too much time to determine the pattern matching with the distance distribution of the reference posture with respect to the distance distribution of the plurality of surveillance fields of view, in the present invention, in general, one field of view is provided in the direction of the central axis of the airbag deployment, that is, the center of the seat. For example, by monitoring by R3, only the distance distribution in this field of view R3 is matched with the standard distance distribution pattern, and the rider attitude is roughly discriminated.

In the event of a collision, the result is used when the deployment control of the airbag is possible by the above-mentioned rough determination. However, when more detailed attitude information is needed, detailed posture determination processing focusing on multiple views does not require another emergency. By prioritizing the processing that is not performed, high-speed processing is enabled and safety is ensured.

According to the present invention, various processing is performed on the distance distribution data in the field of view to minimize the amount of data for extracting or employing the feature and compressing the object for pattern matching, thereby reducing the processing time for the attitude determination. The advantage of greatly shortening can be obtained.

Further, according to the present invention, since the object is compressed for pattern matching, focusing on the characteristics (similarity) between the fields of view of the intra-field distance distribution data, the processing time can be greatly reduced. In the case of a collision, the attitude determination is usually performed by the minimum data. Thus, when the airbag deployment control is possible, the result is used. Otherwise, the processing for the attitude determination does not require any emergency. By prioritizing, processing can be speeded up and safety can be improved.

1 is a block diagram showing the overall configuration of a device to which the present invention is applied.

2 is an explanatory diagram for explaining a relationship between an optical sensor array and a field of view used in FIG. 1;

3 is a configuration diagram showing a specific example of the distance measurement arithmetic processing unit.

4 is an explanatory diagram of a distance measurement principle.

5 is an explanatory diagram of correlation calculation in the distance measurement principle.

6 is an explanatory diagram of a distance measuring principle at a plurality of points.

7 is a configuration diagram showing a specific example of a rider discrimination processing apparatus.

8 (a) to 8 (c) are first explanatory diagrams for explaining the first embodiment of the present invention.

9 (a) to 9 (c) are second explanatory diagrams for explaining the first embodiment of the present invention.

10 (a) to 10 (c) are first explanatory diagrams for explaining a second embodiment of the present invention.

11 (a) to 11 (c) are second explanatory diagrams for describing the second embodiment of the present invention.

12 (a) to 12 (c) are third explanatory diagrams for explaining the second embodiment of the present invention.

13 (a) to 13 (c) are fourth explanatory diagrams for explaining the second embodiment of the present invention.

14 (a) to 14 (c) are first explanatory diagrams for explaining a third embodiment of the present invention.

15 (a) to 15 (c) are second explanatory diagrams for explaining the third embodiment of the present invention.

16A to 16F are explanatory views for explaining the fourth embodiment of the present invention.

17 (a) and 17 (b) are explanatory diagrams for describing the fifth embodiment of the present invention.

(A)-(b) and (a) '-(b)' of FIG. 18 are 1st explanatory drawing which shows the relationship between a seat, a rider posture, etc., and its vertical monitoring visual field.

(A)-(b) and (a) '-(b)' of FIG. 19 are 2nd explanatory drawing which shows the relationship between a seat, a rider posture, etc., and its vertical monitoring visual field.

(A)-(d) of FIG. 20 shows (a)-(b) and (a) '-(b)' of FIG. 18, (a)-(b) and (a) '-(b) of FIG. Explanatory drawing which shows the example of distance measurement in case of "'.

21 is a conceptual diagram showing a conventional example.

(A)-(c) and (a) '-(c)' of FIG. 22 are 1st explanatory drawing which shows the relationship between a seat, a rider posture, etc., and its horizontal monitoring visual field.

(A)-(c) and (a) '-(c)' of FIG. 23 are 2nd explanatory drawing which shows the relationship between a seat, a rider posture, etc., and its horizontal monitoring visual field.

FIG. 24 is a third explanatory diagram showing a relationship between a seat, a rider's posture, and the like, and a horizontal monitoring field of view; FIG.

25 (a) to 25 (c) are explanatory diagrams of a distance measurement example corresponding to FIGS. 22 (a) to (c) and (a) 'to (c)'.

(A)-(c) is explanatory drawing of the distance measurement example corresponding to (a)-(c) and (a) '-(c)' of FIG.

27 (a) and 27 (b) are explanatory diagrams of a distance measurement example corresponding to FIG. 24 and an example of distance measurement at the time of mounting a forward child seat when unattended;

Fig. 28 shows determination results according to four discrimination algorithms.

29 to 34 are diagrams showing in detail the difference of the determination result according to the change of the sheet position.

FIG. 35A is a diagram showing distance values in respective viewing areas obtained by the occupant sensor 1. FIG.

35B is a diagram showing a distance value at an intermediate position in each field of view obtained by the extractor 1061 of the discriminating unit 106;

36 to 38 show model patterns related to the rider's posture or the shape of the object.

<Explanation of symbols for main parts of drawing>

1, 1A: passenger sensor

2, 2A: rider

3: car

4: auxiliary light source

10: multi-stage light receiving IC

11, 12: light sensor array

21, 22: imaging lens

101: distance measurement operation processing device

102: rider determination processing device

103 ～ 106: discrimination unit

107, 1033, 1043, 1053, 1063: memory

108, 1032, 1042, 1052, 1062: comparators

201: central processing unit

202: shock detection sensor

203: Airbag Inflator

204: seat sensor

1011, 1012: register

1013: partial data extraction unit

1014: correlation detection distance calculation unit

1015: controller

1031, 1051: discriminator

1041: Operator

Claims (13)

As a posture determination device of a rider, A rider in each linear field of view having a pair of imaging lenses and at least two pairs of linear light sensor arrays each comprising a plurality of optical sensor elements, and monitored by said at least one pair of linear light sensor arrays. A sensor forming an image appearing at different positions of the sensor; A distance measurement arithmetic processing unit for detecting a distance distribution from an image of a rider formed by the sensor; And a rider discrimination processing apparatus which extracts an abstracted pattern from the detected distance distribution, compares a plurality of model patterns stored in advance with the abstract pattern, and determines a specific posture of the rider or the object. At least one linear viewing region is set to extend in a substantially horizontal direction with respect to a rider, and the occupant discrimination processing apparatus extracts an abstract pattern by checking whether the distance distribution in each of the linear viewing regions is symmetrical, and the distance The posture of the rider, characterized in that a specific posture of the rider or the object is determined after selecting a model pattern having a narrow range by comparing a plurality of preselected model patterns with the abstract pattern according to whether the distribution has symmetry. Discrimination device. 2. The distance distribution according to claim 1, wherein the left and right symmetry of the abstract pattern of the distance distribution is inverted from the center portion of the distance distribution pattern in each of the linear viewing areas, and thus the distance distribution before being inverted in each of the linear viewing areas. By checking the distance difference at the corresponding position between the pattern and the distance distribution pattern after inversion, and determining whether each said distance difference is not larger than a predetermined value or whether the integrated value of said distance difference is not larger than a predetermined value, The attitude determination device of the rider, characterized in that the. As a posture determination device of a rider, A rider in each linear field of view having a pair of imaging lenses and at least two pairs of linear light sensor arrays each comprising a plurality of optical sensor elements, and monitored by said at least one pair of linear light sensor arrays. A sensor forming an image appearing at different positions of the sensor; A distance measurement arithmetic processing unit for detecting a distance distribution from an image of a rider formed by the sensor; And a rider discrimination processing apparatus which extracts an abstracted pattern from the detected distance distribution, compares a plurality of model patterns stored in advance with the abstract pattern, and determines a specific posture of the rider or the object. The linear viewing region is set to extend in a substantially horizontal direction with respect to the occupant, and the occupant discrimination processing apparatus abstracts by checking the correlation between the respective visual fields with respect to changes in the distance distribution pattern between adjacent regions of the linear viewing region. A pattern determination apparatus for a rider, characterized in that a model pattern having a narrow range is selected by extracting a pattern, comparing a plurality of previously selected model patterns with an abstract pattern according to the correlation, and then determining a specific posture of the rider or the target. . The correlation between the adjacent visual field areas of the linear visual field areas is obtained by obtaining a difference of the variation of the distance distribution pattern in each visual field area at a corresponding position of the adjacent visual field areas of the linear visual field areas. The vehicle occupant's attitude discrimination apparatus is checked by determining whether the difference is not greater than a predetermined value or whether the integrated value of the difference is not greater than a predetermined value. As a posture determination device of a rider, A rider in each linear field of view having a pair of imaging lenses and at least two pairs of linear light sensor arrays each comprising a plurality of optical sensor elements, and monitored by said at least one pair of linear light sensor arrays. A sensor forming an image appearing at different positions of the sensor; A distance measurement arithmetic processing unit for detecting a distance distribution from an image of a rider formed by the sensor; And a rider discrimination processing apparatus which extracts an abstracted pattern from the detected distance distribution, compares a plurality of model patterns stored in advance with the abstract pattern, and determines a specific posture of the rider or the object. The linear viewing region is set to extend in a substantially horizontal direction with respect to the occupant, and the occupant discrimination processing apparatus is a concave or convex pattern of the distance distribution between the first viewing position and the second viewing position of each of the linear viewing regions. Extracts an abstract pattern, compares a plurality of preselected model patterns with the abstract pattern according to whether the distance distribution pattern is concave or convex, selects a narrow pattern model pattern, and then A rider's posture determination device, characterized in that it determines a particular posture. The concave or convex form of the distance distribution pattern is an average value of distance values at the first viewing position and the second viewing position in each of the linear viewing regions, and the first viewing position and the second viewing position. A rider's attitude discrimination apparatus, which is determined by comparing a distance value at an intermediate position with a visual field position. As a posture determination device of a rider, A rider in each linear field of view having a pair of imaging lenses and at least two pairs of linear light sensor arrays each comprising a plurality of optical sensor elements, and monitored by said at least one pair of linear light sensor arrays. A sensor forming an image appearing at different positions of the sensor; A distance measurement arithmetic processing unit for detecting a distance distribution from an image of a rider formed by the sensor; And a rider discrimination processing apparatus which extracts an abstracted pattern from the detected distance distribution, compares a plurality of model patterns stored in advance with the abstract pattern, and determines a specific posture of the rider or the object. The linear viewing region is set to extend in a substantially horizontal direction with respect to the occupant, and the occupant discrimination processing apparatus compares the distance value in a predetermined position of each linear viewing region with respect to the distance value in the adjacent viewing region. Extracting an abstract pattern by checking, selecting a model pattern having a narrow range by comparing the plurality of model patterns previously selected according to the tendency of increase and decrease of the distance with the abstract pattern, and determining a specific posture of the rider or the target; The posture discrimination apparatus of a rider. The apparatus of claim 7, wherein the predetermined position is a center position of a rider sitting in a normal posture when viewed from the lateral direction. As a posture determination device of a rider, A rider in each linear field of view having a pair of imaging lenses and at least two pairs of linear light sensor arrays each comprising a plurality of optical sensor elements, and monitored by said at least one pair of linear light sensor arrays. A sensor forming an image appearing at different positions of the sensor; A distance measurement arithmetic processing unit for detecting a distance distribution from an image of a rider formed by the sensor; And a rider discrimination processing apparatus which extracts an abstracted pattern from the detected distance distribution, compares a plurality of model patterns stored in advance with the abstract pattern, and determines a specific posture of the rider or the object. The linear viewing region is set to extend in a substantially horizontal direction with respect to the occupant, and the occupant discrimination processing apparatus determines the attitude and distance of the occupant based on a comprehensive judgment regarding the symmetry of the distance distribution pattern in each of the linear viewing regions. The correlation between the concave or convex shape of the distribution pattern and the variation of the distance distribution pattern and the adjacent visual field of the at least one linear visual field, and the tendency of increase and decrease of the distance measured at a predetermined position of the linear visual field are shown. The attitude | position determination apparatus of the occupant characterized by the above-mentioned. As a posture determination device of a rider, A rider in each linear field of view having a pair of imaging lenses and at least two pairs of linear light sensor arrays each comprising a plurality of optical sensor elements, and monitored by said at least one pair of linear light sensor arrays. A sensor forming an image appearing at different positions of the sensor; A distance measurement arithmetic processing unit for detecting a distance distribution from an image of a rider formed by the sensor; And a rider discrimination processing apparatus which extracts an abstracted pattern from the detected distance distribution, compares a plurality of model patterns stored in advance with the abstract pattern, and determines a specific posture of the rider or the object. The linear viewing region is set to extend in a direction substantially perpendicular to the occupant, and the occupant discrimination processing apparatus checks the correlation between the distance distribution patterns of the pair of viewing regions at positions substantially symmetric with respect to the deployment axis of the airbag. Thereby extracting the abstract pattern and comparing the abstract pattern with a plurality of pre-selected model patterns in accordance with the correlation to determine a specific posture of the rider or the object. 11. The method of claim 10, wherein the correlation between the distance distribution patterns of the pair of viewing areas is determined by comparing each distance difference measured at a corresponding position in the pair of viewing areas with a predetermined value or by subtracting the integrated value of the difference. A rider's attitude discrimination apparatus which is determined by comparing with stationary. 12. The passenger according to claim 11, wherein the occupant determination processing apparatus normally measures only a distribution of distances in a field of view related to the deployment axis of the airbag to determine a rough attitude of the occupant, and at the time of a collision, the occupant including the rough attitude determination. The attitude | position determination apparatus of the occupant which carries out priority of the attitude | position determination process of this process over the process which does not require other emergency. In the occupant attitude determination device for use in an airbag system having an airbag accommodated in a receiving portion provided near a seat of a vehicle and a sensor for detecting an impact of the vehicle and deploying the airbag, A distance sensor having a pair of imaging lenses and at least two pairs of sensor arrays, the distance sensor detecting a distance distribution in a plurality of portions of a rider or an object present in a space between the airbag receiving portion and the seat for receiving the airbag; ; By comparing the pattern of the distance distribution and a plurality of model patterns of the distance distribution, the attitude of the rider or the shape of the object is determined, and according to the determination result, the deployment of the airbag is controlled or whether the airbag should be deployed. And a processing apparatus for determining The processing apparatus extracts an abstract pattern by checking whether the distance distribution has left and right symmetry, and compares a plurality of preselected model patterns with the abstract pattern according to whether the distance distribution has symmetry, thereby narrowing the model in a narrow range. And a posture of the rider or a shape of the object after selecting a pattern.