DE102005025963A1 - Object detecting method for motor vehicle, involves detecting reflections of all individual rays on seat of motor vehicle as common image by common receiver and evaluating brightness of local reflections of rays within image - Google Patents

Abstract

The method involves adjusting individual rays on a seat of a motor vehicle, and directing locally limited illuminations on the seat area. Reflections of all the individual rays on the seat are detected as a common image by a common receiver. Brightness of the local reflections of the rays within the image is evaluated. A position of the brightness maxima is evaluated, and a distance between the brightness maxima is evaluated. Independent claims are also included for the following: (1) an object detecting system for interior of a motor vehicle for detecting an object on a seat of the motor vehicle (2) a motor vehicle with an object detecting system.

Description

The The invention relates to a method for detecting objects the seat of a vehicle according to the preamble of claim 1.

It is not least to fulfillment US law (FMVSS 208) has long been a so-called Occupant classification system (OCS) sensor for occupant occupancy detection and an OOP sensor (Out of Position) for detecting the position of the Head developed. The aim is to cause the airbag deployment To minimize injury to the occupant. There are many solving strategies for this Problem. In doing so, the scenery can be monitored by field sensors (capacitive, resistive), ultrasonic sensors or optical sensors (active, passive) be recorded.

The 2-dimensional detection, for example by means of corresponding gray-scale cameras requires a very uniform lighting without disturbing outside light, to recognize objects exactly. Therefore often there are also actively illuminated systems in the non-visible Wavelength range for use.

The Function principle of active optical sensor systems can be continue to roughly divide into 2-dimensional (2D) and 3-dimensional (3D) acquisition. 3D capture can after the Principle of triangulation, light transit time measurement or light phase measurement or stereo method.

The 3D capture taking into account the Removal of objects allows a higher quality rating than 2D images comparable Resolution, However, it is much more complicated.

at The distance measurement according to the triangulation method will be for everyone discrete beam over the shift at the receiver closed to the distance to the object. Will be on the receiver intensity curve measured, is the most intense Digit selected as the respective corresponding displacement. To This method is usually a one-dimensional distance vector determined.

at Determination of the 3D information according to the principle of the light transit time or light phase measurement, the sensor exposes a space segment the AB module diffuse with high-frequency modulated infrared light. This is reflected by the occupant and the vehicle structure and received by the sensor again. The backscattered phase front of the Contains RF-modulated light the complete Depth information within their local delay. The irradiated wavefront is correlated with a correspondingly modulated electrical Field within each pixel. As a result, the pixels deliver quasi-stationary RF interference pattern.

The The principle is therefore the principle of RF modulation interferometry (RFMI) similar. From the phase relationship between the transmitter drive and the received one Leaves light thus derive the distance of the remission surface.

At the Classic stereo methods are usually two cameras against each other staggered. Corresponding pixels in the Left and right stereo image are identified. Close objects in the image field moved very strongly against each other, however far away hardly. From these shifts, called disparities, the Distances of the objects are calculated. Output of the sensor is thus a pattern, which is further processed. A classification algorithm returns the current occupancy class or head position.

The Pattern extraction by the triangulation method is very robust and takes up few resources. It will be next to a small μC an LED line (e.g., 10 LEDs) and a one-line image sensor needed. A serious disadvantage, however, is the determined one-dimensional Range vector. The information content of this pattern is very much low. The downstream classification algorithm is often not able to make small changes (e.g., twisting a child seat) to interpret correctly. confusion with other occupancy classes are the result.

One multidimensional distance vector (remission matrix) more information. The classification algorithms thus have more Information available for the right decision. In "good" patterns, are almost there are no more mistakes. However, they are currently available 3D sensors in the early Development. The recorded patterns are very noisy, temperature- and remission-dependent. additionally there is a dependency from the intensity the ambient light. The system price (sender and receiver) is also very high.

In the stereo method, physical conditions, in particular illumination variations, but also occlusions, sensor noise as well as different camera characteristics, optical distortions, different apertures etc., ensure that the brightness distributions of the two stereo images generally differ significantly from each other. Search algorithms which assign the pixels in the left and right image to each other must therefore be extremely fault-tolerant. These algorithms are extremely computationally expensive and often have to be integrated into an FPGA.

Also here is the system price (simple transmitter but two camera chips each with a receiving optics) also very high.

The Object of the present invention is therefore to provide a method of to introduce the aforementioned type, which is extremely cost-effective realize and yet a good object recognition accuracy having.

These The object is solved by the features of the independent claims. advantageous Developments of the invention will become apparent from the dependent claims, wherein also combinations and developments of individual features with each other are conceivable.

One essential idea of the invention is that just not the complete scenery with a diffused light possible evenly lit. will, as for conventional Systems usual is, but specifically a locally limited, so quasi punctual Illumination is produced by a plurality of individual beams, however, the reflections of all individual rays from a seat from a common receiver be captured as a common image, so that the reflections accordingly together and in local like amplitude relation evaluated to each other and in the image as a recorded image pattern can be.

The Position and brightness of the individual reflections is clear easier to evaluate than, for example, conventional grayscale images. moreover enable Brightness and / or position of the local brightness maxima within similar to the picture like 3dimensional images inferences the distance of the recorded point and thus the seat occupancy state including the position of the occupant, his head or other parameter. moreover can be clearly seen on these patterns of local maximum brightness apply simpler pattern recognition algorithms.

The Invention will now be described below with reference to an embodiment with the aid closer to the figures explained. In the following you can for functional same and / or same elements with the same reference numerals be designated. Show it

1 : Object detection device comprising a lighting device that generates a plurality of individual beams with locally limited illumination of the seat area or objects thereon, and a common receiver 2

2 : Hardware technical variants of the design of the lighting device, for example, as LED lines or power LED with optical grating for splitting into the individual beams

3 Sketch-like representation of two typical seat occupancy structures and resulting image patterns of the local illumination spots whose overall pattern in the image or its brightness and / or position in the image and / or distribution are evaluated to each other

4 : Sketch of the block construction of an object recognition system

5 : Sketch of such a pattern comparison

1 shows an object recognition system for the interior of motor vehicles. It is a lighting device 1 for generating a plurality of individual beams 101 . 102 , ... such as 201 . 202 , ..., which are directed to a seat in the vehicle and lead to a localized lighting, as outlined by the individual light spots.

The illumination thus takes place quasi-selectively, ie it is preferably used a collimating optics, as in 4 sketchy hinted, so that a divergence of the single beam of less than 10 degrees, preferably about 1 degree is formed. With the distance range of 50 cm to 1 m, which is customary in these areas of application, spot diameters of approximately 1 cm to 2 cm are achieved, the luminescent spots being the coherent areas having a brightness value exceeding a specific threshold value. The individual beams preferably have an angular distance of 8 to 10 degrees to each other, ie the spots are at 1 m distance about 14cm to 17cm apart. This results in the image of the reflections a much simpler image pattern with sufficient significance, as will be explained in more detail below.

With a common receiver 2 the reflections of all the individual beams are detected by a seating area as a common image, ie the seat occupancy and, if necessary, position and position of the occupants on the seat are detected in an imaging pattern of these local light spots.

4 outlined in a block diagram the overall structure. The 2 shows hardware technical Variants of the embodiment of the lighting device 1 eg as LED lines 1.1 and 1.2 or as ever a power LED 1.3 each row with an optical grating 1.4 for the splitting into the single rays (compare to this also 3 ). the receiver is sketched as a logarithmic image acquisition chip with a 120dB dynamic range and a CIF resolution (350 × 280 pixels). This is of course only an example, in particular number of rows and spots and triggering the receiver variable. The image acquisition chip can basically also be used for other applications, for example, capture normal passive gray value images of the scene.

In addition outlined 2 a preferred embodiment, in the lighting device for a group of individual beams, here a row, a common light source 1.3 as well as an optical grid 1.4 has to spread in the individual jets.

thanks the localized lighting with single beams is compared to the classic gray value image a clearly easier to evaluate light spot pattern.

With increasing distance from the lighting device take the brightness in the local maximum and the size of the spots too. The spatial Extension of the objects is therefore evaluable.

In addition, objects in the seating area change the distribution of the light spots in the image and thus the entire image pattern. In particular, the position of the local brightness maxima (point 101 defined by its xy position x (101) and / y (101)) within the image as well as the distances between the local brightness maxima (x (101) -x (102), y (101) -y (102))). ..) among themselves.

• a) Reboard-, also gegen die Fahrtrichtung gerichteter Kindersitz
• b) Unfallbedingt vorverlagerter Insasse

The 3 now shows on the basis of two typical seat occupancy states

• a) Reboard, so directed against the direction of travel child seat
• b) accidentally displaced occupant

The resulting image patterns, on the one hand clearly in the image pattern indeed localized reflections and their significance for them Show situations.

It is now possible single of these occupancy state including concrete location and fixture dependent on the objects Sizes selectively rate.

As 5 illustrated again, both brightness differences between the setpoints (shown as white circles) with the actual values (dark circles) in terms of strength, shown here from the diameter size and the position, preferably the local maximum brightness, in the image, ie with respect to the xy coordinate or a target line (shown here in dashed lines) or the distances of the light spots are evaluated relative to each other.

Especially but preferably an overall evaluation of the recorded image pattern by means of Pattern recognition method, such as a polynomial classifier, by a comparison with stored patterns for typical seat occupancy and Occupant situations occur. In doing so, the images are inherent as well Brightness (z) and / or position (x, y) of the local reflections evaluated with the patterns.

Preferably is for a seat a plurality of individual beams aligned so that in the image of the reflections, an x / y matrix of several rows 10n 20n, each with multiple local reflections arises.

One Such object recognition system is suitable for seats in vehicles as well in other objects and is used to control applications the seat occupancy and / or occupant position are dependent, such as the triggering of Occupant protection devices, on the other hand, however, with regard to the level of detail with the here by the light spots clearly reduced pictorial features still get along, with the number of single rays Naturally depending on the situation is customizable and in particular, in addition to continuous intervals of Single rays and spots also specifically targeted zones of the seating area can be illuminated to maximize significant image patterns for one to get easy pattern recognition.

Claims (11)

Method for detecting objects on a seat of a vehicle, wherein the seating area is illuminated and the reflections occurring are evaluated, characterized in that a plurality of individual beams ( 101 . 102 . 103 , ...) are directed to a seat in the vehicle and lead to a localized illumination and the reflections of all individual beams from a seat of a common receiver ( 2 ) as a common image. A method according to claim 1, characterized in that patterns are stored for typical Sitzbelegungs- and occupant situations and the recorded image pattern in terms of brightness (z) and / or position (x, y) of the local reflections with compared to the patterns. Method according to one of the preceding claims, characterized characterized in that Brightness (z (101), z (102), ...) of the local reflections within the Image is rated. Method according to one of the preceding claims, characterized characterized in that Position of the local brightness maxima (x (101) / y (101), x (102) / Y (102), ...) is evaluated within the picture. Method according to one of the preceding claims, characterized characterized in that Distances between the local brightness maxima (x (101) -x (102), y (101) -y (102))) ...) be rated. Method according to one of the preceding claims, characterized characterized in that for a Seat a plurality of individual beams are aligned so that in the image of the reflections a matrix (x, y) of several rows (10n, 20n), each with a plurality of local reflections. Method according to one of the preceding claims, characterized characterized in that a Single beam a divergence less than 10 degrees, preferably below 2 degrees, and the individual beams to each other an angular distance from 5 to 20 degrees, preferably 8 to 10 degrees. Vehicle interior object recognition system for carrying out the method according to one of the preceding claims, comprising: - a lighting device ( 1 ) for generating a plurality of individual beams ( 101 . 102 , ...), which are aimed at a seat in the vehicle and lead to a localized lighting, - a common receiver ( 2 ), which captures the reflections of all individual rays from a seat of as a common image. An object recognition system according to claim 8, wherein the lighting device for a group of individual beams comprises a common light source ( 1.3 ) as well as an optical grating ( 1.4 ) for spreading the individual beams. An object recognition system according to claim 8 or 9, wherein the illumination device has at least one collimating optics for the individual beams. Motor vehicle with an object recognition system according to one of the claims 8 to 10.