You are on page 1of 9

Make/Model

Vehicle Style/Type

Features

FUNDAMENTALS of STEREOSCOPY
The contents of the first part of this page until the "stereoscopic ratio" was presented at KAPiCA 2002, the international conference on kite aerial photography, on November 5th, 2002 and has been updated and completed since. FOREWORD The key of success in photographic stereoscopy is to have a clear understanding and knowledge of the stereoscopic process and of the parameters which control it. We also have observed that stereoscopy is more or less effective depending on situations, and we will see how to deal with it for successful pictures. This page describes the fundamentals. CONDITION FOR STEREOSCOPY Stereoscopy is a physiological and geometric process. To get a stereoscopic sensation, the image in our left eye and the image in our right eye must be similar, but with slight geometric differences. In this case, our brain will interpret the scene for stereoscopic effect. If the two images have no differences, the stereoscopic effect does't exist. From this, it issues that stereoscopy is effective with the condition that:

Dots P1 and P2 must be differentiated by our eyes

PHYSIOLOGICAL LIMITS

Keenness of eye The keenness is the ability of our eye to make out two close points. We cannot distinguish two points which are within an angle less than one minute. Mathematically, keenness is an angle expressed in radian unit: a = 0.0003 to 0.0004

Color separation We can only distinguish colors if their area is large enough, and if the surroundings have a different tint. This may affect considerably the stereoscopic effect because the left and right image may not be different.

PARAMETERS OF THE SCENERY There are one non-geometric and two geometric Distance between camera and subject parameters to consider This parameter will apply as angle between lines L1 and L2 must be greater than the keenness limit. Background disposition

If the background is of the same uniform color, there is As Dincreases, the angle decreases. no stereoscopic effect. It is the case of a seagull seen flying in an uniform deep blue sky. Our eye cannot see any difference between P1 and P2. If the background is variegate, with many different colored and contrasted areas, P1 and P2 are really different, and there will be stereoscopic effect. The background disposition is a Distance between subject and background non-geometric parameter. The further is the background behind the subject, the On a photograph, there is not only one foreground and more P1 and P2 will be distant. It means that just behind only one background, but a succession of plans, one of a far subject, there is an area where stereoscopy is non them being the foreground of the next one, and the effective because the dots P1 andP2 will not be differentiated. background of the one before. The formulas can be calculated for each couple of successive plans depending on their real arrangement.

PARAMETERS OF THE OBSERVER The observer can use his eyes, but also binoculars, mirror devices, cameras, and so on. All these are optically different, and will affect stereoscopic effect. Applied to photography, the camera constituted of the lens and of the sensor, and the restitution device can change the stereoscopic rendering. The Base For us the base is the distance between our eyes, common value is 63mm. Using cameras, the base can be modified and it is the only parameter that we can change for stereoscopic effectiveness control. The larger is the The orthostereoscopy is when the base is equal to the base, the more stereoscopy will be effective distance between the eyes. RULES AND GEOMETRIC FORMULAS

Let's consider the distances B, D, L, and keenness a, From these, formulas are easy to get: when applying stereoscopic condition we have two basic Non effective zone N = a . D / (B - a . D) relations: Base B > a . D (D+L) / L e/(D+L) > a and B/D = e/L STEREOSCOPIC RATIO

The base formula will give the minimum base to use. As we said, the stereoscopic sensation can be strong or weak. So, it is interesting for us to appreciate and scale this stereoscopic sensation, at least from the geometric parameters. For this, we will use the stereoscopic ratio SR I must admit that I don't know any universal recognized rule or formula for this purpose. For years, I applied the formula that I published in Aerial Eye summer 1998, which I draw from a book. But I observed that it was not

The minimum space of two dots that the eye can see on the background is s, and the gap of projected points of the subject on the background is e: let's compare it! The Stereoscopic Ratio is SR If e=s Stereoscopic Ratio is 1 With e = B L/D and s = a (D+L) we obtain: SR = B L / a D (D+L) Tables can be calculated. For the human eye: 0< SR< 1 no stereoscopic effect. 1< SR< 5 weak stereoscopic effect 5< SR< 10 normal stereoscopic effect 10< SR< 20 high stereoscopic effect 20< SR< 50 hyper stereoscopic effect 50< SR excessive stereoscopic effect = e / s

working properly in some cases even it was satisfactory in most cases. After unsuccessful researches in a few books I went on my own, reconsidered the whole thing, expressed a new line of argument, and set a polyvalent formula. Note: The formulas are set for a stereoscopic angle of view identical to the human vision which doesn't exceed 60 LIMITS of STEREOSCOPIC EFFECT Maximum Stereoscopic Ratio: Maximum stereoscopic distance of the eye: Start from the formula: SR = B L / a D (D+L) and define t = L/D The maximum stereoscopic effect is got when the background is at infinity. Let's find the maximum we get SR = B/ (a.D) . t / (1+t) distance of a field to which the stereoscopic effect It can be verified that 0 < t/(1+t) <1 which gives the formula of the maximum stereoscopic becomes imperceptible to the human eye. In this case, SR = 1 ratio. Mathematical considerations lean to: The before formula is then written D = B / a SR max = B / (a . D) which occurs when the background is at infinity and The space of eyes is 63mm, Dmax = 0.063 / 0.0004 = 157 m which varies decreasing with the distance to the field. Behind this distance, everything is a neutral zone.

Neutral zone: Already mentionned, the neutral zone is where there is no stereoscopic effect. It can be defined two ways: Zone for which the field behind the subject is too close. Zone for which there is no more stereoscopic effect whatever the distance of the background. Physically, in both cases, for the eye, it is the zone where the stereoscopic ratio is inferior to 1. Thus SR<1 is also e<s. As example, the neutral zone of human eye is 1m for a subject 12m far, and further than 157m, all is neutral zone.

General formula of neutral zones: The extent of neutal zone behind the subject is calculated by : N = SR . a . D / (B - a . D) The distance behind which all is neutral zone is calculated by : Dn = B /(a . SR) It is interesting also to be able to calculate the neutral zones for different stereoscopic ratios. The use of cameras with various lenses, of binoculars, of fieldglass, makes the stereoscopic ratio in relation to their magnification factor.

CONVERGENCE of OPTICAL AXIS In all before, it is supposed that the two cameras have their optic axis strictly parallel. The stereoscopic pairs thus done are perfect for subjects moderately distant, or in the far. It is true for aerial photography, for landscapes. It is quite different for very near subjects, and when the base is important. Beyond 50m our eyes don't converge any more. On the contrary, more near is the place we are looking at, the more our eyes are converging.

In general, each eye can converge up to 12 and will normally adapt to a converging angle of 5. So, as the base is not greater than 1/5th of the distance to an object, B>D/5, we should theoretically be able to see the If the last background plan is not too far from the stereoscopic pairs. subject, the convergence will be benefic. On the other hand, if the last background plan is very far, the two However, it will depends on the means of visualization. images of the background will be shifted, and it will be With viewers, it will be possible. When projecting, it is necessary to mask the lateral parts not superposed to more difficult. avoid the trouble when looking at the views. KEYSTONING EFFECT

The horizontal and vertical lines are never displayed parallel, but they run to the infinity end of the axis lines, upward, or downward as well as toward the right and the left sides. It is well known that when turning the camera upward the vanishing of vertical lines is more pronounced. This keystoning effect bothers in stereophotography because the images cannot be exactly superimposed. When the two optical axis run to a near point, the keystoning is visible. In red the vanishing when the left camera is turned to the right, and in green when the right one is turned to the left. The bottom figure shows the difference when superimposed. CONCERGENCE RULE Thus, Let us recall that the convergence do not modify the when the background is very far from the foreground, stereoscopic ratio. the optical axis will be parallel. when the background is close to the subject, it is The good rule is that the backgrounds are perfectly possible to make converging the optical axis if assuring superimposed. that the background is perfectly superimposed on both views. Copyright becot.info

Browse posts by:


All Categories

Continental developing stereo cameras to cut down pedestrian injuries


by Aaron Richardson, Autoblog (RSS feed) on May 9th 2011 at 7:58AM

Continental has developed a stereo camera system to help cut down on the number of serious injuries and deaths related to people being hit by cars. According to Continental, in Germany, nearly half of all accidents that cause injury or death involve pedestrians. Continental's system utilizes two high-resolution cameras working in stereo to detect pedestrians the driver might miss. If the system detects someone stepping in front of you, it automatically performs an emergency stop. The stereo camera not only detects pedestrians, but can determine how far away they are, and how tall they are. Single cameras can see the person, but can't accurately estimate their distance from the car or height. If sudden braking isn't an option, or the pedestrian is too close for braking to be effective, the system can look within its field of vision to find a course of evasive action. The system is still in development, but the ability to

accurately and reliably detect pedestrians pushes it that much closer to production. Read the full press release from Continental after the jump.

The stereo camera reliably recognizes pedestrians and crossing traffic Frankfurt am Main. Continental, the international automotive supplier, will add a stereo camera to the comprehensive ContiGuard safety system as an integral element of its forward looking braking systems. This will help prevent or at least reduce the seriousness of the frequent accidents involving pedestrians or with vehicles at intersections; to date, accidents like these make up almost half (46.6 percent) of those traffic accidents in Germany that result in major personal injury. "What is expected of our accident prevention and avoidance systems is that, instead of prioritizing obstacles, they should be able to help in every hazardous situation. This means that we must look for new ways of monitoring a vehicle's surroundings", said Dr. Andreas Brand, Head of Passive Safety & ADAS Business Unit at Continental's Chassis & Safety Division. Since the stereo camera has two 'eyes', it is able to use the difference in the images within one camera shot to detect every type of obstacle, from loads that have fallen onto the road to people and animals, and can determine their size and the distance to them. This cannot be done sufficiently reliably with mono-cameras, which also have to be taught to recognize a car or a motorcycle and which are then only able to identify objects that they have learned. "Since the stereo camera also realizes the already familiar assistance systems, such as Lane Departure Warning, Traffic Sign Recognition, and Intelligent Headlamp Control, we think that it will set a new trend in the medium to long term and will be available for all vehicle categories, from compact cars to premium vehicles", added Brand. Two cameras and full image analysis within a single unit The stereo camera consists of two high-resolution CMOS mono-cameras, housed approximately 20 centimeters apart behind the windshield. Whereas a mono-camera only estimates distances, the stereo camera measures the distance to an object and its height from the road surface. This is made possible by the differences in the perspective between the lefthand and the right-hand optical paths. In other words, the stereo camera's analyzing electronics exploit the same effect that gives humans spatial vision, i.e. the parallax shift between two images. At medium distances of 20 to 30 meters, the stereo camera can determine the range to the object with an accuracy of between 20 and 30 centimeters. The stereo camera retains its high resolution capability even under difficult circumstances in which other technologies for object recognition might well reach their limits; for example, when several objects are in close proximity to each other, when objects are partially obscured, or when there is poor contrast between the object and its background. The fundamental strength of the stereo camera is its ability to compare the two optical paths because the redundant information obtained when both images contain identical zones with matching characteristics enhances the reliability of the data. In addition, the optical paths support each other in poor visibility, at dusk for example, so that they function better.

Safety through six-dimensional analysis In addition to the spatial position (3-D) of any object that it detects, the stereo camera provides particularly crucial supplementary data for the active driving safety systems. It can determine the direction in which every pixel of an identified object is moving along the horizontal, vertical, and longitudinal axes. This six-dimensional (6-D) identification makes absolutely clear whether an object is moving and in which direction. Combined with object classification, based on common characteristics, this process invests the stereo camera with such a high standard of decision-making certainty that it is able to initiate emergency braking (up to 1 g) if the driver fails to react to the object. The accuracy of the system enables the stereo camera to calculate the precise point of impact of a potential collision and to make the best possible use of the remaining time to prepare appropriate protective measures. The stereo camera functions through the whole speed range. Since the stereo camera can also identify potential ways, within its field of vision, in which the vehicle could take evasive action, other options are for a collision warning to be issued or for automatic braking to be applied earlier if no evasive maneuver is possible. The advantage of this is that a few hundred milliseconds are sufficient for an emergency stop to make the difference between sustaining bruises and suffering far more severe injuries. With its range of up to 60 meters, the stereo camera provides the best possible basis for developing braking systems that are truly looking-ahead. "In the future, the stereo camera will even be able to detect children, who are small pedestrians, cyclists, and wheelchair users crossing the road. We are, in fact, realizing a comprehensive obstacle recognition system, the like of which has never before been possible", said Wilfried Mehr, Head of Business Development for advanced driver assistance systems,,,

Two Eyes Are Better Than One The Stereo Camera

05/04/2011 The stereo camera reliably recognizes pedestrians and crossing traffic Frankfurt am Main. Continental, the international automotive supplier, will add a stereo camera to the comprehensive ContiGuard safety system as an integral element of its forward looking braking systems. This will help prevent or at least reduce the seriousness of the frequent accidents involving pedestrians or with vehicles at intersections; to date, accidents like these make up almost half (46.6 percent) of those traffic accidents in Germany that result in major personal injury. What is expected of our accident prevention and avoidance systems is that, instead of prioritizing obstacles, they should be able to help in every hazardous situation. This means that we must look for new ways of monitoring a vehicles surroundings, said Dr. Andreas Brand, Head of Passive Safety & ADAS Business Unit at Continentals Chassis & Safety Division. Since the stereo camera has two eyes, it is able to use the difference in the images within one camera shot to detect every type of obstacle, from loads that have fallen onto the road to people and animals, and can determine their size and the distance to them. This cannot be done sufficiently reliably with mono-cameras, which also have to be taught to recognize a car or a motorcycle and which are then only able to identify objects that they have learned.

The stereo camera reliably recognizes pedestrians that for example step onto the road from the side and half covered and is able to initiate an emergency brake. Click here to download this image

Since the stereo camera also realizes the already familiar assistance systems,
such as Lane Departure Warning, Traffic Sign Recognition, and Intelligent Headlamp Control, we think that it will set a new trend in the medium to long term and will be available for all vehicle categories, from compact cars to premium vehicles, added Brand. Two cameras and full image analysis within a single unit The stereo camera consists of two high-resolution CMOS mono-cameras, housed approximately 20 centimeters apart behind the windshield. Whereas a monocamera only estimates distances, the stereo camera measures the distance to an object and its height from the road surface. This is made possible by the differences in the perspective between the left-hand and the right-hand optical paths. In other words, the stereo cameras analyzing electronics exploit the same effect that gives humans spatial vision, i.e. the parallax shift between two images. At medium distances of 20 to 30 meters, the stereo camera can determine the range to the object with an accuracy of between 20 and 30 centimeters. The

You might also like