Background of the Pedestrian Detection System
Recognition systems for the army and civilian security systems began to be addressed by various developers in the 1990s. In consumer electronics, facial recognition systems first appeared in Japanese production cameras at the beginning of the 21st century. In 2010, the Pedestrian Detection System for the first time was offered by Volvo.
In addition to the PDS from Volvo, there are existing systems of Advanced Pedtrian Detection System of the company TRW and EyeSight firm Subaru.
Purpose of the Pedestrian Detection System
Not only does the system recognize the carriageways of the carriageways, but also warn the driver that there are pedestrians. If, according to the computer setup program, the computer decides that there is a risk of collision with a pedestrian, it will enable the automatic protection against the collision.
Studies conducted after the beginning of the use of car detection systems showed that in the case of a car equipped with a pedestrian detection system, the probability of death is reduced by 20%, and the infliction of grievous bodily harm by 30%
Difficulties faced by developers of pedestrian detection systems
Human behavior, its growth, the nature of gestures, the color of clothing, and many other parameters are too diverse, so a highly complex program that can analyze all of the above factors is necessary to create an effective system for the detection of pedestrians. The program requires significant computer memory resources and other powerful and reliable components. The equipment of the car is powerful, which means expensive computer equipment, first of all, leads to a significant increase in its price.
The second major problem faced by developers is the overlay of shapes, objects in the hands of pedestrians and the colors of clothing, on each other, and on objects or structures in the background. This makes it extremely difficult to recognize individual shapes.
Methods of detection of pedestrians
A holistic discovery
If the system "secrets" the movable object, it first enters it into a frame according to its dimensions. The computer then analyts the specific characteristics of the discovered object. This may be the level of infrared (thermal) radiation indicating that the object is "warm", that is, it is a living being, a person. The alternative method is to analyse the histograms of flowers inside the frame and compare them with the samples contained in the database. The disadvantage of this method is the continuous interference of foreign objects caught in the frame.
Partial recognition
When this algorithm is used, the shape of a person is considered to be a collection of fragments and fragments. The system analyts the contours, their location, and other parameters. The program forms a common picture and compares it with the sample. This method is more accurate, but requires more technical and computing resources.
Recognition method by model
This recently proposed algorithm combines the merits of the first and the second, but with certain improvements. The recognition system does not only provide information about the shape of the body or parts, the colors and the similar general characteristics, but the samples of the shapes photographed in the place where the system is operated. Thus, in the course of training, it is "trained" on certain local "types" for greater precision of recognition.
Multiple-compartment recognition
In this case, the individual cell is targeted for each mobile object (pedestrians). The general picture is divided into fragments, and each square is assessed separately for the person to be found in it.
How the Pedestrian Detection System works
When the sensors-radar and (or) the video camera, the mobile objects (in the Subaru system, for example, use two cameras) on the move, the system determines the direction of movement of the pedestrian, the speed of motion, and predicts its location at the time of the vehicle's appearance at the point of detection of the object (the system response time-40 meters to the object). After making sure the computer looks like a pedestrian, the computer prints a picture from the camera on the screen of the multimedia system.
If the computer is of the opinion that there is a possible collision at the available speed of the car and pedestrian, a sound signal is heard. If the driver reacts (turns the steering wheel or presses the brake), the system helps to stop the machine by means of a secondary braking system.
At the moment, the detection systems guarantee security at speeds of not more than 30-35 km/h. The recognition algorithm includes samples of generic objects that can be in the hands, for example, umbrelets or bags. However, in conditions of poor visibility (due to worsening weather conditions, for example) and night, the detection system is not yet useful.
Examples of vehicles equipped with a pedestrian detection system
Pedestrian detection can be seen if, for example, Volvo S60 is picked up.
The EyeSight system is equipped with Subaru-Legacy and Outback models. The second generation of the system was announced at the 2012 New York Motor Show.
