Onboard camera configuration and positioning

This problem, described in its entirety in [1-2], is that of the configuration and positioning of an onboard camera. This system is composed of sensors (one to n) whose role is to ensure the stabilization of the image as well as the automatic focus. These sensors are themselves connected to processors (CPUs) (one to n) via a digital video bus. Then, the processors will process the image and compress it to detect obstacles and trigger the vehicle’s automatic decision making by transferring the information through « Transceivers » (one to n). In addition, the transceivers are connected to the CPUs through a parallel digital port. Sensors, processors and transceivers have their cost and reliability characteristics available in catalogs. Thus, the camera is stereoscopic when it has two sensor-cpu-transceiver chains or simple in the case of a single chain. All sensors, CPUs and transceivers in the catalogs are not compatible with each other.

In order to obtain a camera that is both efficient and as inexpensive as possible, it is necessary to optimize cost and reliability, which are the two main performance indicators when designing the system. This is done by minimizing the total cost of the system and maximizing the reliability. In addition to optimizing the cost and reliability of the system, it is also necessary to optimize the performance of the system by maximizing it. To do this, the positioning of the camera in the vehicle must be addressed. Correctly positioned, the camera must be able to detect an obstacle that can be very close or far away.

We rely on the principle of stereoscopic optics to calculate the distance that separates us from an object from the comparison of images in the two views of the cameras. This is called the disparity which is a nonlinear function of the height of the obstacle, the horizontal position of the obstacle, the till angle of the camera, the height of the camera and the characteristics of the sensors (focal length, and size of a pixel).

The system must also be able to operate in two modes: detection of a close obstacle and detection of a distant obstacle. For that, we rule that the image of a distant obstacle on the sensors must have a minimal size that we fix at 50 pixels and the image of the close obstacle must be able to be contained by the sensors.

The camera is characterized by the following six to nine design variables:

• The number of channels nbChan (one or two) of discrete nature.
• The reference of the sensorRef chosen in each discrete channel.
• The reference of the CPU CPURef chosen in each channel of discrete nature
• The reference of the transceiver TransRef chosen in each channel of discrete nature.
• The positioning height h of continuous nature
• The positioning angle 𝜃 of continuous nature.

The requirements are as follows:

• Compatibility of the components in each channel.
• Minimum cost of the architecture.
• Maximum reliability of the architecture.
• Detection of near obstacles.
• Detection of distant obstacles.