In the case of automatic parking, different people have different solutions. One focuses on machine vision and autopilot technology, while the other is optimistic about automation equipment.
I remember that when the first two years of the drop were just emerging, many taxi drivers in second- and third-tier cities felt that their interests were infringed and they joined forces to boycott the network. An O2O software became the public enemy of the taxi industry. . However, the buddies never imagined that the automatic driving that really subverted the taxi industry was already on the road, and artificial intelligence was desperate to become the terminator of taxi drivers.
Speaking of autonomous driving, what we are discussing now is no longer a question that can't be done, but a question of how long it will fall. However, autonomous driving is a big topic, and today we are talking about a technology that is expected to be implemented earlier: automatic parking.
Before talking about automatic parking, imagine a scene where a primitive person is stopped by a river, but he wants to go to the other side of the river. What should he do? To throw this question to everyone, the answer is nothing more than two: the first, learn to swim, swim through the river; the second, build a bridge across the two sides.
In fact, behind this answer, there are different behavioral logics: one is to change oneself to adapt to the world, and the other is to transform the world. In the case of automatic parking, it is also divided into two groups: the machine vision group that adapts to the world and the automation equipment school that transforms the world.
Machine vision solution
The most suitable machine vision solution is the camera. The ordinary camera is also called the non-depth camera. The principle is to capture the image data and then calculate the distance information from the image data. However, the non-depth camera is divided into a monocular camera and a binocular camera. The monocular camera performs target recognition (various models, pedestrians, objects, etc.) through image matching, and then estimates the target distance according to the image size.
Because there is a step of identification, the problem is also easy to appear in this part. Visual recognition has high requirements for database and image quality. Once it is not recognized, it will lead to errors in estimating distance. The reverse case is the Tesla crash. Under the condition of strong light, the camera on the Tesla car judged the white truck as a cloud, and did not slow down, which eventually led to tragedy.
The binocular camera scheme is different from the single-purpose mode, and the front image is obtained by the two cameras respectively, and then the parallax of the two images is calculated to obtain the distance information of the front object. This kind of scheme simulates the way the human eye acquires 3D information. Although the cost is relatively high, the accuracy is improved, and it is also a commonly used solution.
At the Consumer Electronics Show (CES) in 2017, BMW United Segway launched a parking assistant robot . The robot is equipped with a binocular camera that captures distance information in front of the car and guides the BMW's automatic driving system to slowly drive the car into the parking space.
However, the binocular camera is not perfect, because it relies on image data to obtain the distance, so in the scene of strong light or dark, the performance of the binocular camera is not stable enough, which promotes the development of the depth camera solution.
HuiZhou Superpower Technology Co.,Ltd. , https://www.spchargers.com