SIASUN Robotics – 3 common vision algorithms for autonomous mobile robots

Algorithm 1: Deep information extraction

The principle is to use two parallel cameras to triangulate each point in space. By matching the positions of the imaging points in the left and right cameras, the distance of the corresponding 3D points in space is calculated.

The robot wants to obtain the three-dimensional coordinates of the target through several images, and the more important work in binocular vision technology is to match the images, first determine the points where the objects match each other in the left and right images, and then obtain the parallax and depth information of each point.

Binocular stereo vision has the advantages of simple and inexpensive equipment, high accuracy and fast speed, and can calculate distance and depth information without touching objects, and it has important applications in UAV power line inspection and industrial construction robots.

　　Algorithm 2: Positioning and navigation

Robot navigation is a relatively complex system, involving the following technologies:

· Visual odometer VO;

· Mapping, using VO and depth maps;

· relocation, which identifies the current location from a known map;

· Closed-loop detection to eliminate the closed-loop error of VO;

· global navigation;

· visual avoidance;

· Scene tagging, which identifies objects in the room and tags.

To put it simply, the surrounding environment of the robot is optically processed, the image information is collected with the camera first, the collected information is compressed, and then it is fed back to a learning subsystem composed of neural networks and statistical methods, and then the learning subsystem links the collected image information with the actual position of the robot to complete the autonomous navigation and positioning function of the robot.

This method, known as SLAM (Simultaneous Localization And Mapping), is the best embodiment of the intelligence level of mobile robots, and whether it has the ability to synchronously map and locate is generally regarded as the key prerequisite for whether the robot can achieve autonomy.

At present, the commonly used SLAM technologies are mainly divided into two categories, one is VSLAM based on vision sensors, and the other is laser SLAM based on laser sensors.

Visual SLAM refers to the use of depth cameras such as cameras and Kinect for indoor navigation and exploration. So far, indoor visual SLAM is still in the research stage, far from the extent of practical application; Laser SLAM technology is relatively mature, and it is also the most stable and reliable high-performance SLAM method at present.

　　Algorithm 3: Obstacle avoidance

The problem solved by navigation is to guide the robot closer to the target. When the robot does not have a map, the method of approaching the target is called visual obstacle avoidance technology. The obstacle avoidance algorithm solves the problem of avoiding static obstacles and dynamic obstacles according to the data of the visual sensor, but still maintains the movement in the direction of the target and navigates autonomously in real time.

There are many obstacle avoidance algorithms, and the traditional navigation obstacle avoidance methods such as the view method, the grid method, the free space method and other algorithms can deal with the obstacle avoidance problem when the obstacle information is known, but when the obstacle information is unknown or the obstacle is movable, the traditional navigation method generally cannot solve the obstacle avoidance problem well or cannot avoid the obstacle at all.

In real life, in the vast majority of cases, the environment in which the robot is located is dynamic, changeable, and unknown, in order to solve the above problems, people have introduced some algorithms in the fields of computers and artificial intelligence.

At the same time, thanks to the improvement of the computing power of the processor and the development of sensor technology, it has become easy to perform some complex algorithms on the platform of mobile robots, resulting in a series of intelligent obstacle avoidance methods, such as genetic algorithms, neural network algorithms, fuzzy algorithms, etc.

SIASUN Robotics – 3 common vision algorithms for autonomous mobile robots

SIASUN Robotics – 3 common vision algorithms for autonomous mobile robots

Sections viewed

Contact

working hours