Jointly with researchers from North China Electric Power University, scientists from South Ural State University have developed a digital twin of a test bench and a new method of multiprojection omnidirectional machine vision for robotics. The system combines images from several panoramic cameras with a structured laser light line, allowing robot to reconstruct the shape of obstacles and measure the distance to them with high accuracy. In addition, the system is dozens of times cheaper than existing analogues.
Thanks to integrated multiprojection omnidirectional vision and laser, robot can independently avoid spatial obstacles in industrial environment and accurately interact with surrounding objects.
“In the simulator developed on Unity 3D platform, it is possible to change the number of cameras on robotic manipulator, their positions, lens parameters, and laser settings, and immediately send image data to an external computing server,” explained Ivan Kholodilin, Associate Professor of the SUSU Department of Electric Drive, Mechatronics and Electromechanics. “Then a special algorithm merges data from all cameras into a single environment map. As a result, we obtain the shape of obstacle objects along the robot’s path and the distance to them. Tests in virtual environment and on real test bench showed that the system confidently “sees” the entire scene a full 360 degrees and measures the distance with error of only a few millimetres.”
Successful experiment opens way to safer and more accurate navigation of mobile robots, robotic manipulators, and machine vision systems in industrial settings.
To measure the distance from the robot to surrounding objects, the SUSU developers combine images from several panoramic cameras with projected laser line, obtaining the depth map and the obstacle contours. To ensure the system operates even in complete darkness, the scientists integrated a laser emitter directly into the robot design.
“When the laser beam intersects with an object, it leaves a coloured line on its surface, so even in dark conditions we can capture it using cameras and calculate the distance,” Ivan Kholodilin explained. “The algorithm merges images from multiple cameras, reconstructs the object contour from this line, and simultaneously calculates the distance to it.”
Among the key competitive advantages of the Chelyabinsk scientists’ development are the extremely high accuracy of the distance measurement to surrounding objects, including in darkness, and the cost reduced by dozens of times compared to analogue systems equipped with expensive 2D LIDAR rangefinders costing about 400,000 roubles. Results of the research are described in detail in the Sensors international journal.
Unique system of multiprojection omnidirectional vision with laser emitter can be applied at industrial enterprises where robotic manipulators are used for object sorting.