This solution is an AI visual recognition obstacle avoidance module specially designed and developed for sweeping robots. It supports mainstream chip platforms, with a built-in object recognition convolutional neural network model, which can realize intelligent recognition of objects in home scenes; at the same time, active low-light depth detection will provide full-field visual obstacle avoidance for lidar sweeping robots.
In-depth algorithms, AI neural network algorithms, and SLAM algorithms are deeply integrated to greatly improve the robot's adaptability and perception of the environment.
The background covers various home scenes: living room, bedroom, kitchen, study room, bathroom, dormitory, etc.
Marked objects include: slippers, fabrics (socks, cloth, etc.), wires, drag strips, bases (fan bases, lamp bases, etc.), scales, table legs, chair legs, household pet poo.
Update the obstacle map to the SLAM grid map, so that the path planning module can choose a strategy according to the obstacle situation, making the cleaning more efficient and intelligent.
Through the training of the graphics library, the robot uses the AI algorithm to recognize the object and the semantic meaning of the image input by the sensor.
Calculate the depth distance of the front object through binocular vision algorithm, and provide the data needed by the robot for obstacle avoidance countermeasures and path planning.Generate obstacle maps and mark them according to the types of obstacles.
