2results about How to "Improve target positioning accuracy" patented technology

This application provides a vehicle control method for a two-wheeled vehicle and a two-wheeled vehicle. The method includes: during the operation of the two-wheeled vehicle, acquiring current attitude data collected by an attitude sensor and current perception data collected by a sensing device, wherein the acquisition time of the current perception data matches the acquisition time of the current attitude data, and the current perception data is perception data in the device coordinate system of the sensing device; correcting a preset mapping relationship between the device coordinate system and the world coordinate system based on the current attitude data to obtain a current mapping relationship; performing coordinate transformation on the current perception data using the current mapping relationship; performing target identification on the coordinate-transformed current perception data to obtain a set of perceived targets; and controlling the two-wheeled vehicle based on the set of perceived targets. This application solves the problem of low target positioning accuracy caused by unstable sensor coordinate systems in related technologies, thereby improving target positioning accuracy.

The application discloses a three-dimensional target tracking method, system, device and medium based on space-time enhancement, and relates to the technical field of automatic driving. The method comprises the following steps: inputting three-dimensional point cloud sequence data into a trained SMTrack network for processing to obtain a final bounding box prediction result. The SMTrack network comprises a target-specific encoder, an STFM module and an STT module connected in sequence. The target-specific encoder is used for extracting target-specific features from the point cloud sequence. The STFM module is used for modeling appearance information and motion information in stages according to the target-specific features to generate a preliminary bounding box prediction result. The STT module is used for optimizing the preliminary bounding box prediction result. The application adopts a twin structure, extracts features from historical frame and current frame sequences, and introduces part perception motion modeling and coarse-to-fine bounding box regression mechanisms, so that the target positioning accuracy can be greatly improved.