Three-dimensional target detection method and system based on auxiliary task learning network, and storage medium

The invention provides a three-dimensional target detection method and system based on an auxiliary task learning network and a storage medium, and the method comprises the steps: 1, carrying out set abstraction: designing a set abstraction layer which is composed of a sampling layer, a grouping layer and a point network layer; 2, inputting D-FPS and FS output results into a voxel feature coding and self-attention mechanism layer in parallel to form an auxiliary task learning network; step 3, in the candidate generation layer, using a representative point in the F-FPS as an initial center point, and transferring the initial center point to a corresponding instance under the supervision of the relative position of the initial center point and the correction of the auxiliary network center point estimation, so as to generate a new candidate point; and step 4, data enhancement. The method has the beneficial effects that the target detection precision is high, the real-time performance is good, the generalization