Human-tracking methods, systems, and storage media

A human-tracking system, implemented in a server, the server being implemented in a smart retail scene comprising a plurality of cameras, and the apparatus comprising:

• the plurality of cameras, each of the plurality of cameras being a binocular camera;
• and
• the server, configured to receive a plurality of human point clouds of a current frame from the plurality of cameras, generate a total point cloud of the current frame by integrating the plurality of human point clouds of the current frame, receive a plurality of human point clouds of a next frame from the plurality of cameras, acquire a total point cloud of the next frame by integrating the plurality of human point clouds of the next frame, and perform human tracking based on the total point cloud of the current frame and the total point cloud of the next frame,
• wherein the server is configured to:
• determine external parameters of the plurality of cameras;
• and
• perform coordinate transformation on the plurality of human point clouds of the current frame or of the next frame to a plurality of new human point clouds of the current frame or of the next frame;
• and
• integrate the plurality of new human point clouds of the current frame or of the next frame into a same coordinate system based on the external parameters of the plurality of cameras, to acquire the total point cloud of the current frame or of the next frame,
• wherein the server is configured to:
• cluster points in the total point cloud of the current frame to acquire 3-dimensional location information of the current frame;
• cluster points in the total point cloud of the next frame to acquire 3-dimensional location information of the next frame;
• and
• perform human tracking based on the 3-dimensional location information of the current frame and the 3-dimensional location information of the next frame,
• wherein the server is configured to:
• sample and de-noise on the total point cloud,
• wherein the server is configured to sample and de-noise on the total point cloud by:
• converting, with a preset length as a basic unit, coordinates of the points in the total point cloud of into discrete values, wherein a value in each unit is a number of points in a vertical direction;
• and
• de-noising the points in the total point cloud based on the discrete values.

A non-transient computer-readable storage medium having computer programs stored thereon, wherein when the computer programs are executed by a processor, a method is implemented in a smart retail scene comprising a server and a plurality of cameras, and the method, and the method comprises:

• acquiring a plurality of human point clouds of a current frame from the plurality of cameras, each of the plurality of cameras being a binocular camera;
• generating a total point cloud of the current frame by integrating the plurality of human point clouds of the current frame;
• acquiring a plurality of human point clouds of a next frame from the plurality of cameras;
• acquiring a total point cloud of the next frame by integrating the plurality of human point clouds of the next frame;
• and
• performing human tracking based on the total point cloud of the current frame and the total point cloud of the next frame,
• wherein generating the total point cloud of the current frame or of the next frame, comprises:
• determining external parameters of the plurality of cameras;
• and
• performing coordinate transformation on the plurality of human point clouds of the current frame or of the next frame to a plurality of new human point clouds of the current frame or of the next frame;
• and
• integrating the plurality of new human point clouds of the current frame or of the next frame into a same coordinate system based on the external parameters of the plurality of cameras, to acquire the total point cloud of the current frame or of the next frame,
• wherein performing human tracking based on the total point cloud of the current frame and the total point cloud of the next frame, comprises:
• clustering points in the total point cloud of the current frame to acquire 3-dimensional location information of the current frame;
• clustering points in the total point cloud of the next frame to acquire 3-dimensional location information of the next frame;
• and
• performing human tracking based on the 3-dimensional location information of the current frame and the 3-dimensional location information of the next frame,
• wherein the method further comprises:
• sampling and de-noising on the total point cloud,
• wherein sampling and de-noising the total point cloud comprises:
• converting, with a preset length as a basic unit, coordinates of the points in the total point cloud into discrete values, wherein a value in each unit is a number of points in a vertical direction;
• and
• de-noising the points in the total point cloud based on the discrete values.