[feat] Detect background and moving object

Goal

The goal of this MR is to improve the motion detector filter and adapt the filter to solid state lidar data.

The filter detects the motion points and extracts clusters among the motion points.

Method

The principal idea is to create a spherical map which each bin has a gaussian mixture model (GMM) of depth value. The GMM is updated for each frame if a point from new frame is located in this bin.

Result

Please find a demo with velodyne 32 data here

To do:

• refact code structure, make gaussian, gaussian mixture class internal
• analyse livox scan pattern
• add reset detection button
• sort cluster id by distance
• Compute cluster info
• Display bounding box
• Add frame info in cluster output
• Classification of clusters by geometry
• improve display
• add vtk test
• improve background detection
• Add algorithm description in MR

In the next MR

• Improve execution time by changing extraction of clusters

Notes: number of points detected by livox hap on function of the distance video

Runtime test On velodyne vlc-32C data before omp: EstimateMotion ~0.024s ExtractCluster ~0.06-0.2s