EMGU-WPF LiDAR Research-moving Object Tracking 2

Original: EMGU-WPF LiDAR Research-moving Object Tracking 2

The initial implementation of the dry tracking, and using dots to mark the obstacles

80780205

During the test, when there are two obstacles overlapping with the radar scan ray, the information of the obstacle which is far away from the teaching is lost. That is, the data is re-detected when no data is obtained and does not overlap. However, due to the loss of data frames, the algorithm is judged to remove one, add a new one, resulting in tracking failure or large error. Such as:

When the obstacles overlap, the blocked information is lost, resulting in mis-processing ( obstacle marker color Change---lost tracking target, removed and new obstacles ), in order to solve the problem, based on the previous algorithm, optimization of the algorithm can circumvent the problem, described as the Obstacle Survival index discriminant method.

Algorithm validation effect

Independent design algorithm and verification, pit more, brain hole is large, self-encouragement.

Technology Extension Scenario : In my scene, my lidar position is fixed and the obstacles are moving. Suppose the lidar is also moving slowly. The obstacle position information detected by the newly acquired frame data is transformed based on the displacement data of the LiDAR, and the displacement data of the LIDAR is passed through the algorithm to the detected obstacle on the new data frame. It can also construct a pseudo-model with no moving of the LiDAR, and then track and detect the obstacles based on the constructed pseudo-model series data.

such as: The original position of the Laser radar O, the original position of the obstacle A. The next frame of data detects that O moves to OT, and a moves to at; the angle α on the way is one of the numbers in my tracking algorithm. The obvious change is too large, about 60 degrees to 120 degrees, and there are other calculation errors, in the algorithm should cause tracking failure. Middle Left: The displacement of O is passed on to the at, the formation of a new obstacle position av, which constructed a simple pseudo-model (middle row right, the back row), it is equivalent to converting to O, a moved to the AV. This is exactly what I can track with my prior tracking algorithm.

The idea has yet to be further verified.

Slam 2D Technology is still more in the study, the commercial landing is also in the inquiry, the dynamic information collection equipment tracking dynamic objects This aspect of the information seems to be not many, to be explored.

EMGU-WPF LiDAR Research-moving Object Tracking 2