Such as engineering target detection and target tracking.

Classical method: background difference method

A better method is: background subtraction without parameter estimation -VIBE.

Advantages of the algorithm: simple thinking and easy implementation; Sample attenuation is the best; High operational efficiency

Disadvantages of the algorithm: treating shadows as foreground; The moving target is incomplete.

Advantages: simple algorithm implementation, low programming complexity and fast running speed; The dynamic environment is adaptive and insensitive to the change of scene light.

Advantages: high real-time performance.

In the MDP process, the problem of single target tracking is regarded as a decision-making problem, so multi-target tracking becomes a multi-target decision-making problem.

Modeling objects with MDP mainly includes four parts:

Application example 1: video surveillance

Application Example 2: Missile Aircraft Identification

(1). Firstly, the infrared image of the aircraft is processed, including translation, rotation and scaling to simulate the motion of the aircraft, and continuous frame images are obtained as experimental samples.

(2) Using clustering algorithm to judge the complexity of the sky background in the infrared image, and then using segmentation algorithm to segment the aircraft; If the sky background is simple, use OTSU algorithm to segment the plane and go backwards.

(3) Extracting Fourier descriptors as the features of the research object.

(4) Finally, the target tracking algorithm is used.