Occlusion local feature point extraction target Methods

The present invention provides a partially shaded target local feature point extraction method. Technical solutions are: First, based on a scale-invariant feature transform algorithm extracted in most cases are relatively stable changes in local feature point, a feature point as a crude. Then, the amplitude ratio and a high average gradient curvature crude feature points calculated from the three confidence aspects including contrast gradation, the threshold value of the crude filter characteristic point; crude final feature points robust screening for partial The feature point extraction. The present invention overcomes the problem of slow conventional large amount of data local image feature point extraction calculation speed, while ensuring the robustness of the local feature points.