Feature Recognition Technology Study Based On Landform Classification Of UAV Landing Point

The issue of autonomous landing based on image feature recognition technology is a hot topic in unmanned aerial vehicle (UAV) autonomous flight domain. The image feature recognition technology has the important significance to reduce the landing system cost and enhance UAV autonomy. The object of study is the natural landform images. The recognition environment is complex landform environment which contains the lawn and the road. The main work is to recognize UAV landing point without manual mark.The research content includes: image preprocessing, image decomposition, feature recognition, match cluster and experimental verification. The image preprocessing analysis image photographed when the UAV autonomous landing. It can realize eliminate clutter and unified standards through algorithms tailor, filter, gradation, gradation normalizing and so on. The image decomposition use multiscale geometric analysis method to obtain sub-images in different direction and different resolution. The feature recognition extracts the Hu moments, the Zernike moments, the texture features, and analysis distribution and characters of these features, and then calculates the contribution rate of different sub-images. The image feature recognition library is built by certain features which have the biggest contribution rate.The match cluster find the large range of landing safety zone though rough matching between the aerial image and the image feature recognition library, then use the k-means cluster algorithm to achieve classification recognition, finally determine the appropriate sites for UAV landing.The experimental verification completes an imagery software system through the theoretical analysis and the Matlab actual programming.The experimental result shows that this algorithm can recognize the lawn and the road, and the recognition rate arrives 86.7%.This design can realize UAV landform classification recognition. It reduces the dependence on outside information and the other equipments, and enforces the independence of the UAV.