Research on application of BPT node region merging algorithm in UAV rivers and lakes patrol inspection
In view of the problems of the high intra-domain heterogeneity and inter-domain homogeneity in UAV images of rivers and lakes patrol inspection,incomplete target region extraction results are easy to be obtained by using single-stage segmentation algorithm,and a region merging algorithm based on BPT nodes is designed.The algorithm consists of two stages:initial segmentation and region merging,which uses the watershed algorithm for initial segmentation to form the over segmentation region set with low mis-segmentation rate.The similarity measurement of the region pairs in the over segmentation region set is carried out,and the BPT tree is constructed according to the similarity measurement results of the over segmentation region set.Thus,by traversing the BPT tree,the merging order of the optimal over segmentation region set is determined.The comparison between the algorithm presented and adaptive region merging is carried out in rivers and lakes inspection data set.The result shows that the region merging algorithm based on BPT nodes has better segmentation accuracy and time efficiency than the adaptive region merging algorithm,which can achieve accurate extraction of the object in UAV images of rivers and lakes patrol inspection.
BPT node region mergingregion merging algorithmrivers and lakes patrol inspectionUAVintelligent image processing
万方数据
维普
