Research On Target Detection Method Based On Density Peak Clustering

As a technology of data processing,clustering algorithm develops rapidly and is widely used in image processing and computer vision.Target detection,as an interdisciplinary research subject in these two fields,also attracts much attention.With the emergence of clustering algorithms,researchers try to apply it to detection and achieve some results.However,these target detection algorithms often require a large number of prior conditions,and the actual target obtained is not complete.Aiming at this shortcoming,this article focuses on the target detection algorithm based on Density Peak Clustering(DPC).In order to apply the DPC algorithm to the image well,we make a series of improvements to it,and propose an adaptive target detection method.The experimental results and evaluation indicators perform well.The main contents of this article are as follows:(1)In the process of target extraction,strong and weak edges will be blurred,completeness is low,and hollowing.For this reason,an adaptive DPC target detection algorithm is proposed and named GS_DPC.The algorithm firstly combines the image gradient to correct the cutoff distance dc of the sensitive parameter,and then corrects the selection mechanism of the important parameter K according to the degree of data cohesion.Experiments show that the GS_DPC algorithm can effectively separate the strong and weak scale edges of the image,protect the edge contour of the detected object,and reduce the detection hole phenomenon.After the algorithm distinguishes all targets in a single video frame at the pixel level,it can perform secondary clustering on all the distinguished detection objects at the inter-frame area level,which is expected to extract the inter-frame based on such a two-step detection mechanism.Dynamic goals.(2)A target detection method combined with barnacle optimization algorithm(BMO)is proposed.The image information entropy is selected as the objective function of the BMO algorithm,and the optimal parameter ? after the solution is used as the sensitive parameter dc in the DPC algorithm,which overcomes the shortcoming of the DPC algorithm that is sensitive to parameters.SLIC superpixels are used for target detection in video images,a method of adaptively selecting the compaction coefficient m in superpixels is proposed,and the superpixel merging mechanism is improved.The experimental results and evaluation indicators show that the algorithm can effectively improve the accuracy of target object detection,the degree of aggregation between classes is higher,the redundant area is effectively reduced,and the algorithm has a certain degree of adaptability.