Research On Detection Technology Of Communication Base Station Antenna Based On Computer Vision

In the recently fast developing field of artificial intelligence,computer vision technology acquires widespread applications in real life.Image classification,object detection,image retrieval,image segmentation and other technologies are gradually maturing and emerging in the fields of transportation,medical treatment and industry.With the progress of science and technology,as well as the promotion of digitalization and networking in various industries,the demand of society for intelligence is gradually increasing.At present,the rapid growth of the number of antennas in the communication base station and the low efficiency of the artificial survey and deployment of antennas lead to urgent need of realizing automatic detection of the number of antennas on the holding pole in the field of mobile communication.Therefore,this thesis mainly studies an automatic detection technology based on computer vision for the number of antennas on the holding pole of communication base station,that is,people can judge whether there is space to increase the antenna by detecting how many antennas are on each holding pole in the survey image,so as to complete the task of intelligent survey of regional signals and automatic deployment of antenna.In this thesis,an algorithm based on computer vision is proposed to detect the number of antennas on the holding pole.The algorithm mainly includes two parts:object detection and subordination determination of communication base station antenna and holding pole.The details are as follows:1.In the communication base station antenna and holding pole object detection task,the center point of the object bounding box is used as the detection keypoint,and all other attributes of the object are regressed to determine the object bounding box,including its size and offset.In addition,this thesis proposes a method to realize intra-level features fusion,and the experiment shows that this method improves the object detection result by 0.8AP.2.In the task of determining the subordination between the communication base station antenna and the holding pole,(1)for the image of single antenna and holding pole object,this thesis directly determines the subordination based on the detected antenna and holding pole object;(2)for the image of multi antenna and holding pole objects,this thesis assigns an embedding vector for each detected object keypoint,and the subordination between the antenna and the holding pole object is determined by judging the distance between the embedding vectors of different objects;(3)for the image of multi antenna and holding pole objects with crowding and occlusion,this thesis designs a dual attention module to capture the rich context dependence.The module improves the global correlation of the embedding vectors from both space domain and channel domain.Experimental results based on antenna and holding pole image dataset verify the effectiveness of the proposed algorithm.