Intelligent Antenna Object Detection Based On Deep Learning SSD Algorithm

With the rapid development of communication technology,the number of base station antenna has also increased.The attitude parameters of the base station antenna have a decisive influence on the stable operation of the antenna's communication system.The two main attitude parameters,pitch angle and azimuth angle,are the key factors to ensure the normal operation of the sector antenna,which are often changed by the external environment.Therefore,the accuracy of the antenna object's attitude parameter needs to be detected irregularly.The traditional contact measurement method mainly involves that the surveyors place a traditional measurement tool on the antenna by climbing the antenna,and obtaining the attitude parameters of the antenna.Obviously,there are shortcomings in traditional manual measurement methods.It requires huge labor costs and has great security risks.As a result,researchers are increasingly adopting non-contact measurement as a major solution.This measurement method is mainly to measure the antenna object by controlling the drone to complete the acquisition of the image,and obtain the attitude parameters of the antenna through image processing and 3D reconstruction.However,the current method has problems such as low measurement efficiency and a large number of manual interventions in the measurement process.Aiming at the problems in current non-contact measurement methods,this paper proposes a scheme for intelligent detection of antenna objects based on the deep learning SSD algorithm,which mainly includes two parts: automatic positioning of antenna objects and antenna attitude measurement.This solution uses the drone to capture and collect the antenna pictures that need to be measured,and transmits the antenna object pictures and its own attitude data to the server through the network transmission.The server uses the SSD algorithm to detect and locate the antenna object position in the picture.After that,it extracts the line in the area selected by the SSD algorithm,and combines the attitude data of itself to complete the calculation of the attitude parameters of the antenna.Compared with the existing methods,the measurement scheme proposed in this paper is more intelligent,efficient and real-time.In addition,this solution addresses the real-time requirements of the mobile phone.Based on the SSD network model,the basic feature extraction network is replaced with a more efficient Mobile Net and redundant network compression is performed.At the same time,according to the particularity of the antenna object,the prior information of the antenna object is introduced,and an improved Mobile-SSD algorithm model is proposed.The Mobile-SSD algorithm,while ensuring accuracy,can also meet the specific needs of mobile terminal for target detection in terms of detection efficiency.Based on the proposed measurement scheme,this paper implements a complete system that interacts with the drone,completes the object detection and attitude measurement of the sector base station antenna on the Android mobile platform.Meanwhile,this paper also uses the system to conduct sets of measurement experiments and functional tests by using antenna model.Based on the existing data set,the performance of the proposed improved MobileSSD algorithm was analyzed and compared.The experimental results and system test results verify the feasibility of the proposed antenna object intelligent detection method and the practicability of the designed system are presented.