Design Of Algorithms Based On Machine Learning And Application In Intelligent Systems

Machine learning(ML)is an important branch of artificial intelligence(AI).It involves disciplines in many fields and is widely used in intelligent systems,including computer vision(CV),wireless communications,medical diagnosis,natural language processing(NLP)and intelligent robots.In this paper,the advanced machine learning algorithms are carried out in the two fields of computer vision and wireless communication.In the national grid,inspection of transmission lines is a necessary condition to prevent accidents and power outages.With the development of the power industry,the intelligent inspection remotely using surveillance cameras gradually replaces manual inspections with high cost,high risk,and low efficiency.In order to save costs and reduce risks,the paper designs a high-voltage transmission line detection method based on image processing in the field of computer vision,which mainly analyzes images taken by drones and fixed cameras.First,after the picture to be analyzed is read in,the image is pre-processed through operations such as grayscale processing,Gaussian blur denoising,image arithmetic operations and grayscale linear transformation.Second,using Canny operators to detect edge pixels of transmission lines,and the region of interest(ROI)is set to extract the transmission line region to suppress background interference.Then,classify the edge detection results and fit the transmission line.Finally,the transmission lines in the image are drawn with red lines,and the number of detected lines is also output.And by comparing the number of transmission lines manually set with the number of transmission lines output by the algorithm,it can be determined whether the state of the high-voltage transmission line is normal.Experiments prove that the detection of high-voltage transmission lines is achieved with certain accuracy.In addition,it can eliminate the interference of the background of the sky and clouds,etc.,and it is enough to distinguish the crossing high-voltage lines,and it can handle edge detection breakpoints,which has certain robustness and practicability.Aimed at the millimeter-wave(mm Wave)-based massive multiple-input-multiple-output(MIMO)systems in the next generation wireless communication systems(5G),many hybrid precoding architecture and schemes are proposed.Since the number of radio-frequency(RF)chains is greatly reduced in this system,the traditional hybrid precoding technology will cause severe performance loss.To solve this problem,this paper introduces the switch and inverter(SI)-based hybrid precoding architecture as an energy-efficient solution for these challenges.In addition,a detailed performance analysis on sum-rate as well as energy-efficiency is provided through simulation on the two-stage hybrid precoding,antenna selection(AS)-based hybrid precoding,and adaptive cross-entropy(ACE)-based hybrid precoding.It is aimed to prove that the performance of the ACE based scheme is much superior to that of the others with the limited ranges of values of all parameters.At last,the suitable parameters are determined and we prove that they can lead to the optimal performance.