Research On Application Of Landslide Detection System Based On Monocular Ranging

The geomorphic characteristics of our country lead to frequent occurrence of landslides.The occurrence of landslides on both sides of railway and highway lines may cause huge economic losses and casualties.Therefore,it has important practical significance for the research of landslide detection technology.At present,landslide detection technology has been widely used in practical applications,such as traditional detection methods such as close-range photography,remote sensing and GPS,but these detection methods have disadvantages such as poor real-time performance and high investment.Therefore,this thesis presents a landslide detection scheme using monocular ranging.Moving target detection and tracking and monocular ranging algorithms are deeply analyzed,and a landslide detection system based on monocular ranging is designed and implemented.The main content of this article includes the following parts:First of all,algorithms such as illumination processing,moving target detection and tracking,and monocular ranging have been studied in depth,while traditional moving target detection and target tracking algorithms have been compared.The mountain background is complex,and there are many interference factors when performing motion detection.The existing tracking algorithm has poor tracking effect on the target under the complex background.Therefore,a number of calibration objects are fixed in advance in the mountain prone to landslide area and only the calibration objects are monitored.Continuous inter-frame adaptive algorithm is used to track fixed calibration objects,and color-based moving target recognition algorithm is used for recognition.In the monocular ranging algorithm,the traditional algorithm only considers the longitudinal distance measurement of the target,considering the actual movement of the object on the mountain,using an improved algorithm of the traditional monocular ranging algorithm,and introducing the horizontal distance to the target in the algorithm measuring.The above algorithm have been programmed and tested.Secondly,the overall design of the landslide detection system was completed.In view of the function and performance requirements of the system,the overall system scheme is designed,the software and hardware schemes of each part of the system are realized,and the data flow between the modules of the software system is analyzed.Thirdly,a landslide detection system based on monocular ranging is realized.The system is composed of a lower computer(data acquisition platform)and an upper computer(server).The main content includes: 1)The lower computer is built,the hardware is selected,and the data acquisition and transmission function of the lower computer is realized.On the JZ2440 embedded development board,the industrial camera and temperature and humidity sensor module are used to collect mountain video images and The temperature and humidity data is wirelessly transmitted to the server through the 4G network.The lower computer uses power modules and 4G routers for power supply and network connection;2)The server software is implemented,data reception,parameter calculation,and landslide real-time detection Wait for the function to be completed.The server receives the data sent by the lower computer in real time,and the mountain is monitored in real time after the internal and external parameters of the camera are calculated.The video image will be preprocessed by the system,moving target detection and tracking,real-time ranging and judging the target moving distance,and issuing an alarm.Finally,the landslide detection system has been tested.Taking 50 video image sequences with different light intensities and changing complex backgrounds as test samples,the system was tested and analyzed according to the sample library.The results showed that all functions of the landslide detection system were normal,and the missed detection rate of the system during the day is 4.08.%,the false detection rate is 2.08%;the missed detection rate at night is 8.33%,and the false detection rate is 2.17%.Compared with the traditional image monitoring method based on SVM algorithm and BP neural network algorithm to identify whether a mountain is landslide,the accuracy of the detection system in this thesis has been significantly improved,which has certain reference value for practical engineering applications.