Research On Infrared Characteristic Detection Method Of Personnel Invasion Under Low Visibility Dust Environment

Underground excavation operation requires no personnel near the dangerous area of excavation.However,due to the dark underground construction environment or lack of safety awareness,there is often operating personnel into the dangerous area,causing unnecessary damage.Existing methods,such as infrared sensor detectors,can not accurately detect the intrusion information,thus causing a high false alarm rate and affecting the safety of operations.In view of the problem,the paper studies a method of infrared feature detection of personnel invasion for low visibility in a dust environment.Based on this,we developed an infrared monitoring and early warning system for personnel intrusion.It can effectively detect a personnel or groups or local human targets for entering into the dangerous area.The interference of vehicle low frequency vibration to imaging detection is eliminated.Taking into account the sound and light alarm and dust removal requirements,it has achieved very good results in practical application.The main research works involved are as follows:(1)Through the imaging comparison experiment and evaluation of the actual working environment of coal mine in the early stage,we established a long wave infrared imaging detection early warning program.The program can maintain better imaging characteristics of the human body in low visibility,uneven lighting,heavy coal dust and humidity and other harsh conditions.Aiming at the problems such as low contrast of infrared image,noise interference and fuzzy edge of target,we studied a fast median filter method to denoise the infrared image,effectively suppressing the noise in the infrared image and making the image smoother,enhancing details of the information and image edges.In view of the system real-time requirements and image loss field problem,we used bilinear interpolation and interlaced legend and interpolation in the image preprocessing process.It effectively retained the image information and met the system application requirements.(2)This paper presents a method for the detection of infrared intruders based on multi-feature cascade.The primary classifier is composed of the aspect ratio aspect feature and the head Haar feature,and improved HOG-SVM completes final pedestrian identification.An improved HOG feature extraction algorithm and an adaptive scaling factor acquisition algorithm were studied and implemented.It can ensure the accuracy of detection on the basis of improving the real-time detection system.We carried out the blockage reasoning and local feature recognition while the target is blocked.The local feature description operator adopts the HOG and CS-LBP cascade features,which further improves the robustness of the detection system for special cases.Experiments show that the detection method can achieve high detection rate and meet the system real-time requirements.(3)We completed the design and construction of the monitoring and warning system for the infrared intrusion.Specifically included: we completed the system hardware design,component selection,site decoration,and we designed the intrusion control and early warning strategy,then we completed the object-oriented humancomputer interaction program design and achieved the system integrated control terminal and the power of the implementation of the communication interrupt control components.It can meet the real-time dust removal,detection and early warning requirements.We completed the coal mine underground field installation and functional validation.A variety of situations,including multiple people and local targets,have been tested for the actual situation and the system's early warning function has been verified.Experiments show that,the personnel monitoring and warning system designed in this paper has good adaptability and practicability in the complex environment of low visibility of coal mine,and can be applied to the actual production of coal mine.