Application Research On Gestures Detection And Recognition Methods Of Underground Rescue Robot

Accident happens from time to time in coal mine production which is a high-riskindustry.And the underground rescue task is very hard and the search environment iscomplex,which result in great danger in the artificial rescue.Because trapped people areused to calling for help in the way of gestures, so it makes great pratical significance tostudy the rescue robot based on gestures tracking stead of the traditional artificial searchand rescue for guaranteeing the the rescue workers’life safety.With the the rescue robot platform based on machine vision, this paper studys thespecific target tracking methods which are applicable in the underground environmentby studying the complicated gesture movements to guide the motion of the rescue robotin order to complete the tracking of the moving gesture.In this paper,the target detection algorithms which are applicable in the undergr-ound environment is studied according to the problems underground such as the lightdistribution is uneven and the dust is too much. Hotelling transform is made for gray-scale images to prominent the gesture edge of the target featurea and increase the signalto noise ratio. Roberts edge detection algorithm is adopted tomake edge detectiong forthe gesture goal and K-means classification method is used to classify it to realize theidentity of static gestures.By the gray level distribution of the target, the Camshiftdifference method is improved,and the accuracy and stability of the target tracking areincreased.According to the size and posture changing in motion of the undergroundgestures goal, the independent transform gestures template is established and the thestretching ratio and rotation angle is changed adaptively to improve the matchingaccuracy of the algorithm. For the large area interference of the tracking, the judgmentfunction is increased and the the target area that is get by the and operation of thebackground difference and frame difference is introduced the as re-initialize the window.A geometric positioning method is also proposed based on space proportional spacing tomake the target positioning in this paper.The simulation results show that: the algorithmprovides good matching accuracy、real-time performance and stability and possess goodadaptability to phenomenons such as the change of the size and state of the target andand the dust cover.Above-mentioned algorithms are realized in the rescue robot platform based onDSP chip in this paper.The debugging result show that, the system is able to realize the gesture tracking in the underground which achieves the desired objectives. At last,thegesture tracking methods of the underground rescue robot are summarized through thestudy results,and the application prospects outlook is taked in this paper.