Design Of Monocular Vision Based Pedestrian Warning System For Mine Locomotive

Mine locomotive driverless system is a trend and hot spot of the underground mining technology currently.However,this system hasn’t used generally due to the immature correlation technology.Realization of mine locomotive driverless system requires the system automatically detect the pedestrian on track lines,so that the locomotive issues a warning and automatically takes measures to decelerate to prevent the occurrence of accidents.Therefore,how to accurately detect the track pedestrian and give early warning is an important research content.Aiming at the above problems,this paper presents a pedestrian warning system for underground locomotive based on DSP+FPGA.The main work is as follows:(1)A Hough transform track detection method based on polar angle and pole constraint is designed.The number of voting units is greatly reduced in the parameter space by limiting the polar angle and polar diameter of the left and right rail lines of the locomotives,so that the influence of a large number of non-target disturbances on the orbit detection can be effectively removed.Improving the accuracy of track detection,while reducing the amount of computation,improve the real-time detection track based on DSP.(2)A low dimensional feature extraction method is designed for the problem of redundancy calculation in the histogram of the traditional HOG feature.The method divides the image into a number of non-overlapping block units and no longer divides it into smaller cell units,extracts the features for each block unit,concatenates the characteristics of all the block elements together to obtain the characteristics of the image.This can effectively avoid the traditional HOG feature due to the block unit slip to bring cell cell overlap,and thus cause redundant computing problems.At the same time in the statistical gradient direction histogram using the lookup table method,it can effectively reduce the DSP by calling the arctangent function of the time consumed.(3)In this paper,the pedestrian early warning system is completed on the hardware platform of DSP + FPGA.First,the analog signal is sampled by the FPGA to control the video decoding chip,and the data is read and written in the monolithic external memory SRAM,so that the system data sampling and transmission can be continuous processing,and then FPGA preprocessing the image,and the pre-processed image information through the EMIF interface to the DSP,DSP read data,you can carry out subsequent pedestrian detection algorithm.If a pedestrian is detected,the FPGA sends an instruction to the locomotive controller via the serial port to start the alarm module.