Video Surveillance Software Design Based On DSP For Obstacles On Railway Line

The accidents that happened on the railway line bring more and more loss of lives and wealth. However, the traditional monitoring methods for these cases are not effective because of their shortages in getting dangerous information exactly and real-time.Our subject presents a new obstacles-detecting method which equipments a DSP system beside the railway to detect obstacles on the railway and to send an alarm to the coming train for emergency operation. The system comprises CCD vidicon and DSP board adopting a lot of video processing algorithms.Two-values and edge-abstract are adopted as pre-process. This paper adopts Hough transform to abstract the straight rail which is hard to be extracted under complex background. Connectedness detecting is used to abstract the curve rail. For moving obstacles, we adopt the method of difference between adjacent frames, then the illumination influence is reduced. Whether the alarm will be given is determined by the moving trend. For still obstacles, a simple strategy named background subtraction is adopted. Miss-alarm is reduced by a new adaptive background updating with supervisal. Then, whether the alarm will be given is determined by the size of the obstacle.The algorithms are realized in C language and validated on PC. Then the code replanting into the DSP system TMS320DM642EVM is researched by using CCS and DSP/BIOS which can configure the system visually. Six tasks are partitioned in this program: system initialization, software initialization, image capture, moving obstacles detecting, still obstacles detecting, and sending an alarm. Tasks scheduling is realized by synchronization information exchange. Moving obstacles detecting and still obstacles detecting are implemented alternatively and constantly.What's more, several optimization methods are researched in some aspects to decrease the operation cost.