The Study Of Real-time Image Processing For Weak And Small Target Based On DSP

Detection of weak and small targets under complex background is a hot point of image processing field.But,due to the characteristics of small area,weakness of shape of image and low SNR etc,detection of weak and dim targets is still a difficult point.And Matlab,that is a simulating method,has more distance to engineering application in the detection of weak and small targets.This paper depended on the National Natural Science Foundation of China “The automatic detection method and experimental study of hydrogen content in magnesium melt based on real-time image processing”(51374007).Based on the theory of detection of weak and small targets,and platformed on DSP,the sequential images detection of weak and small targets was studied.Then,some real-time detection algorithms of weak and small targets were optimized.Automatic detection of weak and small targets system of multithreaded tasks based on DSP/BIOS was also developed.This layered the foundation of the real-time detection of weak and small targets under complex background in the engineering application.After comparing the detection of weak and small targets based on Matlab,The method of detection of weak and small targets for sequence images with double-pipeline based on DSP was raised up.Not only could it use double-pipeline algorithm and median filtering,but also combined with high performance digital media processor DM642 and Code Composer Studio 3.3.The detection program of weak and small targets was programmed by C language,and to be compiled,debugged and downloaded to Flash.Experimental results showed that the double-pipeline algorithm based on DSP could detect the weak and small targets in sequential images.Meanwhile,in order to solve timeliness question of detection of weak and small targets,the optimization research of detection algorithm of weak and small targets was proposed from the two aspects of preprocessing algorithm and software optimization.Experimental results showed that the speed of the detecting algorithm was improved obviously.Besides,a method,named automatic detection system of multithreaded tasks of detection,was raised up so as to improve the high efficiency and stability of detection algorithm,which was based on DSP/BIOS that was a real-time kernel providing multithreaded task operation and was superior to the single DSP.This system used the DSP/BIOS kernel,the configuration tool and the dynamic API function.It estimated detecting target by observing the output signal of the digital I/O.Experimental results showed that there was a consistent one-to-one match between output signal and target of detection,and not only improved the efficiency of the detecting program,but also made program debugging become faster and easier.Detection of weak and small targets based on DSP overcomes the technical problem of Matlab cannot be applied on industrial scene.Meanwhile,it proved the ideas on detection of weak and small targets based on DSP under complex background.What’s more,that layered the foundation of real-time,accurate and efficient detection of hydrogen content in magnesium melt with DSP.