Detection Of Moving Objects In The Complex Background

Moving target detection based on video is an important branch of video image processing, and serves as the core task in many practical applications. Under ideal condition, moving target detection algorithm is based on a static background, but in practice, the scene of moving target is often influenced by such factors as illumination changes and camera jittering, so that there is no completely static background. Studying soldier shooting countermeasure system, this thesis conducts the detection and tracking of detection target motion based on complex background.The soldier shooting countermeasure system researched in this thesis is that the shooting range is divided into two parts on the left and right sides, of which each forepart is placed with two industrial cameras for real-time record of the soldiers’ motion locus and shooting conditions, and at the same time, with the cameras video, the computer terminal can help with the processing and analysis of detected video images, accurate detection of their motion locus, real-time simulating and reporting of information about the soldiers getting shot, and the judgment of the results of simulating shooting.The main problem of moving target detection is the contradiction between detection algorithm accuracy and real-time detection system. As is mainly reflected in this system, the system requires real-time feedback of soldier motion patterns and shooting state, but the system can only get a big picture from the original video with complex background, and has slow processing speed. This thesis focuses on the detection part of the moving target. The complexity of the system background and the slowness of the processing speed is mainly reflected in the following three aspects:1, the moving target detected are the soldiers dressed in camouflage uniform, making the distinction between foreground and background of the little;2, as the time goes on, the illumination is changing and the camera may shake, which results in the variation of the luminance and relative motion between the background and the current detection frame;3, there is an obstacle in the scene and it is irregular whether the moving object appears, what’s its size and when it appears. To detect moving targets in such a complex background, such problems need to be solved as updating the background, eliminating the noise caused by camera shaking and adaptive segmentation of the image, and the real-time performance of the system must be taken into account.On the detection of moving objects in video streams under complex background, this thesis updates the background model by adopting the background mask method based on the background subtraction method. The updated background mask method is adopted, aiming at connected domain detection for the foreground area and adaptive background updating so as to remove the noise caused by light variation. By responding to relative motion between the background and the current detection frame caused by camera shaking, the concept of "observation window" is proposed to establish the detection area and non-detection area based on the complexity of background and combining the feature that the moving targets appear at the edge of the obstacle in the system. That is to say, the area that a moving target may appear is to be obtained in the background image, the contrast between the current frame and the background image obtained by background subtraction are classified and only the non-zero pixels correspondent with the detection area is searched so as to reduce the calculation of algorithms to ensure timeliness and accuracy of the system. Meanwhile, the "observation window" is optimized to obtain "the capture line" smaller than the range of detection area by adopting Canny edge detection and morphological processing method. In doing so, noise generated due to camera shake can be removed and the misjudgment rate of moving targets can be reduced. At the same time, the approach of establishing threshold pseudo-diagram is adopted to solve the problem that cavity appears in the moving objects when the image is segmented by the common used method of adaptive threshold segmentation, using adaptive threshold to split the entire video sequence images to segment the moving object, which lays the foundation for the follow-up identification and tracking of the moving object.