Detection Of Moving Objects In Dynamic Background

With the development of Internet and digital information technology, video and image frame processing technology gains great attention in many sorts of fields as one of the most accepted and intuitive information access ways. Moving objects detection is the key procedure of video processing as it detects and isolates the moving objects or interested objects for following processing, such as object recognition or classification. Moving object detection is widely applied in digital intellectual monitoring, industrial designing, medical treatment, human interactive and military missile guiding. In real living world, there are moving targets by stochastic motion, such as swaying tree branches, flowing current, varied lights and so on. Therefore, the researches of moving object detection aimed at the complex dynamic backgrounds is always a concern of challenging subject and deserved significant attention. This paper proposes two algorithms based on background subtraction for dynamic scenes and organized as follows:1. Moving target detection algorithm based on fuzzy color coherence vector. The key part of background subtraction is to establish a stable background model, in which case to perform better on similarity comparing between the input frame and the background model.The similarity measure is taken by histogram distance. Color coherence vectors (CCV)include color statistical distribution and color local spatial information; hence it is able to overcome the representation limitation of conventional histogram. For the uncertainty and diversity of the images with dynamic motion,the method applies the fuzzy c-means clustering(FCM) to color coherence sub-vector and incoherence sub-vector and achieves the feature model FCCV which is abstracted from consecutive frames. The model develops a regional fuzzy statistical feature according each pixel of the image with fuzzy membership matrix from FCM. Simulation tests for the proposed detection algorithm with the videos and image frames prove efficient performance in various dynamic environments.2. Moving object detection algorithm based on the aggregation of color auto-correlogram and edge detection. The paper proposes a method of the aggregation of the edge detection and background modeling directing at the poor performance of simply background subtraction in dynamic environment. First of all, the paper extracts the auto-correlograms of the current frame and background frame which including the spatial information and statistical distribution information and conducts the similarity comparing. We will acquire one coarse detection result. Meanwhile, the paper conducts the edge detection for the two corresponding frames with detection Canny operator. After obtaining the edge results and a differential operation, we will get another coarse detection result of the algorithm. Then the paper integrates the two individual coarse results and employs the morphology operation. The algorithm shows great performance in complex dynamic scenes through a series of experiments.