Research On Technologies Of Surveillance Video For Tank Truck Transportation Process

News and network information has shown that, in recent years, the oil of tanktrucks was stolen severely in the process of transport，which caused great economicloss to oil providers and the mass oil users. Right now, vehicle video surveillance fortank truck has been a significant way to keep the oil and the vehicle itself safe.However, with the increase of network scale, the video data grows in mass size and thehuman and time cost to search and see video increases twofold. So, this subjectconducts research of this problem，applies content-based video retrieval algorithm tovehicle video intelligent surveillance. Besides of not omitting significant frames, thisalgorithm can also rapidly and correctly retrieve the expected video information for theusers to meet the demand of different users.This paper’s main work includes:1. For the characteristic of jitter in the vehicle surveillance video, this paperpresents an improved block match-based quick video image stabilization algorithm. Itis based on block match. Draw out the image matching area at reference frame inadvance. Using the absolute difference value combined of the pixel grayscale to foundthe best match in the current frame. And quickly to predict the current frame relative tothe reference frame of the motion parameters, which is used to eliminate jitter.Experimental results show that the new algorithm can ensure eliminate jitter effect, atthe same time, greatly reduce the operation time, to meet user requirements.2. For the demand of video intelligent surveillance, the algorithm does checks ofthe moving objects of the vehicle surveillance video. Since the vehicle surveillancevideo is affected by sunshine variation and nature scene’s slight variation, when themoving object detection this makes difference image full of “false” moving objects.This paper uses multi-frame differential multiplication method to make the relativepeaks of moving edge in difference image for more sharp, and with the threshold, it can effectively eliminate moving object’s affection and correctly calculate movingobject’s edge contour.3. For the demand of video retrieval, it extracts the key frames of vehiclesurveillance video. To counter the movability of vehicle surveillance video camera andthe video content’s diversity, so at the same time of avoiding redundancy, to selectrepresentative key frames, this paper draws lesson from particle swarm-based keyframe extraction algorithm, considers global motion features and local motion featuresas the whole feature of the video, which has more focus on local motion.4. To validate the method mentioned above, this paper extracts low level featuresby means of image sequence, uses OO(Object-Oriented) design method andDirectShow technology in the development environment Visual C++6.0. Finallydesigns and implements a vehicle video intelligent surveillance system model with thefunction of video retrieval.