The Analysis And Recognition For The Motion State Of The Excavator Under Complex Scenes

With the rapid development of the artificial intelligence, computer gradually replace human to complete various tasks in the fields of human activities. In a real file, the excavator is closely related to the human. It was known that excavator which is a kind of equipment in the engineering construction, greatly improve the speed of project construction. Therefore, the excavator will mostly appear on the locale of engineering construction. So it is a good choice to judge whether it occurs an illegal scene in locale by analyzing the locale’s surveillance video and identifying the motion state of the excavator in the video. In addition, because excavator has a strong destructive force, criminals may use it to carry out activities which threaten the safety of human life and property. So it’s important to identify the motion state of the excavator in a video. On balance, a basic task is how to analyze and identify excavator’s motion state based on video. For this task, this paper research the algorithm of the excavator’s detection and identification of the excavator’s motion state. Then analysis and identification of the excavator’s motion state system has been preliminarily designed and developed which lays the foundation of the algorithm. The main rese-arch work is as follows:(1) Research on the algorithm of excavator detection. Because the excavator’s appearance is large deformation which the flexibility of each component of excavator leads to, so this paper study an object detection method that’s called as “Multi Deformable Part Model”. According to the real experiment, we decide to use the model which is consist of three deformable part model as the excavator’s detection model. And each deformable part model contains six component filters. The practice proved that this model works well at the excavator detection.(2) Research on the algorithm of analyzing and identifying the motion state of excavator. In this paper, we analyze a sequence of the excavator video and identify the motion state of the excavator. Firstly, this paper study a shape regression method that uses local binary feature. Because of the large change of the horizontal-to-vertical proportion, we propose an excavator’s shape regression algorithm based on Multi LBF Model. We can use this algorithm to regress the excavator’s shape(i.e., the relative coordinates of the feature points). Secondly, this paper uses the feature points and the horizontal-to-vertical proportion of the excavator to design an excavator’s motion state feature MMF. Thirdly, we use support vector machine(SVM) and MMF to train the excavator motion identification classifier. Finally, we use this classifier to determine the motion state of the excavator in the current video(i.e., determining whether the excavator is in the working state). The experimental results show that this method can effectively identify the excavator’s motion state in the current video.(3) Design and implementation of the system. Under the premise of the verification for the algorithm of the excavator detection and the excavator’s motion state’s identification, this paper uses Microsoft Visual Studio 2008 and Open CV2.4.8 to construct the analysis and identification system of the excavator’s motion state. This system analyzes the input video and identifies the motion state of the excavator. Then the analysis information feedback to the users of the system.