Tracking Of Moving Object Based On Adaptive LTP And Meam Shift Algorithm

The dynamic visual information is one important part of the information, the people can receive with their sensory organs, and it also proves to be a significant research topic in the field of Computer Vision. In the HSI field, the tracking of moving object has a very important value, and is widely used in many field, for example, the image processing, pattern recognition, AI (artificial intelligence), video surveillance, medical diagnostics, faulty detection, automatic control, intelligent transportation, the vision-guided robot visual navigation, security monitoring in the military applications and so on. However, it has many problems to resolve in the practical application because of the change of light intensity of the moving object, the noise, the influence of other objects with same color and others negative factors. Therefore, it has a great challenge and is much valuable to research a stable, versatility and robustness way of the object tracking algorithm.The HSI article proposes the object tracking algorithm, which is based on the adaptive LTP textures and the Mean Shift. In the first place, the HSI article introduces background of the research, the existed problems and the usual imaging processing methods of the object tracking algorithm. And based on the video image acquisition and pre-processing, the optical flow method, background subtraction method and frame difference method, these three common target detection algorithms are studied and proposes the target detection algorithm, which is based on the color image fusion background subtraction method and frame difference method. The experimental result shows that, it can detect the moving object accurately, retain the information of color of the object and reduce the voids phenomena, and make a good foundation for the following object tracking algorithm. Because the video scenes always have large changes, such as the quickly changes of the brightness and the existing of much noises, the traditional Mean Shift tracking algorithm, which only relies the color feature, can’t ensure the accuracy of the tracking algorithm, so it needs to improve the Mean Shift method. Taking into account the texture of the object is a relatively stable characteristic, which is not affected by the light intensity and the background colors, so it makes a combination of the texture feature and the model of the object. THSI article creates the texture model with the Local Ternary Patterns (LTP), which is easy and convenient to represent the texture characteristics of the object. Because LTP computing relies on a fixed noise threshold, the HSI paper presents a method combing the adaptive threshold texture features and the Mean Shift, which improves the robustness of the object tracking. At last the simulation result shows that the average number of iterations and the mean square error of the HSI algorithm are less than the traditional Mean Shift and can also avoid the interference of the obstructions and the background effectively, thus the object can be tracked accurately.