Study On Indoor Visual Localization Based On Adaptive Global Motion Estimation

With the development of society, deepening of information technology, many industries arealso increasing demand for indoor positioning. Visual positioning as a style of indoor positioning,which will be the impact of that indoor obstacles blocking the signal unlike other wireless indoorpositioning, and the visual positioning accuracy is very high. This positioning by virtue of theadvantages of high-precision positioning gradually is becoming the hot spots around the world inindoor positioning research field. The vision sensor of visual location capture the surroundingimage information by the visual sensor, such as a camera, and estimate the mobile cameraorientation information by matching the imagea of the sequence frame, so as to get the positioninginformation of the camera.In order to achieve effective indoor visual location, the main work has the following severalaspects:1. Through performance contrast experiments of SUSAN algorithm, SIFT algorithm andHarris algorithm, verified SUAN algorithm has good real-time performance ensuring accuracy.2. In order to improve the real-time and accuracy of SUSAN algorithm, SUSAN algorithmwere improved: adding a USAN circular template to crude extraction of feature points, using theoriginal USAN template to precise positioning, real-time performance of the improved SUSANalgorithm; Through the analysis of USAN template on pixel gray value, cut out the edge gradientand fuzzy edge feature points, improves the accuracy of SUSAN algorithm.3. In order to demonstrate the effectiveness of the improved SUSAN algorithm, We havedone the accuracy and real-time performance contrast experiments the SUSAN algorithmimproved before and after, the experimental results verified the improved SUSAN algorithmimproves the real-time performance and accuracy.of the original SUSAN algorithm.4. We proposed SUSAN-SIFT algorithm combining the advantages of SUSAN algorithmand SIFT algorithm, and then, did adaptability experiment of the algorithm. Verified the algorithmhas good rotation changes, illumination changes,affine changes and noise changes performance.5. Estimating the two matching image overlap with Kalman filter algorithm, extractionfeatures and matching only on overlap region. To reduce the computation redundancy of theglobal motion estimation method based on SUSAN-SIFT algorithm, improving the adaptability ofthe algorithm. Verified that improved the real-time performance of the algorithm ensuringaccuracy by experiments.6. Building indoor visual positioning system platform combined the software and hardware based on adaptive global motion estimation method, programming on VS2008, overallverification experiment was done on the system, verified the effectiveness of the algorithm.