Research On Recognition And Location Of Odor Source Based On Point Cloud Data

With the development of industry, chemical industry is widely used in the fields of economy, social life, and so on, and becomes more more important.. However, in the process of production, storage and transportation of dangerous chemicals, leak accident, especially the gas leak happens easily and frequently. Therefore, it is necessary to study on odor source detection as the fundamental research, to ensure the safety of life and production.By now, most of the researches are given priority to olfactory information during the odor source recognition and location. Actually, gas concentration olfactory information is complicated, these methods based on olfactory information are often difficult to work. In recent years, some scholars bring visual information into the process of the recognition and location of odor source, but most of these methods are based on the visible light image, and these methods are susceptible to light, shadow, chromaticity, environmental change and other factors. Moreover, the light conditions of chemicals leak environment are usually not ideal, these methods based on olfactory information only could be limited.With the above in mind, the point cloud data is introduced into the recognition and location of odor source in this thesis, which could avoid the decline of the odor source recognition and location accuracy caused by illumination changes, and so on. Studying on point cloud segmentation algorithms, point cloud model datebase construction, and point cloud feature description, the odor source recognition system based on point cloud data is implemented. And local coordinates of the odor source are concluded. Secondly, the point cloud registration based on normal distribution transform algorithm is studied, and the improved normal distribution transform algorithm which is based on euclidean distance segmentation is proposed. Based on the improved algorithm, a robot visual odometry based on the point cloud data is designed and realized. Finally, the key frame extraction algorithm is put forward, which is based on the robot motion position parameters, hardware limitations and the non- plane information abundance parameters. With the combination of the odor source recognition system based on point cloud data and robot visual odometry based on the point cloud data,the odor source recognition and location based on the point cloud data is designed. Test results show that this system could recognize the odor source effectively and locate the source accurately.