Research Of Road Recognition Based On Fuzzy Clustering Algorithm

Road roughness is one of important indexes of road grade. Research of roadrecognition is mainly on the study of road roughness. It is applied in road detection. Inaddition, it can also provide incentive information about the road for car when it doessimulation test of road indoors. Research of road recognition occupies important position inthe field of transportation and more and more is taken seriously. Now, in equipments forroad roughness detection, some are extremely expensive (e.g., laser flatness gauge), somehave low precision (e.g., RSP pavement check-out car). In addition, road roughness which ismeasured by above equipments cannot be used as incentive information about the road forcar when it does simulation test of road. Road vertical load which is collected by our teamcan well reflect the state of road roughness. And it also can be used as incentive informationabout the road for car when it does simulation test of road. Since road load spectrum basedon different road conditions has obvious characteristics of clustering, the fuzzy clusteringanalysis algorithm is a kind of ideal method of pavement classification recognition. It hasvery important practical significance and useful value. The main contents completed of thepaper are as follows:(1) The overall design of the project is introduced, including the acceleration sensor,data acquisition system, data transmission system and host computer, which is used toanalyze road load spectrum. Then there is performance comparison about two groups ofdata transmission system of serial to Ethernet. Technologies related to the road recognitionhave system structure of road recognition, extraction of road characteristic parameters andso on. They are introduced in detail.(2) Knowledge of fuzzy set and fuzzy clustering theory has carried on the simpleintroduction. At the same time, the current main fuzzy clustering algorithm are summarizedand analyzed. Finally, some important fuzzy clustering algorithms are introduced in detail.And their advantages and disadvantages are put forward.(3) Aiming at the fact that traditional fuzzy c-means algorithm easily falls into localoptimum and is easily affected by noise, this paper proposes a new clustering method ofimproved fuzzy clustering combined with genetic algorithm. The new algorithm makes useof good anti-noise performance of the improved FCM algorithm and global search ability ofgenetic algorithm.(4) Using standard IRIS data as test sample set, the improved algorithm is verified to prove its effectiveness. Then it is applied to the road recognition by using real road loadspectrum data. According to the result of clustering, the unknown road surface can beeffectively identified.(5) In order to verify clustering results of fuzzy kernel clustering method, using thelinear inseparable experimental data set, we compare the experimental results of FCMalgorithm and fuzzy kernel clustering algorithm. Then fuzzy kernel clustering algorithm isapplied to the road recognition by using real road load spectrum data. The recognition rateof the pavement is improved.