Image Classification Based On Support Vector Machine

This paper researches on image classification based on support vector machine(SVM) and concentrates on the images of construction roads. Using SVM classification technol-ogy, we can predict whether the road is clean. Construction roads, often with construction vehicles, is easy to be full of dust. Whenever there is a car passing by, the sky is full of dust and people become a "vacuum cleaner". Also the vehicle would take these dusts to city, which influences the cleanliness of the city. Once the dust reaches a certain level, the road need be cleaned. Because the road is near workplace, the dust pollution has high frequency and uncertainty which leads to the difficulty of human monitoring. This paper proposes a system which can be real-time monitoring the road based on the SVM classify method. System will continue extracting images from the camera, recognizing image and determining whether the road needs to clean. This is the practical significance of this paper.In image recognition based on SVM classification, the validity and accuracy of fea-ture extraction of image is directly related to the final result of the system. Because the ve-hicles and pedestrians on the road have an effect on road’s features extraction, the system must segment the image for separating the road and others. The System uses road pixels means, variances, skew and kurtosis as image characteristics. This paper compares vari-ous saliency detection means and clustering segmentation methods, then chooses a better segmentation method. This paper proposes a new algorithm which combines saliency detection method and clustering segmentation algorithm, which can improve the accu-racy of road extraction. The system has low algorithm complexity, high efficiency and accuracy. This problem exists in many places, so it has a very good market prospect.