研究场景构建中纹理图像的三维特征识别,提高识别准确率.由于图像的效果取决于纹理识别,在三维空间中分析纹理图像时,仅使用聚类算法利用三维图像的空间坐标和三维像素灰度特征对其进行分类识别,忽略了不同方向光照对纹理图像的影响而简单聚类分类,导致因提取的特征信息不足而造成识别准确率不高的问题.因此,提出了一种机器学习的三维特征识别方法.通过变换光照的角度得到不同光照下的纹理图像,提取出多面光照下的三维图像特征信息,并利用机器学习算法对特征信息进行准确训练和分类识别,可避免聚类算法利用信息不足的特征进行分类而造成的识别准确率不高的问题.实验表明,这种方法能够有效提取出特征信息并进行准确分类,具有较高的识别准确率,取得了满意的结果.%Research the construction of texture image of characteristics identification to improve identification ac-curacy. In 3-d space image texture analysis, only using clustering algorithm, the 3-d image space coordinates and the 3-d pixel grayscale characteristics to classify and recognise the images, ignores the influence of different direction lights on the texture images, leading to the insufficient information for extraction and causing low identification accura-cy. Therefore, the paper put forward a 3-d characteristics identification method hased on machine learning. Through changing the illumination angles to get different texture images, machine learning algorithm was used to extract the 3 -d image features, train, classify and identify the 3-d image. Experimental results show that this method can effec-tively extract the features and accurately classify the images. It has a higher identification accuracy and satisfactory results.
展开▼
