基于证据理论融合多特征的物体识别算法

摘要

为了提高物体的识别正确率，提出一种基于证据理论融合多特征的物体识别算法。提取物体图像的颜色直方图和尺度不变特征，采用极限学习机建立相应的图像分类器，根据单一特征的识别结果构建概率分配函数，并采用证据理论对单一特征识别结果进行融合，得出物体的最终识别结果，采用多个图像数据库对算法有效性进行测试。测试结果表明，该算法不仅提高了物体的识别率，而且加快了物体识别的速度，具有一定的实际应用价值。%In order to obtain better recognition results, a novel object recognition method based on multi-feature fusion of evidence theory is proposed. Color histogram and scale invariant feature transform features are extracted from object image, and extreme learning machine is used to establish the classifier;the recognition results of single feature are fused to obtain the last recognition results of object based on evidence theory;the performance of algorithm is tested by some image data. The result illustrates that the proposed algorithm has improved the recognition rate and speed, and it has some application vale.