显示图像复杂目标优化识别仿真研究

摘要

对显示图像复杂目标进行优化识别,可以有效提高人机界面中目标识别的速度和准确度.对图像目标的识别需要提取图像信息颜色和纹理特征,并通过决策融合处理完成图像目标的识别.传统方法利用邻域灰度极值提取边界候选目标图像,选取边界候选图像像素的二值模式作为样本集,但忽略了对其进行决策处理,导致识别精度低.提出基于改进证据理论的显示图像复杂目标识别方法.利用Canny算子将界面图像目标的边缘像素识别出来,并组建归一化直方图,提取可以反映目标图像类别信息的颜色和纹理特征,利用神经网络对单一目标特征进行初步识别,将识别的结果作为证据,采用证据理论对初步识别结果进行决策融合处理,利用处理的结果完成对显示图像复杂目标识别.实验结果表明,所提方法目标识别精确度高,可以快速有效地完成对显示图像复杂目标识别.%This article proposes a recognition method for complex target of display image based on modified evidence theory.Canny operator was used to identify edge pixel of interface image target image.Normalization histogram was built in the meantime to extract color and textural feature reflecting class information of target image.Then,neural network was used to identify single target feature preliminarily.Using the result of identification as evidence,the evidence theory was used to carry out decision fusion for result of preliminary identification.Finally,recognition for complex target was completed using the processed result.Following conclusion can be drawn from experiment.The method has high precision of target recognition.It can complete the recognition rapidly and effectively.