基于深度卷积神经网络的模糊字迹图像识别方法

Fuzzy Handwriting Image Recognition Method Based on Deep Convolution Neural Network

在传统的模糊字迹图像识别过程中,忽略了字迹变化尺度对图像的影响,导致识别准确度低识别能力差的问题,提出基于深度卷积神经网络的模糊字迹图像识别方法。通过图像的退化模型,对模糊字迹图像稀疏性特征进行分解,构建模糊字迹图像的多源特征参数检测模型,结合边缘轮廓特征提取方法实现对模糊字迹图像的边界信息采样分析;采用多维参数模拟和模糊度增强处理,结合匹配滤波检测器对图像的多级尺度分解和细节特征进行提取,对提取的模糊字迹图像细节特征进行融合和优化检测,采用深度卷积神经网络训练方法进行模糊字迹图像修复处理,实现模糊字迹图像的识别。仿真结果表明,采用该方法进行模糊字迹图像识别的准确性较高,检测能力较强,提高了模糊字迹图像修复和辨识能力。

In the process of traditional fuzzy handwriting image recognition,the influence of handwriting change scale on image pair is ignored,which leads to the problem of low recognition accuracy and poor recognition ability.A fuzzy handwriting image recognition method based on deep convolution neural network is proposed.Through the degradation model of the image,the sparse feature of the fuzzy handwriting image is decomposed,and the multi-source feature parameter detection model of the fuzzy handwriting image is constructed.Combined with the edge contour feature extraction method,the boundary information of the fuzzy handwriting image is sampled and analyzed.The multi-dimensional parameter simulation and fuzzy degree enhancement processing are used,and the multi-level scale decomposition and fine-tuning of the image are combined with the matched filter detector.Detail features of the extracted fuzzy handwriting image are fused and optimized.The fuzzy handwriting image is repaired by using the deep convolution neural network training method to realize the recognition of fuzzy handwriting image.The simulation results show that the method has high accuracy and strong detection ability,and improves the ability of fuzzy handwriting image restoration and recognition.