雾霾天气条件下车牌信息的识别

雾霾天气条件下,由于大气的散射,降低了拍摄图片中车牌信息的清晰度和对比度,并降低了车牌识别的正确率。针对这一现象,提出了一种基于暗原色的对图像透射率进行改进的算法,通过改进后的方法对图像进行去雾,弥补了暗原色方法针对天空、白色等大片明亮区域无法很好去雾的缺点。算法首先对雾霾天气下拍摄的图像利用改进算法进行去雾处理,然后进行车牌定位和字符分割,最后通过BP神经网络进行车牌信息的识别。实验证明,通过改进后的方法对图像进行去雾后,能够很好地还原车辆信息的原本颜色特征,给后期的车辆信息处理提供了便利。

License Plate Identification in Foggy Weather Conditions

The clarity and contrast of the vehicle license plate information are reduced in the foggy and hazy weather condition because of atmospheric scattering, and the vehicle license plate recognition accuracy is also reduced. In this paper we present a new defogging method based on dark channel prior. The new method improve the image transmission. The shortcoming of the dark channel prior which can not defog in the large bright area of the sky and white area is overcome. First , the image captured under the hazy weather is processed by this new defogging method, and then the license plate is located and the characters are separated. Finally, the license plate information is identified by the BP neural network. Experimental results show that the original color characteristics of the vehicle information can be restored very well by this new defogging method. It also provides the convenience for the post-processing of the vehicle information.