复杂光照下DPM图像自适应多阈值分割方法研究

Adaptive multi-thresholds segmentation of DPM barcode image in complex illumination

复杂工况下,CCD相机采集到的DPM(Direct Part Mark)工业二维码图像受光照影响易出现大片光斑或阴影区域,造成DPM区域的信息遗漏,从而导致识别困难。为此,提出一种基于分段直方图凹度分析的多阈值自适应分割算法。首先在灰度直方图平滑的基础上计算出系列局部峰值,并借此完成直方图分段,再递推计算出每分段区域下凹处的分割阈值。其次通过引入基于阈值点局部区域信息的修正因子,使分割阈值自适应变化而更适用于局部对比度较低的状况。实验结果表明,该方法分割效果优于经典的阈值分割算法,平均运行效率比最快的多阈值分割算法提高17.75倍。经自适应局部阈值分割后,DPM图像复杂光照区域有用信息得以增强,缺失信息得以弥补,为后续的对象识别奠定基础。该方法也可推广于对比度多变的图像增强。

Under complex industrial conditions,two-dimensional DPM barcode captured by a CCD camera easily has large spots or shadow areas owing to the complex illumination.This phenomenon results in the missing information in DPM area and the identification difficulty.Therefore,this paper proposes adaptive multi-threshold segmentation algorithm based on subsection histogram concavity analysis.Firstly,on the basis of smoothed histogram,a series of local peak values are calculated by the simplified formula.Moreover,the histogram is segmented through these local peak values.Then subsection thresholds are computed by recursive algorithm.Secondly,an adaptive correction factor based on the local area information is introduced to modify the subsection threshold for the status of the low local contrast.Experimental results show that the proposed method has superior division performance and more efficient operation to traditional threshold segmentation algorithms.The average running efficiency of this method improves 17.75 times than the fastest one of those conventional algorithms.After the adaptive multi-threshold segmentation,the contrast of uneven illumination area is significantly enhanced and missing DPM regional information is effectively compensated.Therefore,the method in this paper provides a sufficient condition for the accurate identification of the DPM barcode.It can also be applied to the contrast changeable image enhancement.