| 摘要: |
| 为有效管控黑烟排放并保护生态环境,本文提出一种改进的柴油车烟度分级识别方法。首先,针对轻级黑烟被错分至背景的问题,提出了多阈值分割方法。从黑烟扩散特性角度出发,在原有OTSU算法基础上引入新的阈值,用以分离轻级黑烟与背景,对输入黑烟图像实行双阈值、双区域分割策略,这一改进提升了算法对轻级黑烟的分割精度,并将黑烟分割为双区域,从而降低类内差异,这种分区处理为下文捕捉黑烟特征奠定基础。其次,针对黑度值计算不准确的问题,引入一种权重标定机制。通过收集1-4级黑烟数据样本,利用改进后的算法对黑烟进行分割,得到双区域x1、x2,将黑烟真实等级转化为对应的黑度值y,利用最小二乘法拟合,建立了计算机生成的黑度值与实际黑烟浓度之间的映射关系。结果表明:该方法有效反映了黑烟真实黑度,提高黑烟等级评估的精准性,在构建的数据集上,所提方法准确率达到91.06%,显著提升了黑烟监测与评估的可靠性。 |
| 关键词: 烟度分级 OTSU 柴油车黑烟 图像分割 |
| DOI： |
| 分类号:X831 |
| 基金项目:城市道路机动车排气污染物在线监测及表方法征研究,黑龙江省自然基金,E2017001；不同道路类型与交通环境下人-车交互影响机理,国家重点研发计划,2017YFC0803901-2 |
|
| Research on the Identification Method of Diesel Vehicle Emission Smoke Level Based on Improved OTSU Algorithm |
|
xuxin
|
|
东北林业大学
|
| Abstract: |
| To effectively control black smoke emissions and protect the ecological environment, an improved method for classifying diesel vehicle smoke opacity is proposed in this paper. To address the issue of light-colored smoke being incorrectly classified as background, a multi-threshold segmentation method is proposed. From the perspective of black smoke dispersion characteristics, new thresholds were introduced to the existing OTSU algorithm to separate light smoke from the background. A dual-threshold, dual-region segmentation strategy was applied to input black smoke images. This improvement enhanced the algorithm"s accuracy in segmenting light smoke and divided the black smoke into two regions, thereby reducing intra-class variance. This segmentation process laid the foundation for capturing black smoke features in subsequent analyses. To address the issue of inaccurate blackness value calculation, a weight calibration mechanism was introduced. This adjustment improved the accuracy of the blackness values and refined the overall segmentation process. By collecting smoke data samples of levels 1 to 4, the improved algorithm was used to segment black smoke into two regions, x1 and x2. The true levels of black smoke were then converted into corresponding blackness values, y. A least squares fitting method was employed to establish the mapping between the computer-generated blackness values and the actual black smoke concentrations. The results indicate that the method effectively reflects the true blackness of black smoke and improves the accuracy of black smoke level assessments. The proposed method achieved an accuracy of 91.06% on the constructed dataset, significantly enhancing the reliability of smoke monitoring and evaluation. |
| Key words: Smoke level OTSU Diesel car black smoke image segmentation |
