Since decades fingerprints have been the prime source in identification of suspects latent fingerprints are compared and examined with rolled and plain fingerprints which are stored in the dataset. The common challenges which are faced while examining latent fingerprints are background noise, nonlinear distortions, poor ridge clarity and partial impression of the finger. As conventional methods of Segmentation doesn’t perform well on latent fingerprints. The current advancement in machine learning based segmentation approach has been showing good results in terms of segmentation accuracy but lacks to provide accurate result in terms of matching accuracy. As one of the problem faced in matching latent fingerprint is low clarity of ridge-valley pattern which results in detection of false minutiae and poor matching accuracy. A multilayer processing of artificial neural network based segmentation is proposed to minimize the detection of false minutiae and increase the matching accuracy. This approach is designed on binary classification model where the simulation will be carried out on IIIT-D latent fingerprint dataset. Segmentation will be divided into full and partial impression fingerprints which are then compared with minutiae with the database using local and global matching algorithm. An improvised result is received which is more accurate as compared to the previous algorithms.

中文翻译：

---

潜在指纹索引，可通过图像增强技术更快地从数据集中检索

几十年来，指纹一直是识别可疑嫌疑人的主要来源，将潜伏指纹与存储在数据集中的滚动指纹和普通指纹进行比较和检查。检查潜在指纹时面临的常见挑战是背景噪声，非线性失真，不良的脊线清晰度和手指的部分印象。由于传统的分割方法在潜在指纹上表现不佳。基于机器学习的分割方法的最新进展在分割精度方面已显示出良好的结果，但在匹配精度方面却无法提供准确的结果。作为匹配潜在指纹所面临的问题之一是脊-谷图案的清晰度低，这导致检测到错误的细节和较差的匹配精度。提出了一种基于人工神经网络的分割的多层处理方法，以最大程度地减少对错误细节的检测，并提高匹配精度。这种方法是在二进制分类模型上设计的，该模型将在IIIT-D潜在指纹数据集上进行仿真。分割将分为全部和部分印象指纹，然后使用局部和全局匹配算法将其与细节进行数据库比较。收到的即席结果比以前的算法更准确。分割将分为全部和部分印象指纹，然后使用局部和全局匹配算法将其与细节进行数据库比较。收到的即席结果比以前的算法更准确。分割将分为全部和部分印象指纹，然后使用局部和全局匹配算法将其与细节进行数据库比较。收到的即席结果比以前的算法更准确。