Родное программное обеспечение для распознавания лиц

Обзор

Система распознавания лиц и сопоставления, созданная с помощью алгоритмов обработки изображений и распознавания образов, была разработана независимо от аппаратного обеспечения, лицензии и источника для создания системы распознавания лиц, которая является одним из все более важных механизмов биометрического обнаружения. Он состоит из этапов обнаружения лиц и сопоставления приложений.

Архитектура облачной системы: все точки, в которых будет применяться система, контролируются централизованно. После биометрической регистрации передача записей в базу данных и решения, подходящие для нужд пунктов обслуживания, предоставляются централизованно.
Однократное и многократное использование: все точки, в которых будет применяться система, контролируются централизованно. После биометрической регистрации передача записей в базу данных и решения, подходящие для нужд пунктов обслуживания, предоставляются централизованно.

Mia Teknoloji Yerli Görüntü İşleme ve Örüntü Tanıma Algoritmalarıyla Oluşturulmuş Yüz Tanıma ve Eşleştirme Yazılımı

Это родная система распознавания лиц в соответствии со стандартами NIST. Это первый инновационный аспект системы. С другой стороны, выполнение предварительных процессов, распознавание лиц, распознавание лиц (сопоставление) и совместные операции с сервером вместе с разработками платформы представляют собой еще один инновационный аспект.

Highlights

In order to have the source photo quality at the desired rates, 95.9% in the matching phase, 91% in the identification phase; In case the source photo quality is not at the desired rates, depending on the quality, it provides 0-92.5% accuracy in the matching phase and 0-88.7% in the identification phase.

Accuracy

Cross-match matching of different algorithms to be selected according to Age, Gender, Race and Crowd ratios with an integrated upper software provides an accuracy increase in the range of 0.1-50% according to different scenarios.

Algorithm Effect

As the number of people registered in the database increases, a matching pre-order is provided through the algorithm so that the increase in similar face rates does not reflect negatively on the accuracy.

Effect of Increase in Database
Sizes

A ranking algorithm is obtained that can be used in criminal and forensic cases where the threshold value is kept very low. Thus, in order to assist decision-making mechanisms, a ranking proportional to the accuracy percentage is provided.

Sort by Criteria for Human Impact Searches

According to the researches, it has been determined that the old people are easily and accurately defined in all algorithms, while the young people are defined difficult and incorrectly. Accuracy has been increased for younger age groups with the efficient algorithm structure and exponential search method.

Accuracy in Age Groups

Increasing the source photo quality is the most important factor in identification. It enables to select the quality conformity of the source photos in the databases with automation. (Controlling eye spacing, light value and conformity of face posture with software)

Quality

This speed decreases as the number of people registered in the database increases and the quality of the captured image increases depending on the hardware. In order to prevent this speed reduction; Intelligent classification over the database and the definition to be made take place in appropriate subfolders from the very beginning.

Time

The accuracy rate in worldwide algorithms leads to an error rate of approximately 1-3% per year. Thanks to the age-face analysis to be carried out throughout the project, this effect is reduced to 0.8-2.5%.

Age Increase and the Effect of Old Records

Big data management also gains importance in the stages of receiving and storing biometric data. In addition, cutting only the necessary biometric part from the image will be automatically provided in the project content, thus providing great reductions in data sizes.

Data Creation

Depending on the 1:N dimensions, the screen display per second can be flexibly adjusted during the acquisition and processing of multiple data. This method provides an increase in processing speed in the range of (x0.1- x2), provided that the processing speed is variable within the scenarios.

Drivers and Image Processing

By taking into account variables such as camera flow, number of people, workload losses, loss matching rates, processing time, age effects, erroneous transactions are corrected by the algorithm retrospectively and reprocessing.

Error and Speed Variables