Loading...
Publication
Sistema de medição de distâncias para aplicações médicas com base na Câmara NanEye Stereo
| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 3.51 MB | Adobe PDF |
Abstract(s)
O presente projeto visou desenvolver um sistema que permita efetuar medições de distâncias em aplicações médicas, como por exemplo, a medição de pólipos e de outras alterações morfológicas, em contexto de endoscopia, para a determinação de diagnósticos médicos, utilizando o sistema NanEye Stereo criado pela empresa AMS Sensores Portugal. O algoritmo proposto por este trabalho poderá ser englobado no projeto do NanEye Stereo, como sendo parte integrante deste, adicionado à biblioteca de software utilizada pelo sensor. Os dois requisitos necessários a considerar são a eficácia, para determinar dimensões das estruturas, e o tempo de resposta, que deverá ser baixo, de forma a não afetar o desempenho do resto do sistema.
A forma como a aprendizagem das estruturas é feita neste projeto aproxima-se das redes neuronais, nomeadamente ao comportamento dos olhos, uma vez que através de duas perspetivas diferentes, o cérebro processa a informação dando a noção de profundidade.
O algoritmo criado permite calcular profundidades e dimensões de estruturas com base na correlação de pontos sobre os alvos nas imagens stereo, de forma simples e precisa. Tendo em conta as pequenas dimensões do sensor NanEye, o sistema criado permite operar a profundidades de 2 a 26 mm. No entanto foram encontradas dificuldades no que concerne à correlação do ponto sobre o alvo nas duas imagens stereo, por apresentarem ligeiras diferenças na imagem obtida por cada um dos sensores de imagem. De modo a ultrapassar este obstáculo, foi estudada e trabalhada a correlação das imagens, originando o desenvolvimento de complexos processos que visam a melhoria da qualidade nas duas imagens usadas para o processamento das distâncias.
This project aimed to develop a system that allows the measurement of distances in medical applications, such as essential endoscopies for the determination of diagnoses, for example such as the measurement of polyps and other morphological changes, in the context of endoscopy, for the determination of medical diagnoses using the NanEye Stereo system created by the AMS Sensors Portugal, where this project took part. The algorithm proposed by this work can be encompassed within the design of the NanEye Stereo project, as an integral part of the project, added to the software library used by the sensor. The two requirements are an effective method for determining frame dimensions and a low response time, so as not to affect the performance of the rest of the system. The way in which learning of structures is done in this project is close to neural networks, in particular to eye behavior, once through two different perspectives it is possible to send the information to the brain and have the notion of depth. The created algorithm allows us to calculate depths and dimensions of structures based on the correction of the points on targets in the stereo images, in a simple and precise way. Given the small dimensions of the NanEye sensor, the system created can operate at depths from 2 to 26 mm. However, difficulties related to the correlation of the points on the target were found in the two stereo images, because they present slight differences in the image obtained by each of the image sensors.. In order to overcome this obstacle, the correlation of images was studied and worked on, leading to the development of complex processes that aim at improving the quality of the two images used to process the distances.
This project aimed to develop a system that allows the measurement of distances in medical applications, such as essential endoscopies for the determination of diagnoses, for example such as the measurement of polyps and other morphological changes, in the context of endoscopy, for the determination of medical diagnoses using the NanEye Stereo system created by the AMS Sensors Portugal, where this project took part. The algorithm proposed by this work can be encompassed within the design of the NanEye Stereo project, as an integral part of the project, added to the software library used by the sensor. The two requirements are an effective method for determining frame dimensions and a low response time, so as not to affect the performance of the rest of the system. The way in which learning of structures is done in this project is close to neural networks, in particular to eye behavior, once through two different perspectives it is possible to send the information to the brain and have the notion of depth. The created algorithm allows us to calculate depths and dimensions of structures based on the correction of the points on targets in the stereo images, in a simple and precise way. Given the small dimensions of the NanEye sensor, the system created can operate at depths from 2 to 26 mm. However, difficulties related to the correlation of the points on the target were found in the two stereo images, because they present slight differences in the image obtained by each of the image sensors.. In order to overcome this obstacle, the correlation of images was studied and worked on, leading to the development of complex processes that aim at improving the quality of the two images used to process the distances.
Description
Keywords
Algoritmo Profundidade Distância Aplicações médicas Distorção de imagem NanEye Stereo Algorithm Depth Distance Medical applications Distortion of image Engenharia Eletrotécnica - Telecomunicações . Faculdade de Ciências Exatas e da Engenharia