Nail Location And Disease Recognition System Based On Perceptual Hash Algorithm And Deep Learning

With the rapid development of information technology,how to use these technologies to fast and efficiently help patients or doctors to diagnose has been a research hotspot problem.The nail can change along with the appearance of human diseases.The nail is a window which can measure the indicator of our health,so we can judge the state of our body by observing nails.If people can discover the disease by checking nails at the beginning of illness and exhaustion,we can prevent disease and get rid of the disease by doing some exercise and going to sleep to enhance our immune system.At present,the continuous improvement and enhancement of artificial intelligence technology have laid a strong foundation for the nail detection and recognition technology in videos and images.In this thesis,through constructing the nail database and utilizing image processing and machine learning methods,some nail disease detection algorithms are proposed and the corresponding nail disease detection systems are designed.It can raise people's understanding of nails and diseases.We can detect the nail disease and give the initial diagnosis and treatment suggestions according to different nail disease at the same time.This system can also detect the nail disease from the video which can improve the experience of users and have a great influence on healthcare and diagnosis.The details are given as follows:(1)In order to reduce the influence of the skin on the finger on the nail image retrieval result,an integrated method based on the RGB color threshold and illumination compensation is used to separate the skin around the nail and extract useful nail information.Based on the separated nail information,a comprehensive nail detection method combining HSV color space and perceptual hash texture feature was proposed and the corresponding system was developed.Experimental results on images of different nail diseases show that the newly proposed method can detect different nail diseases and give some further instruction.(2)In order to extend the application of nail disease detection methods and to address the problem of nail disease detection in video and images,this paper first proposed a video and image nail locating method based on LBP method,and designed a nail disease detection model based on a convolutional neural network to improve the detection performance.The nail detection and recognition system based on deep learning can express nail information more intuitively and realistically,play an auxiliary role in disease diagnosis,and has certain application prospects.(3)In the matching phase by using the similarity among nail images,there is a problem that the matching speed is too slow.In this paper,we use multi-threading technology to transfer the tasks that need a long execution time to the background.It can reduce the matching time and improve the retrieval speed and efficiency of the system.The nail detection and recognition system which is based on deep learning,it can facilitate the understanding of the relationship between nails and disease for people,and it provides a new form of information technology assisted medical diagnosis.