Big data human face recognition system

Study Abroad Would like to continue in the graduate student this road has been to the black, but time is not enough, the silver in the pocket is always not enough to use , The study of Things is also a little , There is no way to write a paper handed over the graduation, a sigh, think of school for more than 20, always feel that nothing is enough, especially the brain , sometimes even forget to take Lab life is also a relaxation decided to go out to work, but before that, simply sort out the contents of the study, The right to be a catalyst, forget not to see this joke , future content will be biased and data protection DLP Field <- I am engaged in the work let's get back to the chase. : Big Data This field is absolutely inevitable. , especially image recognition / Two areas of speech recognition This is definitely the two big problems that must be solved in the field of artificial intelligence This has begun to bear fruit . words like the ultimate goal of big data clusters have been misunderstood by many people , Many people feel that they can pass Hadoo P The formation of cheap clusters is High , a little treasure-raising consciousness >.< Google still short of money ..... Is Artificial intelligence so far not independent of the strong brain ( There is better ), but need a a robust cluster and a Strong coverage Communication network These two areas of domestic start late study less slow progress Enlightenment is also surprisingly low, all aspects have to Burenhouchen This time, the content is how to classify high-volume images by face recognition technology on the Hadoop platform . Tagging and classifying facial Images in Cloud environments Based on KNN using MapReduce The content is simple to write, not complex, I hope for the students to help get started Here are a number of keywords: Hadoop /mapreduce, Human Face image recognition , Classification Algorithm First of all, get familiar with the picture. With the sharp increase in the number of images, not to mention social networking sites or the like, the world's Daily collection of CCTV images are enough to drink a pot of Storage technology is also decades in situ, this is something, will slowly involve this aspect , the corresponding picture processing technology still stays in the 560 's , tens of billions of photos how to handle this is a very big problem. , So here's a hypothetical scenario. : How to analyze the country through real-time Cctv/sns image of the face in the site image tracking criminals Hadoop making it possible for a cluster to process pictures , image recognition technology has not followed the Times , the stove was replaced with a rice cooker , cooks is still the original cooks, is not boiled out feast But it doesn't affect our practice of practiced hand. here the most primitive and basic simplest picture feature extraction algorithm : pca  principal component Analysis , ( ICA , LDA these codes write , did not try) The classification algorithm also uses the most primitive simplest and most brutal classification algorithm : KNN talked about PCA dimensionality Reduction algorithm, I think of a multidimensional space problem and matrix dimensionality reduction , in my opinionmultidimensional space isoverlap of three-dimensional space, for example, two three-dimensional space two-dimensional coincidence is4dimension space, two x4The three-dimensional coincidence of the dimensional space is5Dimension Space,etc.... Matrix dimensionality reduction: The existing matrix dimension is relatively low, if it is a 10000*10000 matrix, how to reduce dimensions, this slightly studied, will be a separate discussion. PCA/KNN the interpretation of the correlation algorithm can be written in a few separate, and later to see the empty , There are a lot of online information as shown in this picture, the feature factor is extracted by the image learning first , and then map the space to a certain dimension. , Finally, using KNN algorithm to find the closest value ( human face ) the entire construction map is put into Hadoop The post-process flowchart will be as follows What, you asked me the effect the speed is fast. the evil image recognition technology , How bad is the oil? , Recognition rate General Think of it and mend it. Conclusion: The utility is very strong, but the algorithm needs to be improved , and the time is limited, and no more advanced algorithm is used to calculate practiced hand. Initiate

Big data human face recognition system