News classification
Contact us
- Add: No. 9, North Fourth Ring Road, Haidian District, Beijing. It mainly includes face recognition, living detection, ID card recognition, bank card recognition, business card recognition, license plate recognition, OCR recognition, and intelligent recognition technology.
- Tel: 13146317170 廖经理
- Fax:
- Email: 398017534@qq.com
Research on Face Recognition Algorithm of Embedded System
Research on Face Recognition Algorithm of Embedded System
In this paper, based on S3C2400 embedded devices, the use of Linux operating system in the JZ2440 to achieve a complete set of embedded face recognition system, due to the embedded camera is relatively fixed position, so the main focus on the face recognition part . Mainly divided into the following two steps:
(1) image acquisition. Use the camera to obtain dynamic pictures, when the need to identify, just touch the screen, you can save the picture down.
In this case,
(2) face recognition. At this point, the image is preprocessed and then identified by PCA + Euclidean distance or PCA + SVM.
6 Experimental results analysis
In this paper, we use the method of PCA and SVM to perform experiments on the ORL face image database (shown in Fig. 1) and the human face image database of our laboratory (Fig. 2). The experiment proves that the recognition works well. ORL face database has 40 individuals, each with 10 photos, this paper uses a person's first 5 photos as an experimental sample, after 5 as a test sample. As shown in Figure 1, 1 to 5 for a person's training sample photos, the corresponding test sample is 10 001 ~ 10 005. The same training sample 6 ~ 10 corresponds to the test sample is 10 006 ~ 10 010.
In order to verify the versatility of this method, but also the use of some laboratory personnel of the face images, recognition rate can still achieve a good recognition effect (as shown in Table 2).
Image 004.png
In this case,
7 Conclusion
The recognition rate of PCA + SVM is up to 94% in ORL face image database, and 98% in PCA + SVM. In the case of BP neural network, the initial training time is more, The experimental results show that the neural network is not time-consuming, and the experimental results show that the accuracy of SVM is not high, so the final selection of this paper PCA + SVM as an Experimental Method for Embedded Face Recognition.
(1) image acquisition. Use the camera to obtain dynamic pictures, when the need to identify, just touch the screen, you can save the picture down.
In this case,
(2) face recognition. At this point, the image is preprocessed and then identified by PCA + Euclidean distance or PCA + SVM.
6 Experimental results analysis
In this paper, we use the method of PCA and SVM to perform experiments on the ORL face image database (shown in Fig. 1) and the human face image database of our laboratory (Fig. 2). The experiment proves that the recognition works well. ORL face database has 40 individuals, each with 10 photos, this paper uses a person's first 5 photos as an experimental sample, after 5 as a test sample. As shown in Figure 1, 1 to 5 for a person's training sample photos, the corresponding test sample is 10 001 ~ 10 005. The same training sample 6 ~ 10 corresponds to the test sample is 10 006 ~ 10 010.
In order to verify the versatility of this method, but also the use of some laboratory personnel of the face images, recognition rate can still achieve a good recognition effect (as shown in Table 2).
Image 004.png
In this case,
7 Conclusion
The recognition rate of PCA + SVM is up to 94% in ORL face image database, and 98% in PCA + SVM. In the case of BP neural network, the initial training time is more, The experimental results show that the neural network is not time-consuming, and the experimental results show that the accuracy of SVM is not high, so the final selection of this paper PCA + SVM as an Experimental Method for Embedded Face Recognition.