A real-time two-stage and dual-check template matching algorithm based on normalized cross-correlation for industrial vision positioning

One classic algorithm usedin template matching is normalized cross correlation method. It often achieveshigh precision. But it does not meet speed requirements for time-criticalapplications. To solve that issue, a speed-up way of template matching isproposed. The fast matching way bases on pyramid hierarchical searchingalgorithm. It adopts two template matching methods to match images, which baseon rough matching proceeds local matching precision. Firstly, the coarsematching is performed based on gray-scale projection algorithm. Secondly, theprecise matching is made based on several small block matching. The new way iscompared to conventional approach without pyramid hierarchical searching byexperiments. Experimental result demonstrates that the proposed way efficientlyimproves the speed of template matching and the precision is unchanged.

Download Full-text

Identifikasi Ciri Garis Telapak Tangan Berbasis Template Matching dan Metode K-Nearest Neighbor

MIND Journal ◽

10.26760/mindjournal.v3i2.25-35 ◽

2019 ◽

Vol 3 (2) ◽

pp. 25-35

Author(s):

Asep Nana Hermana ◽

Irma Amelia Dewi ◽

Irwan Susanto

Keyword(s):

Template Matching ◽

Cross Correlation ◽

Nearest Neighbor ◽

K Nearest Neighbor ◽

Normalized Cross Correlation

Telapak tangan merupakan ciri unik yang dimiliki oleh manusia yang dapat digunakan pada sistem identifikasi. Proses template matching membutuhkan perhitungan pencocokan untuk menentukan bagian kecil gambar yang memiliki nilai terbesar dikarenakan semakin besar nilai maka tingkat kecocokan semakin tinggi. Sehingga untuk pencocokan dibutuhkan perhitungan normalized cross correlation dengan perhitungan konvolusi yang setiap bagian pixel akan dilakukan pencocokan, diawali dari pixel bagian pojok kiri atas hingga pojok kanan bawah dan akan mendapatkan nilai pencocokan terbesar.Setelah mendapat nilai terbesar dilakukan k-nearest neighbor yang merupakan pengelompokan berdasarkan jarak dan untuk menentukan jarak k digunakan perhitungan euclidien distance. Selanjutnya pengelompokan berdasarkan voting terbanyak yang dimulai dari nilai jarak ketetanggaan terkecil hingga terbesar. Tingkat akurasi pengujian dari 30 sampel telapak tangan didapatkan presentase sebesar 86,67% teridentifikasi benar dan 13,33% salah.

Download Full-text

Optical Disc and Blood Vessel Segmentation in Retinal Fundus Images

Academic Journal of Nawroz University ◽

10.25007/ajnu.v8n3a398 ◽

2019 ◽

Vol 8 (3) ◽

pp. 67

Author(s):

Amira B. Sallow ◽

Hawkar Kh. Shaikha

Keyword(s):

Blood Vessel ◽

Template Matching ◽

Cross Correlation ◽

Region Of Interest ◽

Processing Technique ◽

Image Processing Technique ◽

Original Image ◽

Matching Algorithm ◽

A New Technique ◽

Retinal Fundus

Segmentation of optical disk (OD) and blood vessel is one of the significant steps in automatic diabetic retinopathy (DR) detecting. In this paper, a new technique is presented for OD segmentation that depends on the histogram template matching algorithm and OD size. In addition, Kirsch method is used for Blood Vessel (BV) segmentation which is one of the popular methods in the edge detection and image processing technique. The template matching algorithm is used for finding the center of the OD. In this step, the histogram of each RGB (Red, Green, and Blue) planes are founded and then the cross-correlation is founded between the template and the original image, OD location is the point with maximum cross-correlation between them. The OD size varies according to the camera field of sight and the resolution of the original image. The rectangle size of OD is not the same for various databases, the estimated size for DRIVE, STARE, DIARTDB0, and DIARTDB1 are 80×80, 140×140, 190×190, and 190×190 respectively. After finding the OD center and rectangle size of OD, a binary mask is created with Region of Interest (ROI) for segmenting the OD. The DIARTDB0 is used to evaluate the proposed technique, the result is robust and vital with an accuracy of 96%.

Download Full-text