基于機(jī)器視覺(jué)的圖像亞像素測(cè)量技術(shù)的應(yīng)用研究

1、 論文題目：基于機(jī)器視覺(jué)的圖像亞像素測(cè)量技術(shù)的應(yīng)用研究 專 業(yè)：通信與信息系統(tǒng) 碩 士 生： 吳軍妮 (簽名) 指導(dǎo)老師： 李白萍 （簽名） 摘 要 近年來(lái)隨著傳統(tǒng)制造業(yè)的轉(zhuǎn)型升級(jí)，工業(yè)自動(dòng)化技術(shù)的大規(guī)模應(yīng)用，尤其是在半導(dǎo)體和消費(fèi)電子產(chǎn)業(yè)。產(chǎn)品大批量生產(chǎn)、對(duì)質(zhì)量和加工精度要求越來(lái)越高。在

2、這種情況下要求檢測(cè)設(shè)備也要有很高的檢測(cè)速度和檢測(cè)精度。提高檢測(cè)精度可以采用更換硬件和優(yōu)化軟件兩種方法。但更換硬件代價(jià)較高，效果也并非很理想。所以在合理的范圍內(nèi)，用軟件的方法提高檢測(cè)精度就成為一個(gè)重點(diǎn)研究方向。 本文開(kāi)發(fā)了一款專業(yè)檢測(cè)軟件，可以檢測(cè)產(chǎn)品的尺寸以及 ROI 區(qū)域各種的缺陷。軟件通過(guò) RS485 從 PLC 接受檢測(cè)信號(hào)，檢測(cè)結(jié)果也通過(guò) RS485 傳送給 PLC，用于控制收料機(jī)對(duì)產(chǎn)品進(jìn)行分類。由于泰勒多項(xiàng)式插值不依賴于插值像

3、素的坐標(biāo)位置，抗噪性強(qiáng)、對(duì)圖像對(duì)比度的變化不敏感。所以本文將經(jīng)典的雙線性插值與基于泰勒多項(xiàng)式的亞像素插值結(jié)合，提出了一種改進(jìn)的泰勒亞像素插值算法。該算法對(duì)工業(yè)現(xiàn)場(chǎng)的噪聲、光照的變化（俗稱光害） 、產(chǎn)品的清潔狀況（被測(cè)的工件表面可能有一些污漬）具有一定的魯棒性，在計(jì)算結(jié)果精度和穩(wěn)定性提高的同時(shí)檢測(cè)速度沒(méi)有明顯下降。 銅帽（帶線）是半導(dǎo)體電子行業(yè)中常用的一種電子元器件，是生產(chǎn)多種電子產(chǎn)品的原始材料。它的加工精度和質(zhì)量與最終的電子產(chǎn)品的壽命和

4、功耗有直接關(guān)系。本文結(jié)合蘇州福弘電子工業(yè)有限公司的銅帽（帶線）尺寸測(cè)量和缺陷檢測(cè)項(xiàng)目，將改進(jìn)的泰勒亞像素插值算法應(yīng)用到鉚點(diǎn)直徑和和錐點(diǎn)尺寸的測(cè)量中，與傳統(tǒng)的雙線性插值結(jié)果進(jìn)行對(duì)比，計(jì)算精度和穩(wěn)定性有明顯提高。改進(jìn)的算法在客戶工作現(xiàn)場(chǎng)的使用效果非常理想，微小的質(zhì)量問(wèn)題也檢測(cè)了出來(lái)，大大提高了產(chǎn)品的質(zhì)量。 關(guān) 鍵 字: : 機(jī)器視覺(jué)；銅帽（帶線）檢測(cè)；泰勒亞像素插值 研究類型：應(yīng)用研究 Subject : Research and Ap

5、plication of Image Sub-pixel Measurement Based on Machine Vision Specialty : Communication and Information System Name : Wu Junni (Signature) Instructor : Li Baiping

6、 (Signature) ABSTRACT With the development of the traditional manufacturing industry, Industrial automation technology’s large-scale application. Especially in the semiconductor and consumer e

7、lectronics industry, with the measurement precision, the products; quality, the detection speed needed is increasing. Require the detection device should have a high detection speed and detection precision. But simply in

8、crease the camera and the lens’ resolution to improve the detection precision, is not economical. So within a reasonable range, improve the detection precision of the method using the software is a very important develop

9、ment direction. This paper developed professional measure software, it can detect the size of the products and all kinds of defects in ROI areas. The software receives the detection signal from the PLC through the RS485,

10、 the results are sent to PLC through RS485, used to control the material receiving machine used to carries on the classification to the product. The taylor polynomial interpolation not depends on pixels’ coordinates, not

11、 sensitive to noise, also not sensitive to the change of the contrast of image. This paper combines the classical bilinear interpolation and Taylor-based sub-pixel interpolation to an improved taylor pixel interpolation

12、algorithm. The algorithm has certain robustness for noise and slight variations in illumination. Copper cap (with needle) is a commonly electronic component used in the semiconductor electronics industry. It’s machining

13、accuracy and quality are directly related to the final electronic product life and power consumption This paper led copper cap (with needle) measurement and defect detection project as an example, apply the improved tayl

14、or sub-pixel interpolation algorithm to cap’s inner diameter measurement and cone size measurement. Compared with the traditional bilinear Interpolation algorithm, the precision and stability have improved significantly.