即時影像處理應用於電阻色碼辨識

Abstract

本實驗發想，是因為顏色辨別障礙人士在分辨細小且多種顏色之電阻色碼時，需耗費大量時間或求助於人，因此發想了顏色影像辨識暨自動分類收納之系統。 我們使用ESP32-CAM進行本專題之辨識拍攝，並透過無線網路與伺服器搭配進行照片、辨識出的數值進行資料傳輸，減少線材在電腦佔用太多序列阜。透過修改鏡頭參數使成像對比達到電腦辨識所需要求，並使用高倍凸透鏡縮短焦距，採用HSV顏色理論來對每個色彩進行定義，並加入侵蝕膨脹等顏色處理，減少誤差。 經實驗後，發現與電阻對比較鮮明、被環境光影響小者，較容易辨識成功；中性色(如白色、灰色)較容易被環境影響(如光線、轉盤之顏色等)，目前辨識成功率約六成，若可以加入更多色彩模組增加穩定性，則有更高的辨識正確率與應用機會。本實驗之鏡頭辨識與市面之影像辨識(如:車牌辨識)相比，有更高需求於鏡頭之焦距、顏色對比、形狀不一定、方向性、順序性等。若未來可開發完全，可幫助更多顏色辨別障礙患者快速、方便辨識電阻色碼。

The idea for this experiment was that people with color discrimination disabilities need to spend a lot of time or seek help from others when distinguishing small and multi-colored electronic color code. Therefore, a color image recognition and automatic classification and storage system was invented. We use ESP32-CAM to perform identification and shooting for the topic, and use the wireless network to cooperate with the server to transmit the photos and the identified values, so as to reduce the number of cables occupying too many serial ports in the computer. By modifying the lens parameters, the imaging contrast meets the requirements of computer recognition, and using a high-power convex lens to shorten the focal length, the HSV color theory is used to define each color, and color processing such as erosion and expansion is added to reduce errors. After experiments, it was found that those with a clear contrast to the resistor and less affected by ambient light are easier to identify successfully; neutral colors (such as white and gray) are more easily affected by the environment (such as light, the color of the turntable, etc.), and the current identification is successful The recognition rate is about 60%. If more color modules can be added to increase stability, there will be higher recognition accuracy and application opportunities. Compared with the image recognition in the market (such as license plate recognition), the lens recognition in this experiment has higher requirements on the focal length, color contrast, shape uncertainty, directionality, sequence, etc. of the lens. If it can be fully developed in the future, it can help more patients with color discrimination disorders to quickly and conveniently identify an electronic color code.