軟體擴充通常依賴於原始程式碼的存取權限或是預先設計的擴充框架。然而，在無任何官方擴充手段的情況下，本實驗室開發的 Venom Programming（舊稱 Parasite Programming）技術已經能夠透過 SDK 取得應用程式的 UI 資訊，並利用透明視窗達成軟體擴充的效果。然而，現代應用程式往往包含影片播放、圖片、以及其他動態的渲染機制，導致無法單純使用抓取UI的方式進行分析。為了充分發揮 Venom Programming 在這些複雜應用場景中的潛力，強化其動態內容處理能力變得至關重要。 本論文透過整合電腦視覺技術，大幅強化了 Venom Programming 的處理能力。在原有架構基礎上新增的跨語言電腦視覺後端，結合了 OpenCV 影像處理技術與 ONNX 模型推論能力，並採用工作流機制讓開發者能夠靈活組合各種視覺分析功能。透過這個創新架構，開發者能夠在不依賴額外工具的情況下，從目標應用程式中取得動態內容的資訊，並實時追蹤其變化狀態。這項技術突破顯著擴大了 Venom Programming 的應用範圍，使其能夠處理更加複雜的動態內容擴充需求。 ;Software extension usually relies on access to source code or pre-designed extension frameworks. However, our laboratory has developed Venom Programming (formerly Parasite Programming) technology without any official extension methods. This technology can obtain application UI information through SDK and achieve software extension using transparent window technology. However, modern applications often contain video playback, images, and other dynamic rendering mechanisms. UI extraction methods alone cannot perform analysis on these contents. To make full use of Venom Programming in these complex application scenarios, we need to strengthen its dynamic content processing capabilities. This thesis enhances Venom Programming′s processing capabilities by integrating computer vision technologies. We added a cross-language computer vision backend to the original architecture. This backend combines OpenCV image processing technology with ONNX model inference capabilities. It uses a workflow mechanism. Developers can combine various visual analysis functions through this mechanism. Developers can obtain dynamic content information from target applications through this architecture. They can track changes in real-time without additional tools. This enhancement expands the application scope of Venom Programming. It enables the technology to handle more complex dynamic content extension requirements.
