智慧型數位影像監控開發平台之實作與應用

Abstract

在目前的與影像相關的應用中，如大樓監視系統、車用電子、死角監控、危險區域管制等等方面，大多系統仍然需要人類的監控作為依據，如加上了機器視覺的功能，仍然有許多的影像演算法無法達到Real-Time。由於監視器畫素許多皆為D1或VGA畫質，無法有效的得到有效資訊；如果使用高解析之攝影機，更多的演算法無法實現於系統中。 本論文針對這類的應用提出一個較為平衡的概念，使用部分的硬體影像演算法作為影像「前處理」、「後處理」等等的運算，這些「前處理」、「後處理」在影像處理的概念中有許多相同的部分，這些項目由硬體快速的完成處理，並不用耗損過多的微處理器的負載，當微處理器的負載越低，所能加入的應用就可以越高，並且透過適當的修改完成Real-Time的需求。 在韌體框架中，修改原廠提供的韌體建立出符合影像處理角度的應用框架，使用簡單的指令就可以完成轉編譯的動作，並且新增了些易於開發者使用的子程式，以利便於後續開發的應用，使得開發的速度可以加快。 在系統的軟體應用中，本文提出了一個廣泛型的應用層面，包含背景的建立、物體的追蹤、物體特徵點的選取、判斷警報的條件等等的層面，皆可以做更深入的探討，在此論文中僅做出一個目標性的應用，可由開發者自行往更深入的演算法探討。

此些動作可以使系統軟硬體共同運算加快影像演算法的效率，並可以使開發者專注於影像演算法的構思，避免周邊設施的複雜影響了開發效率，也同時使系統共同運算邁進了一步，未來更可以使用於許多與影像處理有關的設備。

In the current video-related applications , such as building surveillance systems, automotive electronics, corner monitors, hazardous area. Most system still require human monitoring as basis, if the machine has vision functions, there are still many image algorithms can not achieve Real-Time. Because most of monitor pixels are still D1 or VGA quality, it’s unable to obtain useful information effectively. and if high resolution of the camera is used, more algorithms can not be achieved in the system.

This thesis proposed a more balanced concept for these kind of applications, used parts of the hardware algorithm to calculate "pre-treatment," and "post processing" of images. There are many same concept in "pre-treatment," and "post processing" of the image processing, these projects complete by fast hardware processing which doesn’t waste too much loading of microprocessor. When the loading of microprocessor is lower, the joined the application could be higher, and reach Real-Time through appropriate modifications.

In the framework of firmware, modifying the original firmware to establish applications of image processing, using simple instructions to cross-compile. Adding easier programs to use the function, following-up to facilitate an easy application development, making development speed up.

In the application of system, this thesis proposed a widespread application level, including establish the background, object tracking, object feature points selected to determine the level of alarm conditions, etc., All could do more in-depth discussion, this thesis only hypothesize a goal of applications, other developers do more in-depth study of algorithms by themselves.

These applications could common compute the hardware and software of the system which enhance the efficiency of image algorisms, and make designer more concentrate on the idea of image algorithms, avoid the complex surrounding facilities environment effect developing efficiency. These applications make co-operation system has a step forward, and it can apply to equipments which related image processing in future.