本論文對於不同應用面的照片或影片，提出了三種物件辨識解決方案，分別如下：跟骨骨折的分割和分類、使用無人機 (UAV)對運動物體的檢測與識別及使用RGB-D攝影機的人數統計應用。在第一項研究中提出以電腦輔助法檢測跟骨骨折以獲取更快、更詳細的檢測結果。首先選擇輸入圖像中骨的解剖平面方向以確定跟骨的位置，然後偵測跟骨圖像的幾個碎片並使用顏色分割進行標記。Sanders系統可以在橫向和冠狀影像中根據碎片的數量將骨折分為四種類型，在矢狀影像中，根據骨折區域的參與程度可將骨折分為三種類型。在第二項研究中利用無人機擷取的影像設計出了一種新穎且高效的技術用於檢測和識別移動物體。首先找到兩個連續幀之間的特徵點用來估計攝影機的運動以穩定圖像序列，然後將物體的感興趣區域（ROI）檢測為移動物體候選（前景）。除此之外，根據前景和背景中的最大運動向量對靜態物體和動態物體進行分類。在第三項研究中提出了一種使用RGB-D攝像機進行實時人數計數的新技術。 首先，提出了圖像校準以獲得深度和RGB圖像之間的比率和偏移值。 在深度圖像中，通過去除背景將人物檢測為前景。 然後，根據檢測到的人的位置註冊他們的關注區域（ROI），並將其映射到RGB圖像。 基於具有通道和空間可靠性的判別相關濾波器，在RGB圖像中跟踪註冊人員。 最終，當人們越過興趣線（LOI）且位移距離超過2米時，就算在內 。;This dissertation proposes some practical solutions for object identification based on different image acquisitions such as segmentation and classification of calcaneal fractures, detection and recognition of moving objects using an Unmanned Aerial Vehicle (UAV), and people counting applications using an RGB-D camera. In the first study, a computer-aid method for calcaneal fracture detection to acquire a faster and more detailed observation is proposed. First, the anatomical plane orientation of the tarsal bone in the input image is selected to determine the location of the calcaneus. Then, several fragments of the calcaneus image are detected and marked by color segmentation. The Sanders system is used to classify fractures in transverse and coronal images into four types based on the number of fragments. In the sagittal image, fractures are classified into three types based on the involvement of the fracture area. In the second study, a new and efficient technique is proposed for the detection and recognition of moving objects in a sequence of images captured from a UAV. First, the feature points between two successive frames are found for estimating the camera movement to stabilize the sequence of images. Then, the region of interest (ROI) of the objects are detected as the moving object candidate (foreground). Furthermore, static and dynamic objects are classified based on the most motion vectors that occur in the foreground and background. In the third study, a new technique for real-time people counting using an RGB-D camera is proposed. First, image calibration is proposed to obtain the ratio and shift values between the depth and the RGB image. In the depth image, people are detected as the foreground by removing the background. Then, the region of interest (ROI) of the detected people is registered based on their locations and is mapped to the RGB image. The registered people are tracked in the RGB image based on the discriminative correlation filter with the channel and spatial reliabilities. Finally, people are counted when they cross the line of interest (LOI) and the displacement distance is more than 2 meters.
