Multi-class target recognition based on adaptive feature selection

Flying target detection and recognition by feature fusion

“…(9) to (10) it is obvious that jS 0 t ðrÞj ¼ N t ðrÞ ¼ N t−r þ : : : þ N t þ : : : þ N tþr : (11) Figure 4 demonstrates this process using one frame from a video with two objects (i.e., K t ¼ 2), where the inner rectangles (darker color) represent the b 1 t and b 2 t bounding boxes enlarged in both directions with 20%. (9) to (10) it is obvious that jS 0 t ðrÞj ¼ N t ðrÞ ¼ N t−r þ : : : þ N t þ : : : þ N tþr : (11) Figure 4 demonstrates this process using one frame from a video with two objects (i.e., K t ¼ 2), where the inner rectangles (darker color) represent the b 1 t and b 2 t bounding boxes enlarged in both directions with 20%.…”

Visual real-time detection, recognition and tracking of ground and airborne targets

“…where d pos is the Euclidean distance of positions, ϑ is a weighting constant, O i is previous frame object i, D j is the detected object and d hist is the histogram distance: d pos(Oi,Dj = (p i,x − p j,x ) 2 + (p i,y − p j,y ) 2 (13) wherep is the predicted position of object O.…”

“…Here we focus on feature based recognition that has the goal of extracting and comparing object features for target recognition purposes. In one work aerial images (taken from airborne platforms or satellites) are the basis of ground target recognition 13 where specific feature sets are pre-calculated that are sensitive to certain object parameters and can help in detection. Such features include geometric descriptors (e.g.…”

“…Weng et al present a flying target detection method 8 based on infrared processing which is robust against weather conditions, but is not able to recognize the flying targets. Other works 12,13 present more recent approaches to target tracking and recognition, based on infrared sensing for detection and adaptive feature selection for recognition. A robust infrared based approach 12 is presented for target detection and tracking, although it is somewhat constrained since it required static cameras.…”

“…A robust infrared based approach 12 is presented for target detection and tracking, although it is somewhat constrained since it required static cameras. Wang et al 13 present target recognition on aerial imagery by a multi-feature method, sensitive to various geometrical shapes (circles, lines, etc.) of ground targets.…”