






















摘" 要:該文針對紅外弱小目標跟蹤場景中未知新生分布先驗信息且傳統MeMBer會造成勢低估的問題,提出一種基于平方根容積卡爾曼濾波勢均衡多目標多伯努利的新生目標自動搜索算法并給出高斯混合實現。該算法首先將平方根容積卡爾曼濾波(SCK)算法引入到紅外圖像的多弱小目標檢測跟蹤中,用來實現CBMeMBer-TBD算法,同時結合新生目標自動搜索算法,自適應地產生新生弱小目標的分布,而勢分布則通過CBMeMBer來實現。仿真實驗表明,所提算法能夠實現在新生目標未知場景下的紅外弱小目標跟蹤,且其跟蹤精度與傳統的MeMBer跟蹤算法相比,有明顯提高。
關鍵詞:紅外弱小目標;平方根容積卡爾曼濾波;勢均衡多目標多伯努利濾波;自動搜索;目標跟蹤算法
中圖分類號:TN713" " " 文獻標志碼:A" " " " " 文章編號:2095-2945(2023)28-0024-04
Abstract: Aiming at the problem that the prior information of new distribution is unknown and the traditional MeMBer can cause potential underestimation in infrared dim and small target tracking scene, a multi-target multi-Bernoulli automatic searching algorithm based on square root cubature Kalman filter potential equalization is proposed and the Gaussian mixture implementation is given. Firstly, the square-root cubature Kalman filter (SCKF) algorithm is introduced into the infrared image multiple dim and small target detection and tracking, which is used to realize the CBMeMBer-TBD algorithm. At the same time, combined with the new target automatic search algorithm, the distribution of new dim and small targets is generated adaptively, and the potential distribution is realized by CBMeMBer. Simulation results show that the proposed algorithm can achieve infrared dim and small target tracking in unknown scenes of new targets, and its tracking accuracy is obviously improved compared with the traditional MeMBer tracking algorithm.
Keywords: infrared dim and small target; square root cubature Kalman filter; cardinality balanced multi-target multi-Bernoulli filter; automatic search; target tracking algorithm
近些年,紅外成像技術因其價格低廉、抗干擾、可全天候工作等優點越來越受到人們的關注,該技術可廣泛應用于軍事和民用領域。在多數應用場景中,由于紅外成像目標占整幅紅外圖像的面積非常小,導致被檢測的小目標很容易被淹沒在復雜場景的雜波中(低信噪比),在極端情況下,紅外目標只能是一個亮點。所有這些原因使得紅外小目標的檢測和跟蹤變得非常困難。
為了解決這一具有挑戰性的問題,在過去的幾十年中,人們提出了各種紅外小目標檢測算法。這些算法可以分為兩類:序列檢測方法和單幀檢測方法。序列檢測方法主要利用目標的時間相關性和運動信息對多幀圖像進行聯合檢測。……