Wenhai Xu

Antivibration pipeline-filtering algorithm for maritime small target detection

Abstract. When searching for small targets at sea based on an infrared imaging system, irregular and random vibration of the airborne imaging platform causes intense interference for the pipeline-filtering, which is an algorithm that performs well in detecting small targets but is particularly sensitive to interframe vibrations of sequence images. This paper puts forward a pipeline-filtering algorithm that has a good performance on self-adaptive antivibration. In the block matching method that combines the normalized cross-correlations coefficient with the normalized mutual information, the interframe vibration of sequence images is acquired in real time and used to correct coordinates of the single-frame detection results, and then the corrected detection results are used to complete the pipeline-filtering. In addition, under severe sea conditions, small targets at sea may disappear transiently, leading to missing detection. This algorithm is also able to resolve this problem. Experimental results show that the algorithm can overcome the problem of interframe vibration of sequence images, thus realizing accurate detection of small maritime targets.

Robust Infrared Maritime Target Detection Based on Visual Attention and Spatiotemporal Filtering

It has always been a great challenge to efficiently detect small infrared targets from complex image backgrounds without any prior knowledge. This is especially true when both strong and weak targets appear in the same image or when the weak targets come up on image borders. The main contribution of this paper is to design a robust infrared maritime target detection method, in which a visual attention and pipeline-filtering model is proposed by integrating a revised visual attention model (VAM) and the antivibration pipeline-filtering algorithm. The revised VAM, a single-frame target detection strategy, will first compute a saliency map (SM) from a specific modality, which is automatically selected according to image background smoothness. Then, an automatic strategy for extracting suspected targets from an SM is also proposed here, which highlights targets and suppresses background clutters in SMs through local saliency singularity evaluation. Moreover, contrary to the original VAM, we adopt border saliency preservation in center-surround difference so that robust detection can be guaranteed for targets on image borders. Finally, to eliminate the interference of sea glints and confirm real targets, we adopt the antivibration pipeline-filtering algorithm, a multiframe-based clutter removal method. Compared with the original VAM and two other existing target detection algorithms, experimental results have proven that our strategy can detect infrared maritime targets much better under different environmental conditions. This research can significantly improve the success rate and efficiency of searching maritime targets in different weathers using infrared imager, especially in heavy sea fog and strong ocean waves.

Color single-pixel imaging based on multiple measurement vectors model

Abstract. Single-pixel imaging (SPI) represents a promising approach to multispectral imaging. We investigated the interband similarity of color images among red, green, and blue bands and found that it is highly possible for their wavelet coefficients to be at the same locations due to edges and transactions. Accordingly, we constructed a multiple measurement vectors model that includes a constraint under which the sparse coefficients of different bands have the same sparse structure, and then joint reconstruction is performed for all bands. We ran both simulated and actual experiments to validate the feasibility and effectiveness of the proposed method and found that compared with similar methods, it significantly improves the reconstruction quality of color SPI.

Texture orientation-based algorithm for detecting infrared maritime targets.

Infrared maritime target detection is a key technology for maritime target searching systems. However, in infrared maritime images (IMIs) taken under complicated sea conditions, background clutters, such as ocean waves, clouds or sea fog, usually have high intensity that can easily overwhelm the brightness of real targets, which is difficult for traditional target detection algorithms to deal with. To mitigate this problem, this paper proposes a novel target detection algorithm based on texture orientation. This algorithm first extracts suspected targets by analyzing the intersubband correlation between horizontal and vertical wavelet subbands of the original IMI on the first scale. Then the self-adaptive wavelet threshold denoising and local singularity analysis of the original IMI is combined to remove false alarms further. Experiments show that compared with traditional algorithms, this algorithm can suppress background clutter much better and realize better single-frame detection for infrared maritime targets. Besides, in order to guarantee accurate target extraction further, the pipeline-filtering algorithm is adopted to eliminate residual false alarms. The high practical value and applicability of this proposed strategy is backed strongly by experimental data acquired under different environmental conditions.

The UAV verification experiment of on-orbit dynamic scene matching

In order to improve the imaging quality of remote sensing camera, this paper proposes a working pattern of metering, calculating and imaging. In order to verify its effectiveness on-orbit, this paper designs a dynamic scene matching verification system based on unmanned aerial vehicle and the flight experiments were carried out. The verification system includes metering camera, imaging cameras, acquisition processing, ground monitoring control and other parts. There is a certain preset angle between the metering camera and the imaging camera, before the imaging camera is push-broom imaging a certain area, the image of this area can be obtained by metering camera in advance, and the highest luminance information of this area can be measured. Then, Exposal parameters of the imaging camera can be calculated by setting the highest luminance information measured before as the saturation luminance of the imaging camera. Finally, when the imaging camera focuses on this area after orbiting to a certain extent, the imaging camera could realize dynamic scene matching by adopting exposure parameters calculated previously. The flight experiments in Dalian obtained the high-resolution image with a resolution of 0.01m. Experiment results show that, the realization of matching method is simple and can achieve in 1s. The proposed method can obtained well exposure images and the image details are rich. This experiment provides support for the practical application on-orbit.

A self-adaptive anti-vibration pipeline-filtering algorithm

The mobile pipeline-filtering algorithm is a real-time algorithm that performs well in detecting small dim targets, but it is particularly sensitive to interframe vibration of sequence images. When searching for small dim targets at sea based on an infrared imaging system, irregular and random vibration of the airborne imaging platform causes huge interference problems for the mobile pipeline-filtering. This paper puts forward a pipeline-filtering algorithm that has a good performance on self-adaptive anti-vibration. In the block matching method using the normalized cross-correlations coefficient (NCC), the interframe vibration of sequence images is acquired in real time and used to correct coordinates of the single-frame detection results, and then the corrected detection results are used to complete the mobile pipelinefiltering. Experimental results show that the algorithm can overcome the problem of interframe vibration of sequence images, thus realizing accurate detection of small dim maritime targets.

A small dim infrared maritime target detection algorithm based on local peak detection and pipeline-filtering

In order to realize accurate detection for small dim infrared maritime target, this paper proposes a target detection algorithm based on local peak detection and pipeline-filtering. This method firstly extracts some suspected targets through local peak detection and removes most of non-target peaks with self-adaptive threshold process. And then pipeline-filtering is used to eliminate residual interferences so that only real target can be retained. The experiment results prove that this method has high performance on target detection, and its missing alarm rate and false alarm rate can basically meet practical requirements.