Evaluation of Camouflage Effectiveness Model Based on Disruptive Coloration and Background Guided Fusion

Optical camouflage technology, which is widely used in the military field, can significantly reduce the detectability and detection probability of high-value targets, and improve the survivability of targets. Evaluating camouflage's effectiveness is one of the critical points of current research, which directly affects the design and selection of camouflage strategy. Traditional evaluation models mainly focus on the similarity of color, brightness, texture between target and background based on background matching strategy. However, the traditional models fail to consider the target's surface's integrity and its edges' coherence, so they cannot accurately evaluate the camouflage effect of the target in the complex and changeable field environment. In this paper, the evaluation of camouflage effectiveness is regarded as a confrontation between visual search and target camouflage strategy. The disruptive coloration feature is introduced into the traditional background matching feature, and a representation model of target surface damage degree based on multifractal theory is proposed to evaluate the target camouflage effect comprehensively. The experimental results demonstrate that the target camouflage effect is in accordance with the probability of detection and mean search time in the Search_2 dataset and CamData dataset. Search_2 dataset is widely used as a public dataset in the field of camouflage effect evaluation. Camdata dataset is a self built dataset to verify the performance of the proposed model. We expect this model to be used to evaluate and compare the camouflage effect of different patterns and can be applied to optimize pattern design. The proposed model is divided into feature registration and feature fusion. In the feature registration stage, the ability of visual system to gather scene information is simulated. Based on the classical background matching features such as color similarity, texture similarity, and structure similarity, the target surface disruption and edge disruption are introduced to represent the disruptive coloration color features. Multifractal theory and Gabor filter are used to measure target edge and surface disruption. In the feature fusion stage, the feature congestion index is introduced to calculate the background complexity. The adaptive adjustment of feature fusion weight is guided by background complexity through Logistic Equation to evaluate target camouflage effectiveness in different changing scenes effectively. The overall framework of the camouflage effectiveness evaluation model is shown. The performances of our method and several existing methods are evaluated by the Pearson Correlation Coefficient (PLCC), Spearman Correlation Coefficient (SRCC) and Root Mean Square Error (RMSE). In order to verify the robustness of the model, experiments are conducted both on the classical dataset (Searach_2) and the self built dataset (CamData); the proposed model outperformed the related evaluation algorithms. Among the competing methods, the PCDGM and UIQI methods are based on background matching strategy, and the GabRat method is based on disruptive coloration features. The experimental results show that the PLCC, SRCC and RMSE of detection probability are 0.888, 0.773 and 0.054 on Search_2 dataset, and 0.835, 0.805 and 0.126 on CamData dataset. The results of target search time are 0.871, 0.775 and 2.340 on Search_2 dataset, and 0.834, 0.760 and 0.727 on CamData dataset. The model in this paper uses scene complexity to guide background matching features and disruptive coloration features to carry out adaptive fusion, which is more consistent with the visual search mechanism and can accurately reflect the changes in target camouflage state and background environment. It is suitable for the evaluation of high reliability camouflage effectiveness of targets in different scenes. The proposed model, from the visual search mechanism against the target camouflage strategy perspective, combines the features of matching disruptive coloration and background matching through background guidance and overcomes the traditional model of the shortcoming of tow reliability in a complex environment. Consequently, the comprehensive assessment of the camouflage effect of different targets in a complex environment is realized. Experiments on the Searach_2 and CamData datasets confirmed the superior performance of the proposed method over other relevant algorithms. Theoretical analysis and experiment prove that the method presented in this paper has good subjective consistency in two scenes of different complexity, and has high operability and stability, which has specific application and promotion value. Besides, our study shows that the camouflage effect can be evaluated more accurately by using background characteristics to guide the fusion of background matching and disruptive coloration features, which is helpful for people to understand the mechanism of camouflage further.