Research On Privacy Protection Method Of In-home Video Surveillance Based On Bionic Vision-intelligent Application Binary Model

As the whole society strides towards intellectualization,people’s lives have changed greatly.It is undeniable that the booming emerging technologies such as big data and artificial intelligence have greatly improved the quality of life and work efficiency.These intelligent algorithms are based on the massive data samples in the era of big data,which makes everyone in them become the object of sampling.At the same time,with the popularization of education and the general improvement of social quality,people have a stronger sense of self,which leads to the emergence of the concept of personal privacy.The problem of privacy disclosure caused by scientific and technological progress will lead to people’s questioning and even exclusion of the intelligent process.In this regard,this paper focuses on the intelligent monitoring system with the possibility of visual privacy disclosure in the field of computer vision.With the development of deep learning and other technologies that can be applied to image analysis,the practicability of intelligent monitoring system is gradually reflected.More and more places have set up surveillance cameras,which also leads to the possibility that more personal behavior habits,life trajectories and other data can be collected and used.At present,the performance of mainstream intelligent algorithms is generally inseparable from the premise guarantee of video quality,and high-quality video images carry a large amount of private information at the visual level,which makes personal privacy at the risk of malicious theft and disclosure in the intelligent monitoring system.Therefore,the video monitoring system is greatly limited in application scenarios,but in private scenes such as home,the intelligent monitoring system is not without security value.For example,the realization of health monitoring for the elderly or risk early warning for children is its future development direction.As for the video surveillance for intelligent applications mentioned above,the collection of massive data will increase the risk of personal privacy disclosure.In this paper,a new concept of privacy protection based on bionic insect vision is proposed to realize information shielding at human visual level by simulating insect vision with low visual level.Secondly,by integrating visual and statistical features,a set of evaluation model for bionic visual privacy protection(VPP)images is proposed to evaluate the degree of visual privacy protection.Finally,a set of correlation model between bionic VPP and intelligent application is given through quantitative statistics to further standardize the application scope of bionic VPP.The specific contents are mainly from the following aspects:(1)In order to protect the privacy of image and video data,the traditional way is to block and cover sensitive information mainly through image encryption technology,but these methods still have various problems in the whole process of intelligent monitoring.Therefore,the research direction of this paper turns to low-vision animals in nature,and uses the visual difference between animals and humans to achieve the privacy protection effect to the extent of human eyes.In terms of species selection of bionic vision,this paper mainly adopts insects with unique compound eye structure to try to ensure sufficient retention of cognitive information at a level far lower than human vision;(2)For the image processed by bionic insect vision,an evaluation system is still needed to judge whether it has achieved VPP and the corresponding degree of VPP.Therefore,this paper refers to the overall idea of image quality evaluation and visual security evaluation.By extracting the features of Generalized Center-Symmetric Local Binary Pattern(GCS-LBP)as the visual features,and Benford law-fractal dimension as the statistical feature of bionic VPP image.Finally,a set of evaluation model for VPP degree was proposed through the fusion of visual and statistical features;(3)When bionic vision meets the privacy protection security in intelligent applications,it still needs to ensure its practical application value,that is,to ensure the intelligent application recognition effect of bionic VPP video.Therefore,this paper takes human pose recognition,which has high common value in home video surveillance,as an example,and builds a set of binary correlation statistical model combining bionic vision and intelligent application through experiments.Experimental results on open data sets show that the correlation statistical model established by this method can obtain an effective range of bionic VPP.