Study On Low Illuminated Target Recognition Method Based On Image Enhancement

With the continuous innovation and development of smart cities and artificial intelligence,advanced science and technology are applied in various fields.Security work is increasingly demanding for the intelligentization of security systems in areas where roads,communities,schools,shopping malls,etc.are densely populated and complicated.The security system should have the ability to effectively monitor the large-scale areas and accident-prone areas around the clock.The monitoring and image processing identification in the normal daylight environment has become more mature,but the image processing recognition for the nighttime,early morning and other accidents The problem has yet to be optimized.Therefore,in the low illumination environment,the problem that the image recognition method of monitoring image recognition accuracy is not high,the image enhancement optimization cost is large,and the adaptive brightness cannot be adaptively changed has become a bottleneck of practical application,and need to optimized.This study intends to combine multi-disciplinary theories and methods such as optical illumination,color science,and image and video processing.By studying the illumination environment,the impact on image quality and target recognition accuracy is significant,especially in low illumination conditions.Aiming at the influence of many factors such as image background,brightness and image noise,a model of low-illuminance target recognition algorithm based on image enhancement is developed.The low-illumination image test and acquisition experimental scenes in line with the national test environment standards were designed and built,and the test samples were collected based on the specifications.Starting from the relationship between the sample collection environment and the Mask R-CNN target recognition model,the characteristics of the captured images in low illumination environment are studied.The international standard acquisition method is used to collect small sample data sets.The relationship between the imaging environment parameters for the image with low system recognition rate,the environmental parameters with reduced target recognition accuracy,and the image enhancement based on the camera response model for the images captured under low illumination to achieve the improvement Robustness of target recognition algorithm model,recognition accuracy,target recognition model applicable environmental scope and target of reducing target recognition cost.