The economic and scientific value that small celestial bodies(SCBs)offer humanity is the main motivation for close exploration of these bodies.However,autonomous optical navigation is challenging due to the light vari...The economic and scientific value that small celestial bodies(SCBs)offer humanity is the main motivation for close exploration of these bodies.However,autonomous optical navigation is challenging due to the light variation caused by the rapid spin of SCBs.In this context,we propose a light prior brightness equalization self-calibration method,which can achieve brightness equalization of SCB images under varying illumination conditions while preserving image details,thereby increasing the number of feature-matching points.First,we design a light prior information function based on the illumination variation law of Lambert’s cosine law.Based on the function,the high-light and low-light areas of SCB images are distinguished.Furthermore,we create a brightness equalization mathematical model that maps the illumination components of high-light and low-light areas.Then,based on the brightness equalization mathematical model,we construct a light prior brightness self-calibration network.The proposed network includes 3 main modules:the illumination component estimation module,brightness self-calibration module,and light prior information prediction module;the proposed network utilizes a multistage illumination sharing approach to achieve separation and optimization of illumination components.Finally,the experimental results show that our method can achieve brightness equalization,markedly increasing the number of correct feature matches.展开更多
Timely identification and tracking of abnormal hens in stacked cages are of great significance for precision treatment and the elimination of sick individuals.The head features of the caged-hens are used to overcome o...Timely identification and tracking of abnormal hens in stacked cages are of great significance for precision treatment and the elimination of sick individuals.The head features of the caged-hens are used to overcome observation difficulties caused by the cage and feathers blocking,but it is still hard to identify similar head states.To solve this problem,the fine-grained detection of caged-hens head states was developed using adaptive Brightness Adjustment in combination with Convolutional Neural Networks(FBA-CNN).Grid Region-based CNN(R-CNN),a convolution neural network(CNN),was optimized with the Squeeze-and-Excitation(SE)and Depthwise Over-parameterized Convolutional(DO-Conv)to detect layer heads from cages and to accurately cut them as single-head images.The brightness of each single-head image was adjusted adaptively and classified through the deep convolution neural network based on SE-Resnet50.Finally,we returned to the original image to realize multi-target detection with coordinate mapping.The results showed that the AP@0.5 of layer head detection using the optimized Grid R-CNN was 0.947,the accuracy of classification with SE-Resnet50 was 0.749,the F1 score was 0.637,and the mAP@0.5 of FBA-CNN was 0.846.In summary,this automated method can accurately identify different layer head states in layer cages to provide a basis for follow-up studies of abnormal behavior including dyspnea and cachexia.展开更多
Simulating ambient light adaptability and polarization sensitivity of biological vision is paramount for developing intelligent optoelectronic devices with multi-dimensional perception capabilities.However,achieving b...Simulating ambient light adaptability and polarization sensitivity of biological vision is paramount for developing intelligent optoelectronic devices with multi-dimensional perception capabilities.However,achieving both functionalities in semiconductor devices has historically necessitated complex architectures and high-voltage operation,posing significant challenges for bionic vision systems.Here,we present a light-adaptable and polarization-sensitive bionic vision utilizing a simple yet effective strategy of semiconductor-metal contact engineering in PdSe_(2)transistors.By exploiting the differential coupling strengths at diverse metal-semiconductor interfaces to modulate the dynamics of photogenerated carriers,the device achieves energy-efficient visual adaptive perception across a broad range of lighting conditions,from dim to bright,without the need for additional gate voltage.Furthermore,this transistor enables multi-dimensional perception of visual information through dynamic polarization angle changes and light intensity(dim/bright)detection,providing rich input features for intelligent recognition in complex scenarios.Capitalizing on the intrinsic anisotropy of PdSe_(2)and contact engineering,we have constructed a bionic light-adaptive visual neural network capable of perceiving and recognizing images in complex lighting environments.When enhanced by a residual-generating adversarial network,the system achieves remarkable recognition accuracies of 98%and 97%under dim and bright adaptation conditions,respectively.This research offers a streamlined,versatile,and scalable approach for developing energy-efficient,highly integrated,and multi-dimensional imaging recognition capabilities in light-adaptive and polarization-sensitive bionic vision devices.展开更多
基金support from the National Natural Science Foundation of China(Grant No.U2341214)the Shandong Provincial Natural Science Foundation,China(ZR2023MF006 and ZR2023QF176)the Stable Support Program(HTKJ2024KL502028).
文摘The economic and scientific value that small celestial bodies(SCBs)offer humanity is the main motivation for close exploration of these bodies.However,autonomous optical navigation is challenging due to the light variation caused by the rapid spin of SCBs.In this context,we propose a light prior brightness equalization self-calibration method,which can achieve brightness equalization of SCB images under varying illumination conditions while preserving image details,thereby increasing the number of feature-matching points.First,we design a light prior information function based on the illumination variation law of Lambert’s cosine law.Based on the function,the high-light and low-light areas of SCB images are distinguished.Furthermore,we create a brightness equalization mathematical model that maps the illumination components of high-light and low-light areas.Then,based on the brightness equalization mathematical model,we construct a light prior brightness self-calibration network.The proposed network includes 3 main modules:the illumination component estimation module,brightness self-calibration module,and light prior information prediction module;the proposed network utilizes a multistage illumination sharing approach to achieve separation and optimization of illumination components.Finally,the experimental results show that our method can achieve brightness equalization,markedly increasing the number of correct feature matches.
基金This work was financially supported by the Jiangsu Provincial Key Research and Development Program(Grant No.BE2019382,No.BE2020378).
文摘Timely identification and tracking of abnormal hens in stacked cages are of great significance for precision treatment and the elimination of sick individuals.The head features of the caged-hens are used to overcome observation difficulties caused by the cage and feathers blocking,but it is still hard to identify similar head states.To solve this problem,the fine-grained detection of caged-hens head states was developed using adaptive Brightness Adjustment in combination with Convolutional Neural Networks(FBA-CNN).Grid Region-based CNN(R-CNN),a convolution neural network(CNN),was optimized with the Squeeze-and-Excitation(SE)and Depthwise Over-parameterized Convolutional(DO-Conv)to detect layer heads from cages and to accurately cut them as single-head images.The brightness of each single-head image was adjusted adaptively and classified through the deep convolution neural network based on SE-Resnet50.Finally,we returned to the original image to realize multi-target detection with coordinate mapping.The results showed that the AP@0.5 of layer head detection using the optimized Grid R-CNN was 0.947,the accuracy of classification with SE-Resnet50 was 0.749,the F1 score was 0.637,and the mAP@0.5 of FBA-CNN was 0.846.In summary,this automated method can accurately identify different layer head states in layer cages to provide a basis for follow-up studies of abnormal behavior including dyspnea and cachexia.
基金grateful to the National Key Research and Development Program of China(No.2024YFA1211400)the National Natural Science Foundation of China(Nos.U22A20138,52302162,and 52202183)the Guangdong Basic and Applied Basic Research Foundation(No.2023B1515120041).
文摘Simulating ambient light adaptability and polarization sensitivity of biological vision is paramount for developing intelligent optoelectronic devices with multi-dimensional perception capabilities.However,achieving both functionalities in semiconductor devices has historically necessitated complex architectures and high-voltage operation,posing significant challenges for bionic vision systems.Here,we present a light-adaptable and polarization-sensitive bionic vision utilizing a simple yet effective strategy of semiconductor-metal contact engineering in PdSe_(2)transistors.By exploiting the differential coupling strengths at diverse metal-semiconductor interfaces to modulate the dynamics of photogenerated carriers,the device achieves energy-efficient visual adaptive perception across a broad range of lighting conditions,from dim to bright,without the need for additional gate voltage.Furthermore,this transistor enables multi-dimensional perception of visual information through dynamic polarization angle changes and light intensity(dim/bright)detection,providing rich input features for intelligent recognition in complex scenarios.Capitalizing on the intrinsic anisotropy of PdSe_(2)and contact engineering,we have constructed a bionic light-adaptive visual neural network capable of perceiving and recognizing images in complex lighting environments.When enhanced by a residual-generating adversarial network,the system achieves remarkable recognition accuracies of 98%and 97%under dim and bright adaptation conditions,respectively.This research offers a streamlined,versatile,and scalable approach for developing energy-efficient,highly integrated,and multi-dimensional imaging recognition capabilities in light-adaptive and polarization-sensitive bionic vision devices.
