Pedestrian re-identification is a crucial research task in the fields of computer vision and video surveillance. The primary challenges include occlusion, illumination variation, and complex backgrounds, which significantly undermine the robustness and generalization capabilities of existing methods. To address these issues, a pedestrian re-identification method that effectively integrates multi-scale features and enhances the attention mechanism is required. This paper proposes an occluded pedestrian re-identification method based on multi-scale feature fusion. The method introduces an AAC (Add Noise and Concatenate) module, which injects noise into the central region of the input image to enhance the model's robustness and improve its generalization ability. The model employs EfficientNetB0 and DaViT_small as backbone networks. EfficientNetB0 processes the original input image, while DaViT_small handles the concatenated image with noise, incorporating a transposed convolution module for upsampling the feature maps to further extract high-level features and enhance spatial resolution. Additionally, a feature blocking and global fusion module is proposed, which splits the feature maps into multiple resolutions and uses different convolutional layers to further extract and fuse global features, ultimately generating a global feature vector. This design ensures the model can extract a rich variety of features from the images, thereby significantly improving the accuracy and reliability of the pedestrian re-identification task. Experimental results demonstrate that the proposed Occluded Pedestrian Re-identification method based on Multi-Scale Feature Fusion (OPR-MSFF) outperforms RNFPR (Relation Network for Person Re-Identification) with improvements of 0.7%, 0.8%, and 1.3% in rank-1 accuracy and increases of 0.9%, 1.2%, and 1.7% in mAP on the Market1501, DukeMTMC-reID, and Occluded-DukeMTMC datasets, respectively. These results validate the effectiveness of the proposed method in enhancing pedestrian re-identification performance.

Traffic sign detection plays an essential role in the technology of self-driving vehicles. Recently, deep learning methods have significantly advanced the field of traffic sign recognition. Nevertheless, faced with increasingly complex traffic scenarios, practical applications of traffic sign detection still encounter challenges, including false detections, missed detections, and reduced accuracy. To tackle these challenges, we introduce an enhanced algorithm for traffic sign detection built on the YOLOv8 model, aimed at improving performance and accuracy. Firstly, a Multi-Scale Convolutional Attention (MSCA) module is embedded into the backbone architecture to improve the model's feature extraction capabilities at multiple scales, enhancing its focus on target areas. Furthermore, a small object detection layer is added during the detection phase, effectively reducing the false positive and missed detection rates for small objects. Finally, we present the Inner-WIoU loss function for bounding boxes, which integrates a dynamic non-monotonic focusing mechanism with auxiliary boxes. This boosts the model's capability to identify objects and enhances overall detection performance. The findings from the experiments demonstrate that the enhanced algorithm obtains an mAP0.5 value of 83.8% on the TT100K dataset, indicating a 7.8% increase compared to the baseline YOLOv8 algorithm. When compared to existing algorithms, the proposed method demonstrates competitive performance.

Significant progress has been made in salient object detection (SOD) based on deep neural networks. However, existing SOD methods are primarily designed for low-resolution image inputs, which often suffer from issues such as sampling depth, receptive field, and model performance when applied to high-resolution image inputs. To tackle this issue, this study introduces a lightweight feature extraction model by replacing the original ResNet network structure with a RepVGG-based model that incorporates the Efficient Channel Attention (ECA) module for lightweight purposes. In order to enhance both the model's accuracy and processing speed, we introduce the Effective Squeeze-and-Excitation (ESE) module for feature fusion. To tackle the issue of unclear boundaries of salient objects, we fuse the Weighted Binary Cross-Entropy, Structural Similarity (SSIM), and Shape-aware Loss into a combined loss function, which replaces the conventional cross-entropy loss. Experimental results demonstrate that the enhanced algorithm (RepPGNet) achieves a 2.8% increase in accuracy compared to the original algorithm, with reduced model parameters and improved clarity of salient object boundaries. The proposed algorithm is also shown to have improved speed and is suitable for high-resolution image scenes.

This paper describes two new interim connection spaces (ICSs), P-2OC and P-3OC, for spectral image compression and reconstruction. For this type of ICS, the weighting table or compression matrix H is modelled as a variable. The associated reconstruction matrix N is chosen as the Weiner estimation matrix. The objective function f(H) is the combination of the averages of the CIEDE2000 color difference (Delta E-00) and the root mean square error (RMSE) between the original and reconstructed reflectance values. Hence, the compression matrix H is determined by solving the nonlinear minimization problem based on the Munsell training dataset. The proposed ICSs, P-2OC and P-3OC, were tested and compared, respectively, with ICS-2SI and ICS-3SI developed by Zhang et al. (JOSAA, Vol. 29, pp. 1027-1034) in 2012 using the NCS dataset, and 2 spectral images. Performance tests showed that the proposed P-2OC and P-3OC ICSs are better than the ICS-2SI and ICS-3SI ICSs, respectively, in terms of RMSE, goodness of fit coefficient (GFC), and Delta E-00 under CIE illuminants D65, A, C and F11. Therefore, it is expected that the P-2OC and P-3OC ICSs can find applications in spectral image compression and cross-media reproduction. (C) 2023 Society for Imaging Science and Technology.

Three new approximately uniform color spaces named MLUV, MLUV1, and MLUV2 were developed by modifying CIELUV lightness, u'v' chromaticity coordinates, and u*v* color coordinates, respectively. Performance tests using the combined and four individual datasets employed at CIEDE2000 development showed that MLUV, MLUV1 and MLUV2 were significantly better than CIELUV. Using values of Standardized Residual Sum of Squares (CIE 217:2016) for predictions of the four individual datasets the ranking (from best to worst) was MLUV, MLUV2 and MLUV1, but for predictions of two ellipses datasets the ranking (from best to worst) was MLUV2, MLUV, and MLUV1. Overall, the MLUV2 space was found to be the best. Hence, it is expected that MLUV2 could be used for color specification and color difference evaluations, especially in industrial applications that depend on additive light mixing, such as color TV sets, video monitors, and lighting.

Pedestrian re-identification aims to retrieve pedestrians across various cameras and scenes. However, the accuracy of re-identification is often affected by factors such as low-quality images of pedestrians and environmental conditions. Consequently, it is crucial for machine learning models to learn features from multiple dimensions. In response to these challenges, this paper proposes a Multi-Dimensional Attention Feature Fusion (MDAFF) method for pedestrian re-identification based on the NFormer approach. This method enables the model to learn and fuse pedestrian features from multiple dimensions, enriching the expressive power of the feature maps and improving the discrimination among pedestrians. By incorporating a PA module into the ResNeXt network for feature extraction, the model enhances its global perception and integrates pedestrian position information into the feature maps. This increases the model's sensitivity to pedestrian positions and reduces the impact of noise on re-identification accuracy. Furthermore, the method extracts channel and spatial correlations from the fused position feature maps and performs feature fusion, facilitating the fusion of multi-dimensional attention features. This alleviates the influence of varying scenarios and poses on re-identification, thereby enhancing the model's performance. Compared to the Res50+NFormer method, which directly models the relationships among different pedestrians after feature extraction, MDAFF integrates multi-dimensional features into the feature maps, improving the model's expressive power and capturing the relationships among different pedestrians more effectively. The proposed MDAFF method achieves a 1.3% increase in mAP and a 1.9% increase in Rank-1 on the Market1501 dataset, as well as a 1.7% increase in mAP and a 0.5% increase in Rank-1 on the DukeMTMC-reID dataset. Therefore, the MDAFF method effectively improves the accuracy of pedestrian re-identification.

The influence of different chromaticity coordinate spaces and different scale factors derived from the tristimulus values L, M, and S of the cone response on the calculation of correlated color temperature (CCT) is studied, and the difference between the calculated CCT and the standard CCT is compared. Tests are carried out on a data set containing 401 spectral power distributions collected by Houser et al. The test results show that the u(c)-v(c), space directly derived from the calculated L, M, and S when the scale factors are all 1 is the best, and the absolute differences of the mean CCT, the median CCT and the maximum CCT are 48, 31 and 851 K, respectively. If the L, M, S are converted to the u(F)-v(F) space derived from the tristimulus value space X(F)Y(F)Z(F) defined by the color matching function to calculate the CCT, the mean CCT difference, the median CCT difference and the maximum CCT difference between the calculated CCT and the standard CCT are 42, 21 and 540 K, respectively.

The development of colorimetry relies on the perceptual phenomena and the experimental data. With the increase of experimental data of visual color difference, people tries to explore and develope new color space, which is as simple as the CIELAB color space, but more uniform. The characteristics of CIELAB color space are analyzed, and a color space with four parameters based on these characteristics is proposed, which is called MLAB color space. The determination of the new color space parameters is a nonlinear constrained optimization problem, which can ensure that the optimized space can well predict the COM-corrected visual color difference dataset collected by the expert group of International Commission on Illumination. Comparison based on COM-Corrected dataset shows that MLAB color space is significantly better than CIELAB color space. Tests based on hue linearity and ellipse datasets show that MLAB color space is also improved to a certain extent compared with CIELAB color space.

A spectrum optimization problem of maximizing the mesopic visual luminous efficiency of phosphor-coated white light-emitting diodes (LED) by exciting green and orange phosphors with a blue LED. The objective function is to maximize the average mesopic visual luminous efficiency at multiple brightness levels. The optimization model for correlated color temperature (CCT) is discrete and exhaustive, and it can only maximize the average mesopic visual luminous efficiency for varied brightness levels under a constant color temperature. A nonlinear constrained optimization model with a simple objective function is suggested in this study to attain the maximum mesopic visual luminous efficiency at varied brightness levels, and the CCT is considered an optimization parameter varying in a specific range. The experimental results show that the model can obtain the optimal LED spectrum, while meeting the requirements of color rendering index, color fidelity, and a given mesopic visual brightness level, and that it truly achieves the purpose of maximizing the medium visual luminous efficiency under the specified brightness level.

Grey scale value inside and outside the contour curve approximates as a constant in CV model. For the target with unever gray scale, the global approximation method can't reflect the change of image gray scale and it is difficult to obtain a satisfactory segmentation results for this kind of image. In this paper the improved CV model is put forward and it overcomes some defects of the original, through the experiment in this paper, the improved algorithm is simpler than LBF algorithm in terms of computational complexity. Better segmentation results can be obtained by using LBF model and ILBF model, but the iterations of ILBF model are less than that of LBF model, and the time consuming of ILBF model is less than that of LBF model.

Here, we report some new observations on the luminescence of carbon dots (CDs) that are passivated with polyethylene glycol, and explored them, for the first time, as highly sensitive H2S detection probe. The as-prepared CDs have an ultraviolet (UV) emission at 350 nm and a green emission at 540 nm when excited at 270 nm, but they have only a green emission at 523 nm when excited at 365 nm. As Na2S is added to the CDs, the UV emission decreases gradually, while the green emission increases slightly, and interestingly a new emission appears at 455 nm is increased linearly in intensity with the increase of Na2S concentration. The blue emission at 455 nm is from the intrinsic core of CDs, and the green emissions at 540 or 523 nm are from their surface states, while the UV emission at 350 nm is from the anchored PEG. Both the emission intensity ratios of F-450/F-350 and F-455/F-523 show an excellent linear relationship with the Na2S concentration in the range of 0-800 mu M, and the detection limit estimated from both of them is approximately 7.0 nM, indicating our method is the most sensitive H2S CDs based detection method reported thus far. Furthermore, the detection was tested successfully for H2S imaging in live cells. We propose that the interactions of H2S with the CDs and the anchored ligands, as well as the energy transfer between the ligands and the CDs, are responsible for the luminescence responses for H2S detection. Our studies enrich the world of carbon dots with plenty of potential applications. (C) 2020 Elsevier Ltd. All rights reserved.

A study by Gao et al found that the Newton method may not converge for certain combinations of the photopic luminance and the ratio of scotopic and photopic luminance for computing the adaptation coefficient m, for the CIE MES2 system for mesopic photometry. Hence, they proposed to use the Bisection-Newton method. In this short note, we propose the Newton method with a new initial guess for computing m. Numerical simulation has shown the proposed method not only converges, but also converges faster than the Bisection-Newton method.

Vehicle detection is the first step and an important part of automatic traffic incident detection systems. It guarantees subsequent vehicle identification and vehicle counting accuracy and has crucial theoretical significance and practical value for traffic safety and control. The model obtained by the original YOLOv4 algorithm is too large to be used in embedded terminals in real time. To overcome this problem, this study replaces the original backbone network of YOLOv4, which is CSPDarknet53, with MobileNetv3 for the feature extraction. To further reduce the number of parameters, deep separable convolution is used to replace the common 3x3 convolution in the original model of the enhanced feature extraction networks SPP and PANet. Because of the imbalance in the object detection data, the loss function is redesigned using a weighting method. The research results show that in comparison to the original YOLO series algorithm, the optimized YOLOv4 algorithm improves the accuracy by 0.53% and reduces the number of model parameters by 78%. In comparison to the other algorithms, the improved YOLOv4 model is smaller and more accurate, which is the basis for realizing intelligent transportation systems.

This paper investigates the effect of the linearization of camera JPG data on the accuracy of spectral reflectance reconstruction based on weighted polynomial regression algorithm and demonstrates whether the JPG data needs to be linearized in the weighted polynomial regression algorithm. This method was trained using the X-Rite Digital ColorChecker Semi Gloss (SG) chart including 110 color and grey patches and tested using the GretagMacbeth ColorChecker chart including 21 color and grey patches, and self-made 11 printed and 18 textile samples. Comparison results based on real camera data have shown that the weighted polynomial regression method with the original JPG data outperformed the weighted polynomial regression method with the linearized JPG data measured in terms of CIEDE2000 color difference and root-mean-square error. Based on the results of this study, linearization of the JPG data does not improve the reconstruction accuracy for reconstructing the reflectance using the weighted polynomial regression method. Without JPG data linearization, higher spectral reconstruction accuracy can also be obtained. The weighted three order polynomial regression method performed the best with original JPG data.

Photoluminescence of Eu3+ in DMSO is intense and ultrasensitive to water, thereby providing a novel method for water detection. Herein, for the first time, we investigated the effects of Eu3+ concentration on luminescence and developed a multiparameter method for trace water detection based on a single luminescence agent. To further extend its practical applications, we explored its performance for water detection in ethanol and gasoline. Our findings demonstrate that it is a sensitive and reliable probe for the detection of a wide concentration range of water in ethanol (0-24.24%) and gasoline (0-32.43%), making Eu-DMSO a promising candidate to detect water in a wide concentration range. These phenomena not only make Eu-DMSO a sensitive agent for in situ water detection in real time but also provide scientifically interesting mechanisms behind its application as a water sensing probe.

Due to the limitation of industrial and civil fields of the polarization camera, we have developed a scheme for synthesizing polarization image based on intensity image and simulation technology. A whole process polarization or intensity transfer model from the incident solar light to the point at which the camera enters the pupil is set up based on 6SV. The indoor simulation test system of the target surface reflectivity is formed. The simulated polarization image is compared with the image directly obtained by the polarization camera under the same condition. Compared with the image from the polarization camera, we get the same trend of contrast information, while the target characteristic information is obviously improved for the intensity image and the error rate of the improved contrast does not exceed 12%.

For the Cellular Yule-Nielsen Spectral Neugebauer (CYNSN) model, the inverse nonlinear Gauss-Seidel iterative algorithm is studied. It is proposed to use the XYZ method of multi-illuminant dimensionality reduction to achieve the accelerated inverse CYNSN model, and use 3-ink and 1-ink printer data to test. It was found that for the 3-ink combination, the XYZ dimension reduction using one light source (D50), and the 4-ink combination, the reverse CYNSN model of XYZ dimension reduction under two light sources (D50, Lambda), not only in accuracy (more chromatic aberration under the light source), and have advantages in calculation speed.

Most viable modern chromatic adaptation transforms (CATs), such as CAT16 and CAT02, can trace their roots both conceptually and mathematically to a simple model formulated from the hypotheses of Johannes von Kries in 1902, known as the von Kries transform/model. However, while the von Kries transform satisfies the properties of symmetry and transitivity, most modern CATs do not satisfy these two important properties. In this Letter, we propose a generalized von Kries transform, which satisfies the symmetry and transitivity properties in addition to improving the fit to most available experimental visual datasets on corresponding colors.

Display characterization is one of the key-problems for colour management, and in the early stage focus is on developing transforms between the display digital driving signals RGB and the colorimetric values XYZ. The GOG and PLCC models were widely considered for this kind of applications in the literature. Recently, in order to reproduce colour match in spectral, display spectral characterization becomes a hot research topic, which has a very important application for the reproduction of multispectral images. In this paper, the well-known GOG and PLCC models are proposed for spectral characterization for the liquid crystal displays. Though the GOG and PLCC models have been widely considered for the display characterization application, it seems that there are no discussions for the display spectral characterization in the literature. It is first shown in this paper that the GOG and PLCC models can indeed be used for display spectral characterization under the assumptions of channel independence and chromaticity constancy for each channel. Performance of the proposed models together with SRPM and SRPPM models are considered using the two widely used professional displays: EIZO CG277 and BENQ PG2401 LCD. At the same time, comparisons are also considered for the GOG and PLCC models trained using the pure red/green/blue colour data and the grey scale (neutral point) data respectively. The comparison results have shown that both GOG and PLCC perform better trained using the grey scale (neutral point) data than those trained using the pure red/green/blue colour data. Furthermore, the comparison results have also shown that PLCC model trained using the grey scale ( neutral-point) data performs better than the SRPPM and GOG models according to both forward and inverse models. Especially, the inverse of the PLCC model is much simpler than the inverse of the SRPPM model. Hence the PLCC model is recommended for the LCD spectral characterization.