3D-reconstructed relocation of spectral feature is a key technology in multi-spectral imaging fusion and 3D reconstruction. Here, the feasibility was to explore about relocating spectral data onto 3D-reconstructed dense point cloud through a new neural network HF-Net, one deep-learning-based AI technology and also a hierarchical localization approach based on a monolithic CNN that is able to simultaneously predict local features and global descriptors for spectral localization. Such an HF-Net was adopted to carry out the spectral relocation of heterogeneous localization. Specifically, the MobileNet and NetVLAD layers were taken to extract the global descriptors among the dataset of three-dimensional color point cloud from the spectral image so as to find the approximate position of spectral image in the three-dimensional point cloud. With the conjectural locations obtained through the prior global retrieval within those candidate places, the SuperPoint was utilized to get the local descriptors and key point scores of the spectral image so that the matching spectral points were found in the three-dimensional point cloud, therewith having mapped out the spectral information and three-dimensional reconstruction pattern. By leveraging the learned descriptors, this assay achieved remarkable localization robustness across large variations of appearance, demonstrating more robust and efficient than SSD algorithm. Due to limits with GPU (graphics processing unit) memory, the extracted spectral features were down-sampled from image to the largest resolution of 6016 × 4512 pixels. With HF-Net being trained through multi-task distillation in TensorFlow 1.12, a spectral image had been able to relocate into 3D reconstruction pattern in 12s under having run on the device of NVidia TESLA V100 with 32G memory and CPU of Intel (R) Xeon (R) Silver 4114 with 12G memory. The approach proposed here can realize the fine positioning of 3D spatial information and spectral features of 3D physical evidence. At present, there seems no such an exploration in forensic science home and abroad, revealing that the exploration tried here is a new application of artificial intelligence to the full-dimensional imaging data fusion technology of physical evidence. Such an HF-Net-based relocation is accurate, scalable, efficient, and a monolithic deep neural network choice for descriptor extraction, capable of achieving higher exactness on large-scale multi-spectral imaging fusion and 3D reconstruction.

Objective Bloodstains, among the most commonly biological evidential materials found and extracted at crime scenes, are capable of providing clues and evidential proof for crime investigation. Hyperspectral imaging technology has presently been attracting attentions of many scholars and forensic professionals. This paper tried to study the ability of spectral imaging technology to reveal and distinguish human bloodstains from the animal's with both the field investigation and laboratory examination. Methods Bloodstains of human and two animals (chicken, snake) were separately made on each of four substrates (black/blue cloth and black/red leather), having their individual image collected. Each collected image was extracted of its spectral values that were converted into functional data and then analyzed with: 1) principal component analysis for quantitative determination of the bands where the three bloodstains' spectra were mainly concentrated into difference; 2) cluster analysis into the scores of principal components; 3) Fisher discriminant analysis for species classification. Results The principal component analysis showed that the first and second principal components contributed more than 96% of the related spectral information throughout the four substrate-harbored samples, demonstrating its ability to identify different bloodstains. Fisher discriminant analysis verified that the three species' bloodstains had been indeed discriminated accurately with the two principal components, rendering a discrimination accuracy of respective 96.6%, 98.8%, 99.0% and 94.6% for the three species' bloodstains on four substrates. Conclusions It is effective to distinguish human bloodstains from the animal's through spectral imaging data and their functional conversion. The spectral imaging technology and functional data analysis are feasible as choice for bloodstains to have their species identified.

Objective To probe into the feasibility and accuracy about the UAV(Unmanned Aerial Vehicle)-carried hyperspectral imaging system to find and recognize the bloodstain on the surface of bitumen road so as to provide a new approach for the application of UAV into criminal investigation. Methods The rotor UAV, equipped with the coaxial reflection convex holographic grating hyperspectral imaging system, was used to collect the visible and near-infrared spectral imaging data of human blood, paint, soy sauce and other sampling traces on the surface of bitumen road. Different algorithms were selected to analyze and tackle the collected spectral imaging data. Results For the human blood, soy sauce and paint, their reflectance spectra are significantly different in the waveband of 600~800nm. With adoption of the spectral angle mapping algorithm and the related maximal angular threshold set to 0.1 radian, the UAV, flying at 30m high, can harvest the spectral imaging data of 12000m² scanning area through its one single takeoff and landing. Consequently, the bloodstain can be accurately discerned from the bitumen pavement if it is of the size over 4cm × 4cm, having been distinguished from paint, soy sauce and other traces. Conclusions The UAV-carried hyperspectral imaging system is applicable for crime scene hunting to discover and/or perceive biological evidential materials, e.g., the bloodstain.

Objective To explore the feasibility of applying the near-infrared (NIR) spectral imaging technology in searching for the evidence at crime scene. Methods With the self-developed portable NIR spectral imager, tests were carried out on blood, firecracker powder, hair, hidden or overlaid letters/text and other evidence-category materials. Spectral images, collected from 650 nm to 1100 nm under the stepping interval of 10 nm, were analyzed. Results The NIR spectral imaging technology could enhance the blood visualization on the fabric of dark or complex background, yet the effect was affected by both the blood concentration and the NIR reflectance of the background substance. Besides, blood, when diluted more than 10 folds or washed, could be hardly detected. Meanwhile, NIR spectral imaging technology can make it easier to discover firecracker powder and hair on dark fabric, especially lessening the influence of complex background from fabric. In addition, the technology could be applied to show hidden or overlaid letters/text and identify materials of different substances. Conclusion It is feasible to quickly find evidence in crime scene with NIR spectral imaging technique. The results here should provide a foundation for the spectral camera to be further developed more compact.

Hyperspectral imaging combines spectroscopy an d digital i maging technology.By the use of an imaging spectrometer，hyperspectral imaging cap tures a serial of images of reflective light or fluorescent light from examined object， and forms a data set of spectral image which contains intensity and spe ctral information of the object.This imaging method can be used to reveal patte rn details and to identify material component in forensic science.

Objective To explore the application of spectral imaging to forensic science. Methods The images obtained from the spectral imaging analysis were compared to the results of currently used techniques and methodology. Results The ideal photography was got with spectral imaging system. Conclusion The spectral imaging method proved to have significant capabilities and potential application in forensic science.

ObjectiveResearch on making spectral imaging applying much standardization. MethodsThe images obtained from the spectral imaging analysis were compared to different datum marks. ResultsVariety results were got with different datum marks. ConclusionFor different inspection objectives choosing different datum marks has very strong disciplinarian.

Objective To explore the capability of spectral imaging for idengtification of same colour cotton fibers.MethodsTen red colour cotton fibers, ten blue colour cotton fibers and ten black cotton fibers were analyzed by spectral imaging.ResultsThe rate of discrimination for the red fiber, blue fiber and black fiber are 62.2%, 77.8% and 69.6% respectively. ConclusionWith the spectral imaging method, most same colour cotton fibers can be identified.

Objective To study the method of removing the background interference of the developed fingerprint using spectral imaging technology. Methods Multi-spectral images of two developed fingerprint samples disturbed by the background pattern were acquired, whose spectra ranges from 400 nm to 720 nm, with the sampling interval of 10 nm. Two specific spectral images were selected from the data cube by comparing the spectra of the fingerprint and the background or observing the spectral images directly, then the division of two images was performed and the resulted image was enhanced by linear stretching, finally the enhanced fingerprint image with a weakened background was obtained. Results The background pattern interfered with the developed fingerprint on Renminbi was effectively removed, and so was the striped background disturbing the developed fingerprint on envelope. In another word, the fingerprints were significantly enhanced. Conclusion The method of removing the background interference by division of two featured spectral images based on visible spectral imaging technology is very suitable for the case of enhancing the developed fingerprint on absorbent object, but the background is required to have a consistent color hue or little color change.

The oil-sweat fingerprints were developed by UV(200nm～400nto)spectral imaging and the results showed that UV spectral imaging call be used for detecting the fingerprints on non-porous surfaces．

目的 考察在运用拉曼光谱阵列面扫描检验朱文时序的方法中印文油墨种类对检验结果的影响。方法 根据拉曼谱图的不同对搜集的21种印泥印油进行扫描并分类，采用拉曼光谱阵列面扫描技术对印泥印油与同一书写笔形成的朱文时序样本进行检验。结果 21种印泥印油被分为11类，21种印泥印油与蓝色签字笔形成的朱文时序样本都得到了良好的检验结果。结论 印文油墨种类用拉曼光谱阵列面扫描检验朱文时序的检验结果没有影响。

Objective To identify the documents printed with Fuji Xerox C4300 Printer.Methods The tracking code on documents printed with six sets of Fuji Xerox C4300 Printer were examined by Multi-spectral Imager.Results The tracking codes of six sets of Fuji Xerox C4300 Printer were different.Conclusion The documents printed with different Fuji Xerox C4300 Printer can be identified by examination of tracking code.

Objective To explore the capability of fluorescence spectral imaging for identification of the authenticity of currency and determine whether the counterfeit money came from the same source.Methods 50 pieces of real 100yuan in 1999 edition, 50 pieces of real 100yuan in 2005 edition, 12 pieces of counterfeit 100yuan in 1999 edition, and 12 pieces of counterfeit 100yuan in 2005 edition were analyzed by fluorescence spectral imaging.Results The counterfeit money and the money coming from the different source were distinguished.Conclusion The fluorescence spectral imaging was a very effective method for distinguishing the counterfeit money and the money of different source.

Objective To explore a method for discrimination of electrical tape with same color.Methods Electrical tapes (red, yellow, blue and green) were analyzed by spectral imaging.Results The rate of discrimination for the blue, red, green and yellow electrical tapes are 99.0%, 94.3%, 93.8% and 93.3% respectively.Conclusion With the spectral imaging method, most electrical tapes can be discriminated.

Objective To establish a method for correcting images aspect ratio of the THEMIS UV spectral imaging system and obtain the parameterβof the system.Methods Test pictures made with Adobe Photoshop were analyzed by THEMIS UV spectral imaging system. The difference of aspect ratio of images were measured and corrected with MATLAB.Results Based on differences between test pictures and images, the imaging defects and the parameterβwere obtained.Conclusion UV spectral image can be recovered to the original aspect ratio and the parameterβcan be used for recovering the original size of objects.

Objective Generally, the optical properties, chemical and physical characteristics of writing inks are helpful to identify the added and overwrite handwriting in the examination of questioned document, but it does not work for most carbonaceous ink marks. This paper provides a non-destructive, convenient and effective testing method for identifying the added and overwrite handwriting by black gel ink. Methods Added and overwrite handwritings with black gel ink were examined by electric resistance measurement and spectral imaging technique. The difference of resistance value and spectrograms between questioned carbonaceous ink marks and known black pen marks were analyzed for determining authenticity of the writings. Results The handwritings written by black gel inks made with different types of pigments were classified by their resistance value and spectrograms. Conclusions The electric resistance measurement and spectral imaging technique, which breaks through the limitation of morphology inspection in detecting altered handwriting written by carbonaceous black gel inks, basing on the quantitative detection, can make the final conclusion more scientific and reliable.

Objective To establish a non-destructive, efficient and direct method for visualizing the handwriting covered and smeared by black gel pen inks. Methods 306 smeared and covered handwriting samples prepared with 18 kinds of black gel pen were analyzed by using Nuance420 Spectral Imaging with the technique of transmission spectral imaging, compared with reflective spectral imaging. The results were divided into two types: visual and non-visual. The standard of classification was whether common people who had never seen the original handwritings could correctly recognize the smeared and covered characters. Results Of 306 test samples, 231 samples were visual and 75 ones were non-visual with transmission spectral imaging technique. The visualized rate was 75.49%, by contrast with the visualized rate of 51.96% with reflective spectral imaging. Conclusions Transmission spectral imaging could differentiate the handwriting smeared and covered by black gel pen inks with the exception that the original writings and the smearing and covering layer used same or similar composition inks. Therefore, the technique of transmission spectral imaging is an efficient choice to uncover the smeared and covered handwritings by black gel pen, which can be applied in questioned document examination.

ObjectiveTo explore the application of spectral imaging for examination of multilayered paint. Methods623 multilayered paint samples collected from twelve provinces were analyzed using spectral imaging technology and the layer structure of paint chips were compared. ResultsThe paint's layer structures were obtained and the all of 623 paint samples can be discriminated. ConclusionThe results of this project demonstrated that spectral imaging offers advantages over conventional visualization techniques and can be used for trance evidence examination.

Objective 　To establish a non‐destructive method for visualizing the handwriting covered by same color inks ． Method 　The covered handwritings were examined by using spectral imaging technology ．Results 　For ball pen and gel pen ， black and blue handwritings covered with the same color inks have been visualized by spectral imaging ．Conclusion 　Spectral imaging technology can be used for visualizing the handwriting covered by same color inks ．