Research On Stereo Footprint Measurement Method Based On Depth Camera

The three-dimensional footprint is an important basis for the analysis and reconstruction of the crime scene,and the effective and simple extraction and measurement of it is the focus of criminal science and technology.Although the traditional three-dimensional footprint extraction methods such as gypsum modeling and silicone rubber modeling can save the three-dimensional information of the three-dimensional footprint more completely,they will also destroy the original traces.The non-contact footprint extraction method is used,but most of them can only mine the plane information of the stereo footprint from the image.In order to combine the advantages of the two,this paper uses depth cameras to replace ordinary digital cameras for non-destructive extraction of stereo footprints,and studies related methods of stereo footprint information extraction based on depth cameras,so as to achieve the purpose of automatic measurement of stereo footprints.In this study,we first used a digital camera to collect nearly 22,500 stereo footprint images on the sandy ground to build a sample data set,and randomly selected 2500 images as a test set to prevent overfitting,and the remaining 20,000 images were used as training sets for Mask-Rcnn deep neural network related training,so as to get the detection model.Then calibrate the color and infrared cameras that come with the SR-300 depth camera,find the internal and external parameter matrix corresponding to the two cameras,and then perform the registration process of the depth camera,that is,the depth camera is captured by the obtained internal and external parameter matrix.Each pixel on the color map and the depth map correspond to each other.After the registration is completed,use the depth camera's color and infrared cameras to take pictures of the three-dimensional footprint at the same time,and use the mask generated by Mask-Rcnn to segment the three-dimensional footprint in the image and extract the extracted All pixel coordinates in the footprint area are saved,and then these pixel coordinates are processed by related algorithms to obtain the required pixel coordinates of the feature points such as foot length,palm width,arch width,heel width,and depth,and finally in their corresponding depth maps.By acquiring the depth information of these feature points,the three-dimensional information of the three-dimensional footprint can be obtained.Compared with the planar footprint,the three-dimensional footprint carries more information about the criminal suspects,which can help the investigators to narrow the scope of the investigation or lock the suspects faster in some cases.This paper proposes a non-contact non-destructive extraction method to measure the three-dimensional footprint of the site,which can effectively obtain the three-dimensional information of the three-dimensional footprint,and makes up for the lack of the current three-dimensional footprint measurement equipment because it is relatively bulky and difficult to bring into the site for measurement and analysis.In addition,on the premise of not needing to manually take feature points,automatic measurement is realized,which has great practical value for the rapid measurement and analysis of on-site three-dimensional footprint.