Jurrien Bijhold

Methods for identification of images acquired with Digital cameras

From the court we were asked whether it is possible to determine if an image has been made with a specific digital camera. This question has to be answered in child pornography cases, where evidence is needed that a certain picture has been made with a specific camera. We have looked into different methods of examining the cameras to determine if a specific image has been made with a camera: defects in CCDs, file formats that are used, noise introduced by the pixel arrays and watermarking in images used by the camera manufacturer.

Content based information retrieval in forensic image databases.

This paper gives an overview of the various available image databases and ways of searching these databases on image contents. The developments in research groups of searching in image databases is evaluated and compared with the forensic databases that exist. Forensic image databases of fingerprints, faces, shoeprints, handwriting, cartridge cases, drugs tablets, and tool marks are described. The developments in these fields appear to be valuable for forensic databases, especially that of the framework in MPEG-7, where the searching in image databases is standardized. In the future, the combination of the databases (also DNA-databases) and possibilities to combine these can result in stronger forensic evidence.

Pattern recognition in a database of cartridge cases

Several systems exist for collecting spent ammunition for forensic investigation. These databases store images of cartridge cases and the marks on them. The research in this paper is focused on the different methods of feature selection and pattern recognition that can be used for comparison. For automatic comparison of these images it is necessary to extract firstly the useful parts of the images. On databases of 3800 images several processing steps have been tested and compared. The results and methods, which have been implemented, are presented. The usual correlation methods based on gray values of all relevant image data have been tested. They were useful in the database. Further invariant image descriptors and the a trous wavelet transform have been implemented. These methods are promising, however more investigation is needed for the use of these methods.

Tracking people and cars using 3D modeling and CCTV

The aim of this study was to find a method for the reconstruction of movements of people and cars using CCTV footage and a 3D model of the environment. A procedure is proposed, in which video streams are synchronized and displayed in a 3D model, by using virtual cameras. People and cars are represented by cylinders and boxes, which are moved in the 3D model, according to their movements as shown in the video streams. The procedure was developed and tested in an experimental setup with test persons who logged their GPS coordinates as a recording of the ground truth. Results showed that it is possible to implement this procedure and to reconstruct movements of people and cars from video recordings. The procedure was also applied to a forensic case. In this work we experienced that more situational awareness was created by the 3D model, which made it easier to track people on multiple video streams. Based on all experiences from the experimental set up and the case, recommendations are formulated for use in practice.

Pilot investigation of automatic comparison of striation marks with structured light

We have developed and tested an algorithm that can compare striation marks that are acquired with a standard camera and sidelight as well as 3D-information acquired with structured light.

Use of gait parameters of persons in video surveillance systems

The gait parameters of eleven subjects were evaluated to provide data for recognition purposes of subjects. Video images of these subjects were acquired in frontal, transversal, and sagittal (a plane parallel to the median of the body) view. The subjects walked by at their usual walking speed. The measured parameters were hip, knee and ankle joint angle and their time averaged values, thigh, foot and trunk angle, step length and width, cycle time and walking speed. Correlation coefficients within and between subjects for the hip, knee and ankle rotation pattern in the sagittal aspect and for the trunk rotation pattern in the transversal aspect were almost similar. (were similar or were almost identical) This implies that the intra and inter individual variance were equal. Therefore, these gait parameters could not distinguish between subjects. A simple ANOVA with a follow-up test was used to detect significant differences for the mean hip, knee and ankle joint angle, thigh angle, step length, step width, walking speed, cycle time and foot angle. The number of significant differences between subjects defined the usefulness of the gait parameter. The parameter with the most significant difference between subjects was the foot angle (64 % - 73 % of the maximal attainable significant differences), followed by the time average hip joint angle (58 %) and the step length (45 %). The other parameters scored less than 25 %, which is poor for recognition purposes. The use of gait for identification purposes it not yet possible based on this research.

Forensic Video Investigation

At forensic laboratories, cases are submitted for forensic investigation of videotapes and recorders. The video investigation is often requested by the court for different types of crime, like robberies and homicides. For these investigations, most laboratories use a computer to digitize the images uncompressed in such that they try to minimize the wear of the videotape as much as possible.

CCD fingerprint method for digital still cameras

We have reported the Charge Coupled Device (CCD) fingerprint method for identification of digital still cameras. The CCD fingerprint method utilizes the nonhomogeneous nature of dark currents in CCDs. In this study, we have measured CCD defects patterns of various digital still cameras including professional cameras and cheap ones with various resolution and compression rates. As a result, CCD defect pattern was detected in all cameras except for a low-resolution cheap camera using only one image. Resolution mode change of digital cameras did not affect the position of defect points in general but in some cases, relative pixel intensity varied. Image compression did not affect the pixel position for blank images within normal compression rate, but when there existed light in the background, the pixel position was blurred as the compression rate became high. In conclusion, it is recognized that the CCD fingerprint method can be applied in principle to digital still cameras, that is, individual camera identification can be achieved in principle by using images taken with the camera.

Image matching algorithms for breech face marks and firing pins in a database of spent cartridge cases of firearms

On the market several systems exist for collecting spent ammunition data for forensic investigations. These databases store images of cartridge cases and the marks on them. Image matching is used to create hit lists that show those cartridges in the database which have marks that are most similar to the marks of the cartridge case under investigation. The research in this paper focuses on the different methods of feature selection and pattern recognition that can be used for optimizing the results of image matching. A fast pre-selection method based on signatures is applied that is based on the Kanade Lucas Tomasi equation. The positions of the points compared with this method are compared. In this way 11 of the 49 images were in the top position in combination with the third scale of the a trous wavelet. Light conditions and the prominence of the marks determines to a large extent whether correct matches are found in the top ranked position. All images were retrieved in the top give percent of the complete database. This method takes only a few minutes, which can be structured for comparisons to be carried out in seconds.

Facial image comparison using a 3D laser scanning system

To reliably perform comparisons of facial images, it is important to position the head corresponding to the facial images available. Techniques using three or more landmark points on the face have been proposed for matching the face and camera positions to the available photographs. However, these methods can be cumbersome, and require the cooperation of the subject. 3D photographs, together with 3D modeling software, offer the possibility of flexible and reproducable positioning of the head of a person corresponding to the face and camera position of the facial images. We will present our experiences with a non-contact 3D laser-scanning system (Minolta VI-900), especially with respect to ease-of-use, reproducabilty, and performance for facial comparison applications.