The supply chain of agricultural products is intricately linked to the daily lives of people. In light of rising import and export quantities, the need for a prompt and efficient inspection system has become increasingly pressing. Without opening baskets and manually sorting, a smart inspection scheme is designed in this work leveraging X-ray images and transformer neural network. Due to its penetrating capabilities, X-ray enables a direct examination of agricultural products within a basket, a task that normal vision devices are unable to accomplish. Taking into account the varying shapes and combinations of agricultural products, we introduce a transformer-based deep neural network for type identification. Additionally, a dataset augmentation process is developed inspired by computed tomography generating 1,6000 X-ray images. Through experiments, the proposed smart inspection scheme is proven to be feasible and works efficiently. The inspection accuracy for both single-type and mixed-type agricultural products on the established dataset exceeds 90%.
Traditional reconstruction algorithm in computed tomography (CT) requires a square reconstruction matrix, regardless of the particular object shape. It becomes inefficient when reconstructing planar objects such as IC chips which have a large area-to-thickness ratio. This article presents a modified cone-beam reconstruction algorithm which is efficient for planar multilayer objects. by detecting the orientation and dimension parameters of the object for the first projection, a reconstruction volume can be defined in a region that would exactly cover the object area and follow the actual
orientation of the object. This new technology is demonstrated in both its general form and a targeted application. Compared to conventional cone-beam CT reconstruction, this new method uses much less computation time and storage, can achieve higher reconstruction resolution in the thickness dimension, and makes layer separation much easier for multilayer objects. These advantages will be demonstrated with two high-resolution CT inspection applications of a
stacked IC and an advanced packaging device.
Image intensifier is widely used in X-ray imaging system for both radiography and tomography internal defects/structure
inspections. A typical image intensifier comprises of a vacuum tube and a CCD camera. It is the combination of the
CCD' s zoom function and the tube's dual-field capability that enables the device to achieve variable field-of-view and
resolution. However, it is also because of this feature, it becomes difficult to correct the pincushion distortion of the
image that is caused by the curvature of the input window of the tube. This paper will present a method that can
dynamically correct this kind of distortion in a commercially available X-ray inspection system. With this method, the
curvature of the tube input window is first calibrated with a calibration unit. Then with each image, three edge points of
the input window are detected, from which both the radius and centre of the field of view are determined in pixel. These
parameters are then used to map the distorted image into a correct image by establishing the transform relationship
between them. This method can be fully automatic, therefore, it is able to be implemented on the existing system to
directly display the distortion-corrected image of an object.
Traditional industrial computed tomography systems generally calibrate the central ray by scanning a wire phantom. That is time-consuming and slow. This paper presents a direct central-ray determination method using the projection data of the object to be examined. The basic idea is that with a fan-beam arrangement, any straight line on the object slice will align with the x-ray point source at only two times during a 360-deg scan. By measuring the mismatching of the two sets of projection data that correspond to these two alignments, the true central ray can be identified at the minimum of the measurement. Our test has proven that this automated method is fast, reliable, and accurate.
High-resolution X-ray is now an essential tool for internal defect and structure inspection in electronics and advanced
materials industry. However, it is always a challenge to use it for accurate dimensional measurement due to the nature of
the fan-beam X-ray source, particularly for cylindrical objects. This paper presents a novel hybrid X-ray and microscopy
method for the profile measurement of the internal hole of a
cylinder-shaped steel component. The part to be measured
has a beer bottle shape but is open at the bottom side. The objective is to measure the diametrical profile of the internal
hole with an accuracy of about 10μm. Traditionally this is measured with using a microscope after cutting and polishing
the specimen. This is not only a tedious work, but is also inaccurate due to the uncertainty in cutting and polishing. This
report demonstrates that the two edge-profiles of the internal hole can be obtained with X-ray inspection by sequentially
placing each of them at the central of the X-ray beam so that the
fan-beam effect can be minimized. The resolution of the
X-ray inspection is about 6µm under a 20x magnification. Subsequently, the diameter of the hole is measured at two
positions through the open end using a microscope with a 20x and a 10x objectives respectively. The results obtained
with the two methods are then combined to generate the whole diametrical profile of the internal hole.
All computed-tomography (CT) reconstruction algorithms used today require knowledge of the central ray, which is the projection of the center of rotation (COR) on the detector. The current common practice is to determine it using a wire phantom made of a dense metal material before the CT scan of the object to be inspected. This work presents four methods that can be used for this purpose, the center-of-sinogram, the opposite-angle-interpolation, curve-fitting, and geometrical methods. To our best knowledge, the last two approaches have not been reported before. The performance of the four methods is evaluated under four different situations. The comparison study shows that for all situations, the curve-fitting method and the interpolation method have consistent performance, and that only when the wire-to-COR is small, can all four methods generate close results. When the wire-to-COR distance becomes large, the center-of-sinogram method is generally not reliable. Although in principle the geometrical method is able to provide an exact solution, in practice its accuracy is limited by the finite size of the detector pixel and can be improved if the detector pixel size becomes smaller.
This paper presents a new three-wavelength ring laser using one erbium-doped fiber amplifier. The simultaneous output of three lasing wavelengths is achieved by controlling the gains of different filters through utilizing a cascaded fiber Mach-Zender interferometer (MZI) with the triangular-shaped interference transmission spectrum. The key advantage of the proposed system over existing technologies is that the output is channel-wavelength selectable. Each channel can be tuned individually from one channel wavelength to the other. The channel wavelengths are selectable over the range of 1522-1665nm.
The application of phase-shifting photoelasticity to real- time dynamic event involves simultaneous recording of the four phase-shifted images. This purpose is realized in this paper by using a special beam-splitting lens called Multispec Imager. The isoclinic and isochromatic parameters for a given time in a dynamic event are determined by using multi-load to phase-shifting technique. A modified median filter algorithm for wrapped phase data is also presented.
In this paper, a photoelastic stress analysis is carried out for a cracked femur bone with compression plate fixation. A loading rig has been designed and manufactured to apply forces in the physical directions on the 2D modal of the femur bone based on a single-leg-stance. Three femur models made of photoelastic materials had been fabricated with three configurations and loaded. A recently developed three- load to phase shifting method is adopted to extract the full-field quantitative information from the fringe patterns of the loaded models. A comparison shows that the configuration where the screws point away from each other had the best effectiveness.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.