scholarly journals Spatial recall index for machine learning algorithms

2021 ◽  
Vol 2021 (1) ◽  
pp. 58-62
Author(s):  
Patrick Müller ◽  
Mattis Brummel ◽  
Alexander Braun

We present a novel metric Spatial Recall Index to assess the performance of machine-learning (ML) algorithms for automotive applications, focusing on where in the image which performance occurs. Typical metrics like intersection-over-union (IoU), precisionrecallcurves or average precision (AP) quantify the performance over a whole database of images, neglecting spatial performance variations. But as the optics of camera systems are spatially variable over the field of view, the performance of ML-based algorithms is also a function of space, which we show in simulation: A realistic objective lens based on a Cooke-triplet that exhibits typical optical aberrations like astigmatism and chromatic aberration, all variable over field, is modeled. The model is then applied to a subset of the BDD100k dataset with spatially-varying kernels. We then quantify local changes in the performance of the pre-trained Mask R-CNN algorithm. Our examples demonstrate the spatial dependence of the performance of ML-based algorithms from the optical quality over field, highlighting the need to take the spatial dimension into account when training ML-based algorithms, especially when looking forward to autonomous driving applications.

Geophysics ◽  
2013 ◽  
Vol 78 (3) ◽  
pp. WB113-WB126 ◽  
Author(s):  
Matthew J. Cracknell ◽  
Anya M. Reading

Inductive machine learning algorithms attempt to recognize patterns in, and generalize from empirical data. They provide a practical means of predicting lithology, or other spatially varying physical features, from multidimensional geophysical data sets. It is for this reason machine learning approaches are increasing in popularity for geophysical data inference. A key motivation for their use is the ease with which uncertainty measures can be estimated for nonprobabilistic algorithms. We have compared and evaluated the abilities of two nonprobabilistic machine learning algorithms, random forests (RF) and support vector machines (SVM), to recognize ambiguous supervised classification predictions using uncertainty calculated from estimates of class membership probabilities. We formulated a method to establish optimal uncertainty threshold values to identify and isolate the maximum number of incorrect predictions while preserving most of the correct classifications. This is illustrated using a case example of the supervised classification of surface lithologies in a folded, structurally complex, metamorphic terrain. We found that (1) the use of optimal uncertainty thresholds significantly improves overall classification accuracy of RF predictions, but not those of SVM, by eliminating the maximum number of incorrectly classified samples while preserving the maximum number of correctly classified samples; (2) RF, unlike SVM, was able to exploit dependencies and structures contained within spatially varying input data; and (3) high RF prediction uncertainty is spatially coincident with transitions in lithology and associated contact zones, and regions of intense deformation. Uncertainty has its upside in the identification of areas of key geologic interest and has wide application across the geosciences, where transition zones are important classes in their own right. The techniques used in this study are of practical value in prioritizing subsequent geologic field activities, which, with the aid of this analysis, may be focused on key lithology contacts and problematic localities.


Author(s):  
K. Tsuno ◽  
T. Honda ◽  
Y. Harada ◽  
M. Naruse

Developement of computer technology provides much improvements on electron microscopy, such as simulation of images, reconstruction of images and automatic controll of microscopes (auto-focussing and auto-correction of astigmatism) and design of electron microscope lenses by using a finite element method (FEM). In this investigation, procedures for simulating the optical properties of objective lenses of HREM and the characteristics of the new lens for HREM at 200 kV are described.The process for designing the objective lens is divided into three stages. Stage 1 is the process for estimating the optical properties of the lens. Firstly, calculation by FEM is made for simulating the axial magnetic field distributions Bzc of the lens. Secondly, electron ray trajectory is numerically calculated by using Bzc. And lastly, using Bzc and ray trajectory, spherical and chromatic aberration coefficients Cs and Cc are numerically calculated. Above calculations are repeated by changing the shape of lens until! to find an optimum aberration coefficients.


Author(s):  
K. Shibatomi ◽  
T. Yamanoto ◽  
H. Koike

In the observation of a thick specimen by means of a transmission electron microscope, the intensity of electrons passing through the objective lens aperture is greatly reduced. So that the image is almost invisible. In addition to this fact, it have been reported that a chromatic aberration causes the deterioration of the image contrast rather than that of the resolution. The scanning electron microscope is, however, capable of electrically amplifying the signal of the decreasing intensity, and also free from a chromatic aberration so that the deterioration of the image contrast due to the aberration can be prevented. The electrical improvement of the image quality can be carried out by using the fascionating features of the SEM, that is, the amplification of a weak in-put signal forming the image and the descriminating action of the heigh level signal of the background. This paper reports some of the experimental results about the thickness dependence of the observability and quality of the image in the case of the transmission SEM.


Author(s):  
William Krakow

An electronic device has been constructed which manipulates the primary beam in the conventional transmission microscope to illuminate a specimen under a variety of virtual condenser aperture conditions. The device uses the existing tilt coils of the microscope, and modulates the D.C. signals to both x and y tilt directions simultaneously with various waveforms to produce Lissajous figures in the back-focal plane of the objective lens. Electron diffraction patterns can be recorded which reflect the manner in which the direct beam is tilted during exposure of a micrograph. The device has been utilized mainly for the hollow cone imaging mode where the device provides a microscope transfer function without zeros in all spatial directions and has produced high resolution images which are also free from the effect of chromatic aberration. A standard second condenser aperture is employed and the width of the cone annulus is readily controlled by defocusing the second condenser lens.


Author(s):  
J. S. Lally ◽  
R. Evans

One of the instrumental factors often limiting the resolution of the electron microscope is image defocussing due to changes in accelerating voltage or objective lens current. This factor is particularly important in high voltage electron microscopes both because of the higher voltages and lens currents required but also because of the inherently longer focal lengths, i.e. 6 mm in contrast to 1.5-2.2 mm for modern short focal length objectives.The usual practice in commercial electron microscopes is to design separately stabilized accelerating voltage and lens supplies. In this case chromatic aberration in the image is caused by the random and independent fluctuations of both the high voltage and objective lens current.


Author(s):  
S. Takashima ◽  
H. Hashimoto ◽  
S. Kimoto

The resolution of a conventional transmission electron microscope (TEM) deteriorates as the specimen thickness increases, because chromatic aberration of the objective lens is caused by the energy loss of electrons). In the case of a scanning electron microscope (SEM), chromatic aberration does not exist as the restrictive factor for the resolution of the transmitted electron image, for the SEM has no imageforming lens. It is not sure, however, that the equal resolution to the probe diameter can be obtained in the case of a thick specimen. To study the relation between the specimen thickness and the resolution of the trans-mitted electron image obtained by the SEM, the following experiment was carried out.


Author(s):  
David A. Ansley

The coherence of the electron flux of a transmission electron microscope (TEM) limits the direct application of deconvolution techniques which have been used successfully on unmanned spacecraft programs. The theory assumes noncoherent illumination. Deconvolution of a TEM micrograph will, therefore, in general produce spurious detail rather than improved resolution.A primary goal of our research is to study the performance of several types of linear spatial filters as a function of specimen contrast, phase, and coherence. We have, therefore, developed a one-dimensional analysis and plotting program to simulate a wide 'range of operating conditions of the TEM, including adjustment of the:(1) Specimen amplitude, phase, and separation(2) Illumination wavelength, half-angle, and tilt(3) Objective lens focal length and aperture width(4) Spherical aberration, defocus, and chromatic aberration focus shift(5) Detector gamma, additive, and multiplicative noise constants(6) Type of spatial filter: linear cosine, linear sine, or deterministic


Author(s):  
Willem H.J. Andersen

Electron microscope design, and particularly the design of the imaging system, has reached a high degree of perfection. Present objective lenses perform up to their theoretical limit, while the whole imaging system, consisting of three or four lenses, provides very wide ranges of magnification and diffraction camera length with virtually no distortion of the image. Evolution of the electron microscope in to a routine research tool in which objects of steadily increasing thickness are investigated, has made it necessary for the designer to pay special attention to the chromatic aberrations of the magnification system (as distinct from the chromatic aberration of the objective lens). These chromatic aberrations cause edge un-sharpness of the image due to electrons which have suffered energy losses in the object.There exist two kinds of chromatic aberration of the magnification system; the chromatic change of magnification, characterized by the coefficient Cm, and the chromatic change of rotation given by Cp.


Author(s):  
K. Ishizuka ◽  
K. Shirota

In a conventional alignment for high-resolution electron microscopy, the specimen point imaged at the viewing-screen center is made dispersion-free against a voltage fluctuation by adjusting the incident beam direction using the beam deflector. For high-resolution works the voltage-center alignment is important, since this alignment reduces the chromatic aberration. On the other hand, the coma-free alignment is also indispensable for high-resolution electron microscopy. This is because even a small misalignment of the incident beam direction induces wave aberrations and affects the appearance of high resolution electron micrographs. Some alignment procedures which cancel out the coma by changing the incident beam direction have been proposed. Most recently, the effect of a three-fold astigmatism on the coma-free alignment has been revealed, and new algorithms of coma-free alignment have been proposed.However, the voltage-center and the coma-free alignments as well as the current-center alignment in general do not coincide to each other because of beam deflection due to a leakage field within the objective lens, even if the main magnetic-field of the objective lens is rotationally symmetric. Since all the proposed procedures for the coma-free alignment also use the same beam deflector above the objective lens that is used for the voltage-center alignment, the coma-free alignment is only attained at the sacrifice of the voltage-center alignment.


Author(s):  
T. Miyokawa ◽  
H. Kazumori ◽  
S. Nakagawa ◽  
C. Nielsen

We have developed a strongly excited objective lens with a built-in secondary electron detector to provide ultra-high resolution images with high quality at low to medium accelerating voltages. The JSM-6320F is a scanning electron microscope (FE-SEM) equipped with this lens and an incident beam divergence angle control lens (ACL).The objective lens is so strongly excited as to have peak axial Magnetic flux density near the specimen surface (Fig. 1). Since the speciien is located below the objective lens, a large speciien can be accomodated. The working distance (WD) with respect to the accelerating voltage is limited due to the magnetic saturation of the lens (Fig.2). The aberrations of this lens are much smaller than those of a conventional one. The spherical aberration coefficient (Cs) is approximately 1/20 and the chromatic aberration coefficient (Cc) is 1/10. for accelerating voltages below 5kV. At the medium range of accelerating voltages (5∼15kV). Cs is 1/10 and Cc is 1/7. Typical values are Cs-1.lmm. Cc=l. 5mm at WD=2mm. and Cs=3.lmm. Cc=2.9 mm at WD=5mm. This makes the lens ideal for taking ultra-high resolution images at low to medium accelerating voltages.


Sign in / Sign up

Export Citation Format

Share Document