Quantum mechanical machine vision system and arithmetic operation method based on quantum dot

Abstract

A quantum mechanical arithmetic operation method for machine vision, based on quantum dots is performed by a quantum processing processor. The quantum mechanical arithmetic operation method comprises, obtaining a first labeled graph connecting between feature points of the first image and a second labeled graph connecting feature points of the second image, generating a point-to-point combination by matching the feature points of the first image with the feature points the second image, generating a conflict graph by adding the largest point-to-point combination by comparing the point-to-point combinations with the threshold, generating non-constrained binary optimization equation for finding a maximum independent set of conflict graphs, converting the non-constrained binary optimization equation into Ising model of the quantum system, and calculating the Hamiltonian of Ising model based on the quantum dots to obtain solution of the non-constrained binary optimization equation.

Description

CROSS REFERENCE TO RELATED APPLICATION

This application claims priority from and the benefit of Korean Patent Applications No. 10-2016-0168244, filed on Dec. 12, 2016, which is hereby incorporated by reference for all purposes as if fully set forth herein.

BACKGROUND OF THE INVENTION

Field of the Invention

The present invention relates to a quantum mechanical machine vision system and an arithmetic operation method, more specifically to a quantum mechanical machine vision system and an arithmetic operation method based on quantum dot.

Discussion of the Background

Human beings now have better analytical capabilities than machine analysis in many areas such as object recognition, knowledge representation, reasoning, learning and natural language processing. Accordingly, in order to imitate or surpass the human way of thinking mechanically, a complicated arithmetic operation method must be used.

An accurate solution to the problem of optimization of machine vision system is required to imitate or surpass human visual recognition ability as an example.

In order to solve the complex computation method of artificial view, there is a method of performing quantum mechanical calculation using quantum computing.

A quantum computer is a physical system that uses one or more quantum effects to perform calculations. A quantum computer capable of efficiently simulating other quantum computers is called a universal quantum computer (UQC).

1. Approach to Quantum Computation

There are several general approaches to the design and operation of quantum computers.

One approach corresponds to a ‘circuit model’ of quantum computation. In this approach, qubits operate in the order of a logical gate, which is a representation of a compiled algorithm. Circuit model quantum computers have some serious barriers in their actual implementation. In a circuit model, qubits are required to be coherent for a longer period of time than a single-gate time. This demand arises because circuit model quantum computers require operations, called quantum error correction, to operate. Quantum error correction cannot be performed without the qubit of a circuit model quantum computer that can maintain quantum coherence for a time interval of about 1000 times one gate time. There have been a number of studies focused on developing qubits with sufficient coherence to form basic information units of quantum computers. This is described in, for example, “Introduction to Quantum Algorithms”, by Shor, P. W. arXiv. org: quantph/0005003 (2001), pp. 1-27. This technical field is still stagnant due to the lack of the ability to enhance the coherence of the qubit to a level suitable for designing and operating real circuit model quantum computers.

2. Computational Complexity Theory

In computer science, computational complexity theory is a kind of computational theory required to solve a given computational problem and the theory of computation to study resources or costs. Costs are generally measured by abstract parameters called computational resources, such as time and space. The time means the number of steps necessary to solve the problem, and the space means the required amount of information storage or the amount of memory required.

Optimization problems correspond to problems where one or more objective functions are minimized and maximized under a set of constraints, sometimes with respect to a set of variables.

Simulation problems typically deal with the simulation of one system by another system during a typical time interval. For example, computer simulations consist of business processes, ecological habitats, protein folding, molecular ground states, and quantum systems. These problems often involve numerous diverse entities that are different from complex interrelationships and behavioral rules. Feynman suggests that a quantum system can be used to simulate several physical systems more efficiently than UTM.

Many optimization and simulation problems cannot be solved using UTM. Because of these limitations, computational elements are needed that can solve computational problems beyond the scope of the UTM. Other digital computer based systems and methods for solving optimization problems can be found.

An example of a technique for solving this optimization problem is described in Korean Patent No. 10-1309677 entitled ‘Method for Calculating Adoptive Quantum’.

The prior art discloses a quantum computing method using a quantum system that includes a plurality of qubits. In the prior art, quantum annealing is possible to obtain a desired minimum energy (or cost), which concurrently tracks a configuration of a superposition state, and especially, Adiabatic Quantum Computation (AQC) technique is used to perform quantum annealing. In addition, AQC uses a technique in which an adiabatic change of Hamiltonian from the initial state to the target state is obtained and a solution of the desired target state is finally obtained.

The above prior art describes the general operation of a quantum computing system to solve a complex problem, and in spite of the existence of this prior art, the selection of an optimized quantum system remains a very important problem to solve the complicated matter.

SUMMARY OF THE INVENTION

Accordingly, it is an object of the present invention to provide a quantum mechanical machine vision system and an arithmetic operation method based on quantum dot, which can facilitate calculation of complexity caused by an increase in the number of feature points for image identification.

A quantum mechanical arithmetic operation method based on quantum dots according to an exemplary embodiment of the present invention is performed by a quantum processing processor. The quantum mechanical arithmetic operation method comprises, obtaining a first labeled graph connecting between feature points of the first image and a second labeled graph connecting feature points of the second image, generating a point-to-point combination by matching the feature points of the first image with the feature points the second image, generating a conflict graph by adding the largest point-to-point combination by comparing the point-to-point combinations with the threshold, generating non-constrained binary optimization equation for finding a maximum independent set of conflict graphs, converting the non-constrained binary optimization equation into Ising model of the quantum system, and calculating the Hamiltonian of Ising model based on the quantum dots to obtain solution of the non-constrained binary optimization equation.

For example, calculating the Hamiltonian of Ising model based on the quantum dots to obtain solution of the non-constrained binary optimization equation may be performed by quantum dots arranged in a matrix shape.

In this case, neighboring quantum dots with the shortest distance in a column direction or a row direction may be connected to each other through the tunnel junction.

For example, the Hamiltonian of the Ising model may be calculated through adiabatic evolve, in calculating the Hamiltonian of Ising model based on the quantum dots to obtain solution of the non-constrained binary optimization equation.

On the other hand, the quantum mechanical arithmetic operation method may further comprise repeatedly learning the non-constrained binary optimization equation through machine learning.

A quantum mechanical machine vision system according to an exemplary embodiment of the present invention comprises an image acquisition module, a quantum processing processor and a memory unit. The image acquisition module acquires an image. The quantum processing processor processes the image obtained from the image acquisition module. The memory unit stores data necessary for computation of the quantum processing processor. The quantum processing processor obtains a first labeled graph connecting between feature points of the first image and a second labeled graph connecting feature points of the second image, generates a point-to-point combination by matching the feature points of the first image with the feature points the second image, generates a conflict graph by adding the largest point-to-point combination by comparing the point-to-point combinations with the threshold, generates non-constrained binary optimization equation for finding a maximum independent set of conflict graphs, converts the non-constrained binary optimization equation into Ising model of the quantum system, and calculates the Hamiltonian of Ising model based on quantum dots to obtain solution of the non-constrained binary optimization equation.

For example, the quantum processing processor may comprise quantum dots arranged in a matrix shape.

For example, neighboring quantum dots with the shortest distance in a column direction or a row direction are connected to each other through the tunnel junction.

For example, the quantum processing processor may further comprise a charge detection unit disposed adjacent to the quantum dots.

For example, the quantum processing processor may calculate the Hamiltonian of the Ising model through adiabatic evolve.

As described above, according to the quantum dot-based quantum mechanical machine vision system and an arithmetic operation method of the present invention, the NP problem generated as the number of feature points increases is replaced by Hamiltonian using Ising model, so that it can be easily calculated using quantum dots arranged in a matrix shape.

BRIEF DESCRIPTION OF THE DRAWINGS

The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate embodiments of the invention, and together with the description serve to explain the principles of the invention.

FIG. 1 is a diagram showing a modeling of interrelationships of feature vectors between interest points according to an embodiment of the present invention.

FIG. 2 is a conceptual diagram showing two quantum dots spin qubit.

FIG. 3 is a conceptual diagram showing the arrangement of quantum dots of a quantum processing processor according to an exemplary embodiment of the present invention.

FIG. 4 is a block diagram showing a quantum-dot-based quantum mechanical machine vision system in accordance with an exemplary embodiment of the present invention.

DETAILED DESCRIPTION OF THE ILLUSTRATED EMBODIMENTS

The present invention is described more fully hereinafter with reference to the accompanying drawings, in which example embodiments of the present invention are shown. The present invention may, however, be embodied in many different forms and should not be construed as limited to the example embodiments set forth herein. Rather, these example embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the present invention to those skilled in the art. In the drawings, the sizes and relative sizes of layers and regions may be exaggerated for clarity.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings.

FIG. 1 is a diagram showing a modeling of interrelationships of feature vectors between interest points according to an embodiment of the present invention.

In the machine vision system, the computer or the robot compares the previously captured reference pattern with the photographed image to recognize the image, and this process is advanced through a training process. In general, an algorithm called a heuristic algorithm is applied to a particular type of image, and a heuristic algorithm can be applied in various ways depending on the image.

Generally, a brain recognizing human visual and visual information extracts feature points that can adequately describe a pattern represented by each image for pattern matching between different images, and the distance and direction between the feature points are synthetically recognized, so that the distance and direction are recognized as a pattern. The human brain then compares the pattern information extracted from the image to determine whether the two images match. This process is difficult by searching only by movement between a point and a point, but easy by obtaining patterns including surrounding points.

However, in machine vision with limited intelligence in general, it is extremely difficult to fully simulate the behavior of human brain or the interpretation of human sensory data. It is known as the NP-Hard Problem to simulate the behavior of the human brain.

In FIG. 1, a process of recognizing a pattern between different images is modeled, and the process is described as a general NP-Hard problem. The images of image A and image B are different images but may be different versions of images having the same physical structure.

In order to perform pattern recognition operations related to machine vision, a combination of feature vectors representing relative position information between respective points of interest (feature points) in each image is used, and a combination of these feature vectors is used as a reference for recognizing each image.

The process of extracting a reference pattern for each image can be performed by a process of extracting feature points and a process of determining relative position information between the feature points as a feature vector. This process is not a deterministic problem, but a non-deterministic problem in which it is necessary to find an optimized value by comparing the results.

Assuming that a reference pattern made up of a set of feature vectors between interest points (feature points) i, j, k in the image A in FIG. 1 is X, and a reference pattern made up of a set of feature vectors between interest points (feature points) α, β, γ in the image B in FIG. 1 is Y, it is difficult to calculate the reference pattern X of image A and the reference pattern Y of image B in FIG. 1.

In order to describe the mapping between the image A and the image B in FIG. 1, a reference pattern X which best describes image A and a reference pattern Y which best describes image B must be obtained, and how the reference pattern X is displaced to the reference pattern Y must be derived through calculation. That is, searching for a relationship describing the mapping between the image A and the image B in FIG. 1 can be regarded as searching for the most optimized combination of the reference pattern X and the reference pattern Y. In the present invention, this process is set as one objective function, and it is considered as an optimization problem in which the objective function is minimized. The optimization problem at this time is known as NP-hard problem as mentioned above.

In the present invention, this optimization problem is solved by using quantum computing, and an arrow representing a connection between respective points of interest (feature points) is modeled as a dipole in an image. In this case, the arrow indicates the direction and length between the respective points of interest (feature points), and may be represented by a vector.

To this end, (i) a term indicating a mismatch between a feature point of the image A and a corresponding position of the feature point of the image B, and (ii) a term indicating spatial consistency between neighboring points can be defined by measuring divergence of matches of the neighboring points.

By utilizing the physical model of quantum computing, the most optimized pattern X in image A and the most optimized pattern Y image B can be found at the same time. By modeling together with a combination of feature vectors between interest points (feature points) i, j, k in the image A and a combination of feature vectors between interest points (feature points) α, β, γ in the image B as a physical model for quantum computing, and by observing the physical model, optimized reference patterns X and Y can be obtained. The optimized state can be obtained by taking the state of the physical model when the energy of the physical model is in the ground state. For example, when a physical model is implemented in a black box capable of quantum computing and physical properties of a physical model in a black box are observed when the target state (ground state) is obtained through an adiabatic evolution process using a physical model, the optimized reference patterns X and Y can be obtained.

At this time, the physical model including the dipole physical characteristics such that the combination of vectors between interest points (feature points) of each image is described by the dipole model can be selected as the physical model in the black box.

In the image A in FIG. 1, a feature vector having the starting point of interest (feature point) i and the ending point of j can be expressed as {right arrow over (g_i,j)}. In this case, the relationship between the points of interest (feature points) i to j includes not only the translation between the feature points but also the difference of the local scale and the orientation. Feature vectors {right arrow over (g_i,j)} can be normalized for global translation, rotation, and scaling.

When the graph of the feature points i, j, k in image A of FIG. 1 is defined as G_A and the graph of the points of interest (feature points) α, β, γ in image B is defined as G_B, following Equation 1 defines the distance between feature points (iεG_A, αεG_B) derived from each image. If the number of feature points in image A is M, the G_A is expressed as a labeled graph having M nodes. If the number of feature points in image B is N, the G_B is represented by a labeled graph having N nodes. Where {right arrow over (f_i)} is the normalized feature vector for the i^th vertex of the G_A and {right arrow over (f_α)} is the normalized feature vector for the α^th vertex of G_B. The normalized feature vector may also be referred to as a local descriptor depending on the document. For example, the normalized feature vector may be a vector based on Gabor wavelets of varying scale and orientation with varying magnitude and direction around the point of interest. The edges of the graphs G_A and G_B represent the geometric relationships between the feature vectors. Similarity between the image A and the image B in FIG. 1 can be confirmed by finding the similarity between the two labeled graphs G_A and G_B.d(i,α)=d_feature({right arrow over (f)}_i,{right arrow over (f)}_α),  [Equation 1]

where d (i, α) is a scalar product between feature vectors {right arrow over (f)}_i and {right arrow over (f)}_α, and d (i, α) can be interpreted as a measure of the similarity of the correlated feature vectors and is a normalized value.

At this time, the combination between the point i obtained in image A and the point a obtained in image B can be defined as (i, α). If we define T_feature as a point-wise threshold that indicates whether the combination (i, α) is a potential match suitable for describing the pattern, the combination (i, α) satisfying d (i, α)>T_feature can be interpreted as a potential match suitable for describing the image pattern.

The conflict graph G_C can be generated from the graphs G_A and G_B as a measure for measuring the similarity of the graphical representations of image A and image B in FIG. 1. The conflict graph G_C can be generated by sequentially adding possible point-to-point combinations suitable for describing the image pattern, starting from the combination (i, α) having the largest d (i, α) value as the vertex V_iα of the conflict graph G_C. The process of generating a conflict graph G_C can be repeated until all suitable point-to-point combinations are included.

Edges (i,α;j,β) in the conflict graph G_C encode geometric consistency between {right arrow over (f)}_i and {right arrow over (f)}_j in the labelled graph G_A corresponding to image A in FIG. 1, and {right arrow over (f)}_α and {right arrow over (f)}_β in the labelled graph G_B corresponding to image B in FIG. 1.

For all vertex pairs (V_iα,V_jβ) in G_C where i≠j and α≠β we calculate the geometric consistency of the two pairs of interest points d(i,α,j,β)=d_geometric({right arrow over (g)}_i,j,{right arrow over (g)}_αβ) which is normalized.

The geometric consistency measures the geometric compatibility of the match pairs (i,α) and (j,β) as the residual differences in local displacement, scale and rotation assignment of the associated interest points after the changes due to global translation, rotation and scaling have been normalized. A pair (i,α) and (j,β) are not allowed to match if they are in geometric conflict, i.e., the residual effects are two large.

If d(i,α,j,β)<T_geometric, the pair (i,α) and (j,β) are considered in geometric conflict for a threshold T_geometric. We draw an edge in G_C between vertex pairs (V_iα,V_jβ) if i≠j and α≠β and they in geometric conflict. By this prescription, we draw the conflict graph with at most L vertices. The maximum independent set of the conflict graph is equivalent to the maximum common subgraph of unlabeled graphs G_A and G_B.

Finding the maximum independent set for the conflicting graph G_C can be translated as a quadratic unconstrained binary optimization problem defined by following Equation 2.

$\begin{array}{cc}{\stackrel{\rightharpoonup }{x}}_{\mathrm{opt}}=\arg \min \left\{\sum _{i\alpha <j\beta }^N{Q}_{i\alpha ,j\beta }{x}_{i\alpha }{x}_{j\beta }\right\},x_{i\alpha }\in \left\{0,1\right\},& \left[\mathrm{Equation}2\right]\end{array}$

where Q_iα,iα=−1 for all vertices and Q_iα,jβ=L when there is an edge between the pair (i,α) and (j,β).

The minimum energy configuration enforces x_iα=1 if and only if V_iα belongs to the maximum independent set and x_iα=0 otherwise. Equation 2 is a well know NP-hard problem which requires tremendous amount of computation time as L grows.

In the following, Equation 2 is modified to apply to adiabatic quantum computation for a quantum Ising Model.

Let X (where

$X=\left(\begin{array}{c}{x}_1\\ x_2\\ x_3\\ ⋮\\ x_N\end{array}\right))$ be a column vector of N Boolean variables and Q is a N×N matrix such that, then Equation 2 can be rewritten as the following Equation 3.X_opt=argmin X^†QX wherein x_iε(0,1}.  [Equation 3]

On the other hand, quantum mechanical Ising problem is formulated as the following Equation 4 by applying the relation of S=2X−1 to the Equation 3.S_opt=argmin{S^†JS+h^†S} where S_iε{−1,1}.  [Equation 4]

The S variables are called quantum-mechanical spin. This quantum Ising model can be solved by one particular model of quantum computation called adiabatic quantum computation (AQC).

In quantum mechanics, the spin states S=±1 are represented by orthogonal vectors in Hilbert space denoted as qubits. The two state of qubits are described by vectors as the following Equation 5.

$\begin{array}{cc}0\rangle =\left(\begin{array}{c}1\\ 0\end{array}\right)\mathrm{and}1\rangle =\left(\begin{array}{c}0\\ 1\end{array}\right).& \left[\mathrm{Equation}5\right]\end{array}$

The qubits can be extended to a vector by a linear combination called superposition, and this process is expressed as Equation 6.|ϕ=α|0+β|1 with |α|²+|β|²=1.  [Equation 6]

Larger quantum systems are constructed through tensor product of the individual qubit vector spaces, for example, as the following Equation 7.

$\begin{array}{cc}01\rangle =0\rangle \otimes 1\rangle =\left(\begin{array}{c}0\\ 1\\ 0\\ 0\end{array}\right).& \left[\mathrm{Equation}7\right]\end{array}$

Superpositions of N qubit states are also possible with the associated amplitudes representing the probability of observing the respective N spin state.

We also define single qubit operators can be defined as the following Equation 8.

$\begin{array}{cc}I=\left(\begin{array}{cc}1& 0\\ 0& 1\end{array}\right)\mathrm{and}{\sigma }^z=\left(\begin{array}{cc}1& 0\\ 0& -1\end{array}\right).& \left[\mathrm{Equation}8\right]\end{array}$

If the operator of Equation (8) is applied to the qubit vector of Equation (5), the following Equation (9) can be obtained.σ²|0=|0,σ^x|1=−|1.  [Equation 9]

On 2-qubit state, an operator σ^z⊗I extracts the classical spin of the first qubit, I⊗σ^z extracts the classical spin of the second qubit and σ^z⊗σ^z extracts the product of the two classical spins.

The quantum-mechanical Ising model on N spins is represented as 2^N×2^N Hamiltonian represented as the following Equation 10.

$\begin{array}{cc}{H}_I=\sum _{i,j}{J}_{\mathrm{ij}}{\sigma }_i^z{\sigma }_j^z+\sum _i{h}_i{\sigma }_i^z,& \left[\mathrm{Equation}10\right]\end{array}$

where σ_i^z is an operator σ^z acting on i^th qubit.

To initialize the quantum system, another kind of spin operator σ^x can be defined as shown in Equation 11.

$\begin{array}{cc}{\sigma }^x=\left(\begin{array}{cc}0& 1\\ 1& 0\end{array}\right)& \left[\mathrm{Equation}11\right]\end{array}$

The spin operator of Equation 11 can flip the state of the qubit.

At this time, the ground state Hamiltonian can be expressed by following Equation 12 using a spin operator acting on the i^th qubit.

$\begin{array}{cc}{H}_0=\Delta \sum _i{\sigma }_i^x.& \left[\mathrm{Equation}12\right]\end{array}$

The eigenstate of the ground state Hamiltonian of Equation 12 can be expressed as following Equation 13.

$\begin{array}{cc}{\Phi \rangle }_0={\otimes }_i\left(\frac{{0\rangle }_i-{1\rangle }_i}{\sqrt{2}}\right).& \left[\mathrm{Equation}13\right]\end{array}$

The time dependency of the quantum state of Equation 13 can be expressed by the Schrödinger equation of Equation 14.

$\begin{array}{cc}i\hslash \frac{\partial }{\partial t}\Psi \left(t\right)\rangle =H\left(t\right)\Psi \left(t\right)\rangle .& \left[\mathrm{Equation}14\right]\end{array}$

To solve the quantum mechanical eigen modeling adiabatically, the convex form of the adiabatic Hamiltonian is obtained by using the initial condition given by |Ψ(0)>=|Φ>0 at t=0.

$\begin{array}{cc}H\left(t\right)=\left(1-\frac{t}{T}\right)H_0+\frac{t}{T}{H}_I& \left[\mathrm{Equation}15\right]\end{array}$

At t=0, the quantum system has the lowest energy state. At this time, the lowest energy state can give equal probability for all classical configurations. On the other hand, at t=T, it is designed to cope with the quantum mechanical Ising model problem to solve the artificial visual problem.

In this way, the NP-hard problem, which is difficult to classically handle, can be solved through quantum mechanical adiabatic evolution of a given quantum system.

The globally lowest classical configuration obtained by adiabatic quantum mechanics (AQC) can be a solution to the second-order, non-definite binary optimization problem, which is a complex computational problem related to artificial view. It has been mathematically proved that quantum computing can provide exponential speed-up in solving NP-hard problems compared to classical methods.

The process of training the adiabatic quantum computing system to solve the secondary unrestricted binary optimization problem defined by Equation 2 begins with hardware training using a classification algorithm. The classification algorithm is expressed by the following equation 16.

$\begin{array}{cc}y=\mathrm{sign}\left(\sum _{i=1}^N{\omega }_i{h}_i\left(x\right)\right),& \left[\mathrm{Equation}16\right]\end{array}$

where xεR^M are input patters to be classified, yε{−1,1} is the output of the classifier, h_i:R^M|→{−1,1} are feature detectors, and ω_iε{0,1} is the weights to be optimized during training.

The training is achieved by solving the discrete optimization problem expressed as the following Equation 17.

$\left[\mathrm{Equation}17\right]{\underset{\_}{\omega }}_{\mathrm{opt}}=\underset{\omega }{\mathrm{argmin}}\underset{\_}{({\underset{\_}{\omega }}^{†}\left(\frac{1}{N^2}\sum _{s=1}^S\underset{\_}{h}\left({\underset{\_}{x}}_s\right){\underset{\_}{h}\left({\underset{\_}{x}}_s\right)}^{†}\right)\underset{\_}{\omega }+\underset{\_}{\omega }\left(\lambda I-2\sum _{s=1}^S\frac{\underset{\_}{h}\left({\underset{\_}{x}}_s\right)}{N}{y}_s\right)}).$

Here, the above Equation 17 is described for S-number of training samples {(x_s,y_s)|=1, 2, . . . , S}.

Hereinafter, a hardware implementation for solving the machine vision problem and a mathematical expression for describing the hardware will be described.

FIG. 2 is a conceptual diagram showing two quantum dots spin qubit.

FIG. 2 shows the electrons confined in quantum dots 121 with spin

${\stackrel{\rightharpoonup }{S}}_{1,2}=\frac{1}{2}{\mathrm{\hslash \sigma }}_{1,2}^z$ connected via tunnel junction 122. Those quantum dots 121 can be fabricated via gate-defined quantum dot structure. The Hamiltonian for this system is given by the following Equation 18.H=J₁₂(t)σ₁^zσ₂^z+h₁σ₁^z+h₂σ₂^z  [Equation 18]

In this equation, J₁₂ (t) is controlled by an external bias. The individual spin detection of the electrons is performed by the charge detection unit 123.

On the other hand, Hamiltonian regarding to N quantum dots can be described by the following equation 19.

$\begin{array}{cc}{H}_I=\sum _{i<j}{J}_{\mathrm{ij}}{\sigma }_i^z{\sigma }_j^z+\sum _i{h}_i{\sigma }_i^z& \left[\mathrm{Equation}19\right]\end{array}$

In Equation 19, J_ij and h_i are variables having a positive value, and are changed as the external power source is applied. The bias voltage applied to the gate can be appropriately adjusted to calculate the Hamiltonian of the quantum processing processor.

Hamiltonian at the initial (t=0) is J_ij=0 for all i and j. In Equation 15, the T value is defined as T=ℏ/ΔE, and ΔE is the interval between the initial ground state and the global minimum energy.

FIG. 3 is a conceptual diagram showing the arrangement of quantum dots of a quantum processing processor according to an exemplary embodiment of the present invention, and FIG. 4 is a block diagram showing a quantum-dot-based quantum mechanical machine vision system in accordance with an exemplary embodiment of the present invention.

Referring to FIG. 3 and FIG. 4, a quantum dot based quantum mechanical machine vision system 100 according to an exemplary embodiment of the present invention may include an image acquisition module 110, a quantum processing processor 120 and a memory unit 130.

The image acquisition module 110 acquires an image. The image acquisition module 110 may include, for example, a CCD camera.

The quantum processing processor 120 processes the image obtained from the image acquisition module 110.

The memory unit 130 stores data necessary for the computation of the quantum processing processor 120.

The quantum processing processor 120 obtains a first labeled graph connecting between feature points of the first image and a second labeled graph connecting feature points of the second image, generates a point-to-point combination by matching the feature points of the first image with the feature points the second image, generates a conflict graph by adding the largest point-to-point combination by comparing the point-to-point combinations with the threshold, generates non-constrained binary optimization equation for finding a maximum independent set of conflict graphs, converts the non-constrained binary optimization equation into Ising model of the quantum system, and calculates the Hamiltonian of Ising model based on the quantum dots to obtain solution of the non-constrained binary optimization equation. The quantum processing processor 120 may learn repeatedly the non-constrained binary optimization equation through machine learning.

For example, the quantum processing processor 120 may include quantum dots 121 arranged in a matrix shape as shown in FIG. 3.

At this time, neighboring quantum dots 121 with the shortest distance in a column direction or a row direction may be connected to each other through the tunnel junction 122.

The quantum processing processor 120 may further include a charge detection unit 123 disposed adjacent to the quantum dot 121 for detecting electrons of the quantum dot 121.

Meanwhile, in calculating the Hamiltonian of Ising model based on the quantum dots 121 to obtain solution of the non-constrained binary optimization equation, the Hamiltonian of Ising model can be calculated through adiabatic evolve. This process has been described in detail above, so redundant description is omitted.

As described above, according to the quantum dot-based quantum mechanical machine vision system and an arithmetic operation method of the present invention, the NP problem generated as the number of feature points increases is replaced by Hamiltonian using Ising model, so that it can be easily calculated using quantum dots arranged in a matrix shape.

It will be apparent to those skilled in the art that various modifications and variation may be made in the present invention without departing from the spirit or scope of the invention. Thus, it is intended that the present invention cover the modifications and variations of this invention provided they come within the scope of the appended claims and their equivalents.

Claims

1. A quantum mechanical arithmetic operation method based on quantum dots, the quantum mechanical arithmetic operation method being performed by a quantum processing processor, in a quantum system, the quantum mechanical arithmetic operation method comprising: obtaining a first labeled graph connecting between feature points of a first image and a second labeled graph connecting feature points of a second image;generating a point-to-point combination by matching the feature points of the first image with the feature points of the second image;generating a conflict graph by adding the largest point-to-point combination by comparing the point-to-point combinations with a threshold;generating a non-constrained binary optimization equation for finding a maximum independent set of conflict graphs;converting the non-constrained binary optimization equation for finding a maximum independent set of conflict graphs into an Ising model of the quantum system; andcalculating a Hamiltonian of the Ising model based on the quantum dots to obtain a solution of the non-constrained binary optimization equation.

2. The quantum mechanical arithmetic operation method of claim 1, wherein calculating the Hamiltonian of the Ising model based on the quantum dots to obtain the solution of the non-constrained binary optimization equation is performed by quantum dots arranged in a matrix shape.

3. The quantum mechanical arithmetic operation method of claim 2, wherein neighboring quantum dots with the shortest distance in a column direction or a row direction are connected to each other through a tunnel junction.

4. The quantum mechanical arithmetic operation method of claim 1, wherein the Hamiltonian of the Ising model is calculated through adiabatic evolve, in calculating the Hamiltonian of the Ising model based on the quantum dots to obtain the solution of the non-constrained binary optimization equation.

5. The quantum mechanical arithmetic operation method of claim 1, further comprising: repeatedly learning the non-constrained binary optimization equation through machine learning.

6. A quantum mechanical machine vision system comprising: an image acquisition module to acquire an image;a quantum processing processor to process the image obtained from the image acquisition module; anda memory unit to store data necessary for computation of the quantum processing processor;wherein the quantum processing processor, obtains a first labeled graph connecting between feature points of a first image and a second labeled graph connecting feature points of a second image,generates a point-to-point combination by matching the feature points of the first image with the feature points the second image,generates a conflict graph by adding the largest point-to-point combination by comparing the point-to-point combinations with a threshold,generates a non-constrained binary optimization equation for finding a maximum independent set of conflict graphs,converts the non-constrained binary optimization equation for finding a maximum independent set of conflict graphs into an Ising model of the quantum mechanical machine vision system, andcalculates a Hamiltonian of the Ising model based on quantum dots to obtain a solution of the non-constrained binary optimization equation.

7. The quantum mechanical machine vision system of claim 6, wherein the quantum processing processor comprises quantum dots arranged in a matrix shape.

8. The quantum mechanical machine vision system of claim 7, wherein neighboring quantum dots with the shortest distance in a column direction or a row direction are connected to each other through a tunnel junction.

9. The quantum mechanical machine vision system of claim 7, wherein the quantum processing processor further comprises a charge detection unit disposed adjacent to the quantum dots.

10. The quantum mechanical machine vision system of claim 6, wherein the quantum processing processor calculates the Hamiltonian of the Ising model through adiabatic evolve.