Motion vector imaging method and system based on neuromorphic visual sensing array

Abstract

The invention belongs to the field of image perception and intelligent processing, and discloses a motion vector imaging method and system based on a neuromorphic visual sensing array. Acquiring continuous images of a moving target through the neuromorphic visual sensing array and outputting synaptic current levels of all pixel points; establishing a coordinate system in the Nth frame of image, selecting two monitoring points, and obtaining corresponding synaptic current level data; extracting time information corresponding to the current based on a synaptic current exponential decay formula; and calculating the moving speed and direction of the target according to the pixel coordinate difference and the time difference. The scheme provided by the invention overcomes the problems of high dependency on multi-frame image difference, large data redundancy, high power consumption and the like in the traditional image processing process, has the advantages of simple structure, high calculation efficiency, suitability for integration and the like, and is particularly suitable for various dynamic visual scenes with intelligent perception, automatic driving systems, target identification and tracking and the like.

Description

Technical Field

[0001] The present invention belongs to the field of image perception and intelligent processing, and relates to a motion vector imaging method and system based on a neuromorphic visual sensor array. Background Art

[0002] Traditional vision systems typically use CMOS image sensors paired with digital signal processing modules to capture and analyze objects in a scene. However, this architecture suffers from significant energy consumption, latency, and data transmission bottlenecks, resulting in poor object detection and tracking performance, particularly in dynamic, high-frequency, low-light, or complex background environments.

[0003] In recent years, neuromorphic vision sensors have garnered widespread attention due to their event-driven, low-power, and high dynamic range advantages. By simulating synaptic response characteristics, they can directly output current or voltage signals related to the input light signal and time series information, enabling pre-processing at the perception level. However, current motion target analysis based on these sensors primarily focuses on event detection, lacking a systematic method for extracting motion direction and velocity vectors.

[0004] The present invention proposes a motion vector extraction scheme that integrates the current response of a neuromorphic vision sensor and an image coordinate system, which has high accuracy and feasibility. Summary of the Invention

[0005] The purpose of the present invention is to overcome the shortcomings of the existing technology and provide a motion vector imaging method and system based on a neuromorphic visual sensor array. By constructing a time-series current change model based on the output data of the neuromorphic sensor and combining it with coordinate information, the target motion direction and speed can be accurately extracted.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] A first aspect of the present invention provides a motion vector imaging method based on a neuromorphic visual sensor array, comprising the following steps:

[0008] S1, acquires continuous images of moving targets through a neuromorphic visual sensor array and outputs the synaptic current level of each pixel;

[0009] S2, selecting the Nth frame image from the continuous multi-frame image, establishing a two-dimensional rectangular coordinate system at the center of the image, and selecting any two monitoring points separated by n pixels to obtain the corresponding synaptic current level data;

[0010] S3, extracting the time information corresponding to the current based on the exponential decay formula of the synaptic current;

[0011] S4, calculating the target movement speed and direction based on the pixel coordinate difference and the time difference.

[0012] Furthermore, the pixel interval between two monitoring points is a natural number n, where n>1.

[0013] Furthermore, S3 is specifically:

[0014] Using the exponential decay formula:

[0015]

[0016] Combined formula:

[0017]

[0018] Derived the imaging time difference t2-t1 between any two pixels,

[0019] The constant τ is an intrinsic parameter of the device, obtained through pre-experimental calibration. I represents the current value at time t, I max is the peak postsynaptic current.

[0020] Furthermore, the position difference: Time difference: t2-t1.

[0021]

[0022] s represents displacement, x1, x2, y1, y2 are the horizontal and vertical coordinates of two pixel points respectively, I t1 ,I t2 Represent the current levels read out by the two pixels respectively.

[0023] A second aspect of the present invention provides a motion vector imaging system based on a neuromorphic visual sensor array, comprising:

[0024] A neuromorphic visual sensor array, which is used to acquire continuous images of moving targets and output the synaptic current level at each pixel;

[0025] The image processing module establishes a coordinate system in the Nth frame image, selects two monitoring points, and obtains the corresponding synaptic current level data;

[0026] The current analysis module extracts the time information corresponding to the current based on the exponential decay formula of the synaptic current;

[0027] The vector calculation module calculates the target movement speed and direction based on the pixel coordinate difference and time difference.

[0028] The present invention improves the ability to recognize dynamic targets and perceive motion trajectories by simulating the integrated sensing, storage, and computing mechanism of biological vision. It can be widely used in autonomous driving, intelligent monitoring, human-computer interaction, and new robot vision systems.

[0029] Advantages and beneficial effects of the present invention:

[0030] 1. The computational logic of the present invention is simple and easy to implement in edge devices or brain-inspired chips;

[0031] 2. The present invention uses the physical response model of the device itself for timing modeling, which has strong noise resistance;

[0032] 3. The present invention can significantly reduce the dependence of traditional image processing on high frame rate and high-precision timers. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 Schematic diagram of the neuromorphic visual sensor array structure;

[0034] Figure 2 Schematic diagram of selecting monitoring points in continuous image frames;

[0035] Figure 3 is a diagram of the synaptic current decay model;

[0036] Figure 4 Flowchart of the target velocity extraction algorithm. DETAILED DESCRIPTION

[0037] The following is a further detailed description of the technical solution of the present invention through the accompanying drawings and embodiments. It should be noted that the following examples are not intended to limit the scope of protection of the present invention, and any improvements and variations made on the basis of the present invention are within the scope of protection of the present invention.

[0038] Example 1:

[0039] like Figure 1 As shown, the neuromorphic visual sensor array includes multiple photosensitive units, and each pixel can output a current signal I(t) that changes with the light intensity.

[0040] At a certain time t0, the image center coordinate system is established by reading the Nth frame image generated by the array. In the coordinate system, any two positions separated by n pixels are selected as monitoring points M1 (x1, y1) and M2 (x2, y2), and the corresponding current levels I are read respectively. t1 , I t2 .

[0041] Assuming the maximum initial current is I0, using the current decay model:

[0042]

[0043] The time difference between the two points can be obtained as Δt=t2-t1;

[0044] At the same time, the coordinate difference of the pixel point

[0045] Then the speed of the target on this path segment is:

[0046]

[0047] Repeating the above operations can construct a continuous velocity vector trajectory of the target for dynamic behavior prediction and path planning.

[0048] The above description is only a preferred embodiment of the present invention. It should be pointed out that a person skilled in the art can make several modifications and improvements without departing from the inventive concept, and these all fall within the scope of protection of the present invention.

Claims

1. A motion vector imaging method based on a neuromorphic visual sensor array, characterized in that: The steps include: S1, acquires continuous images of moving targets through a neuromorphic visual sensor array and outputs the synaptic current level of each pixel; S2, establish a coordinate system in the Nth frame image, select two monitoring points, and obtain the corresponding synaptic current level data; S3, extracting the time information corresponding to the current based on the exponential decay formula of the synaptic current; S4, calculating the target movement speed and direction based on the pixel coordinate difference and the time difference.

2. The motion vector imaging method based on the neuromorphic visual sensor array according to claim 1, characterized in that: The pixel interval between two monitoring points is a natural number n, where n>1.

3. The motion vector imaging method based on a neuromorphic visual sensor array according to claim 1, wherein S3 specifically comprises: Using the exponential decay formula: Combined formula: Derived the imaging time difference t2-t1 between any two pixels, The constant τ is the intrinsic parameter of the device, which is obtained through pre-experimental calibration. I represents the current value at time t, and I max is the peak postsynaptic current.

4. The motion vector imaging method based on the neuromorphic visual sensor array according to claim 1, characterized in that: The expression of the speed v is: s represents displacement, x1, x2, y1, y2 are the horizontal and vertical coordinates of two pixel points respectively, I t1 ,I t2 Represent the current levels read out by the two pixels respectively.

5. A motion vector imaging system based on a neuromorphic visual sensor array, characterized in that: include: A neuromorphic visual sensor array, which is used to acquire continuous images of moving targets and output the synaptic current level at each pixel; The image processing module establishes a coordinate system in the Nth frame image, selects two monitoring points, and obtains the corresponding synaptic current level data; The current analysis module extracts the time information corresponding to the current based on the exponential decay formula of the synaptic current; The vector calculation module calculates the target movement speed and direction based on the pixel coordinate difference and time difference.

Images