Dynamic visual sensor suitable for pulse neural network emergency detection
By introducing MOS tube control circuits into dynamic vision sensors, pulse signals are generated and combined with pulse neural networks, the problem of data redundancy and slow processing speed of sensors when processing complex dynamic scenarios is solved, and fast and accurate detection of emergencies is achieved.
Patent Information
- Application Number
- CN202411669929.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-21
- Publication Date
- 2025-02-25
AI Technical Summary
Dynamic vision sensors have problems of data redundancy and slow processing speed when dealing with complex dynamic scenarios, resulting in poor performance in emergency detection accuracy.
A dynamic vision sensor is designed, which generates pulse signals through the MOS tube control circuit, which is directly combined with the pulse neural network, improving processing speed and detection accuracy.
It realizes rapid and accurate detection and tracking of emergencies, and improves the processing capabilities of dynamic vision sensors and the detection accuracy of emergencies.
Smart Images
Figure CN119521025A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to a dynamic vision sensor, in particular to a dynamic vision sensor suitable for detecting sudden events of a pulse neural network. Background Art
[0002] Emergency detection is the use of computer vision technology to detect and identify sudden, unexpected or potentially dangerous events. This technology has extremely high application value in many fields, such as security monitoring, autonomous driving, industrial production, etc. By timely detecting and responding to emergencies, risks can be effectively reduced and losses can be reduced. Dynamic Vision Sensor (DVS) is a device that can capture changes in light intensity in a scene and generate corresponding electrical signals. It has the advantages of low power consumption and high sensitivity, but when processing complex dynamic scenes, there are problems such as data redundancy and slow processing speed. The pulse neural network has the characteristics of fast processing speed, low energy consumption, and sensitivity to dynamic input. If the signal of the dynamic vision sensor can be directly used in the pulse neural network, the processing capability of the sensor and the accuracy of emergency detection will be significantly improved. Summary of the invention
[0003] The purpose of the present invention is to enable the signal of a dynamic vision sensor to be directly used in a pulse neural network to solve the problems of low sensor processing power and low emergency event detection accuracy, and to propose a dynamic vision sensor suitable for emergency event detection in a pulse neural network.
[0004] The above purpose is achieved through the following technical solutions: A dynamic visual sensor suitable for pulse neural network sudden event detection, which comprises a honeycomb unit frame, a group of honeycomb units, a circular optical lens and a bottom plate; A honeycomb unit frame, used for mounting and fixing the honeycomb unit; A honeycomb unit, used to convert the sensed light signal into a pulse signal; A circular optical lens used to mount and fix some parts of the honeycomb unit; A bottom plate, used to mount and fix some components of the honeycomb unit; The honeycomb unit frame is cylindrical, and the circular optical lens is arranged in the middle and lower part of the honeycomb unit frame, the two are arranged coaxially, and the plane where the circular optical lens is located is parallel to the cross section of the honeycomb unit frame, and a group of honeycomb units are arranged in an array manner in the honeycomb unit frame; Each honeycomb unit includes a hexagonal pyramid optical lens, a group of event generators, and a gap layer; Hexagonal pyramid optical lens, used to select local light; An event generator, used to sense changes in light intensity; A part of the event generators constitutes an emergency detection and location layer, which is used to detect the occurrence of emergencies and locate the area where the events occur; Another part of the event generators constitutes the emergency event tracking layer, which is used to track the displacement changes of the emergency event movement; A gap layer, used to separate the space between the emergency detection and positioning layer and the emergency tracking layer, and used to separate the six sides of the hexagonal pyramid optical lens and the space between the honeycomb units; In the middle and upper part of the honeycomb unit, a hexagonal pyramid optical lens is arranged above the circular optical lens, an emergency detection and positioning layer composed of a part of event generators is arranged on the upper surface of the circular optical lens, below the emergency detection and positioning layer is a gap layer, below the gap layer is a bottom plate, and an emergency tracking layer composed of another part of event generators is arranged on the inner surface of the bottom plate; the emergency detection and positioning layer and the emergency tracking layer are in a corresponding upper and lower positional relationship.
[0005] Furthermore, the number of event generators in the emergency detection and positioning layer is 3, the centers of the 3 event generators are located on the same circumference, and the 3 event generators are distributed according to the angle between their centers and the centers of adjacent event generators being 120°, and the 3 event generators form output groups in 3 directions, each group having 12 outputs; There are 9 event generators in the emergency event tracking layer, of which 3 are set in the middle and the remaining 6 are set in the outer circle; the centers of the 3 event generators in the middle are located on the same circumference, and the 3 event generators are distributed at an angle of 120° between their centers and the centers of adjacent event generators, and the 3 event generators form output groups in 3 directions; the 6 event generators in the outer circle are located on the same circumference, and the 6 event generators are distributed at an angle of 30° with the adjacent event generators, and the 6 event generators form output groups in 6 directions; Each event generator includes a photosensitive diode D1, a capacitor C1, a transistor Q1, a transistor Q2, a transistor Q3, a transistor Q4, a transistor Q5, a transistor Q6, a transistor Q7, a transistor Q8 and an output circuit; The P pole of the photodiode D1 is connected to the D pole of the transistor Q3, and the N pole is connected to the S pole of the transistor Q1; the positive pole of the capacitor C1 is connected to the S pole of the transistor Q2, and the negative pole is grounded; the G pole of the transistor Q1 is connected to RES, the D pole is connected to VDD, and the S pole is connected to the G pole of the transistor Q7; the G pole of the transistor Q2 is connected to the S pole of the transistor Q6, the D pole is connected to the S pole of the transistor Q1, and the S pole is connected to the positive pole of the capacitor C1; the G pole of the transistor Q3 is connected to the S pole of the transistor Q6, the D pole is connected to the P pole of the photodiode D1, and the S pole is grounded; the G pole of the transistor Q4 is connected to the S pole of the transistor Q7, the D pole is connected to the S pole of the transistor Q1, and the S pole is connected to the D pole of the transistor Q5; the G pole of the transistor Q5 is connected to the S pole of the transistor Q6, the D pole is connected to the S pole of the transistor Q4, and the S pole is grounded; the transistor Q6 The G pole of transistor Q7 is connected to the S pole of transistor Q1, the D pole is connected to VDD, and the S pole is connected to Vout; the G pole of transistor Q7 is connected to the S pole of transistor Q1, the D pole is connected to the S pole of transistor Q6, and the S pole is connected to the G pole of transistor Q4; the G pole of transistor Q8 is connected to the S pole of transistor Q7, the D pole is connected to the S pole of transistor Q7, and the S pole is grounded.
[0006] Furthermore, the three event generators in the middle of the emergency event tracking layer are arranged clockwise at 0°, 120°, and 240° as the self-identification area of the current A honeycomb unit 2, and the three event generators in the self-identification area constitute an output group, and the six event generators in the outer circle are arranged clockwise at 0°, 60°, 120°, 180°, 240°, and 300°, respectively, as six temporary identification areas corresponding to adjacent honeycomb units, and the event generators in the temporary identification areas constitute an output group.
[0007] Furthermore, each honeycomb unit of the emergency event tracking layer has a total of 1 self-identification area and 9 temporary identification areas, and a total of 9 event generators. 112 honeycomb units 2 have a total of 784 self-identification areas, including 112 self-identification areas and 672 temporary identification areas.
[0008] The beneficial effects of the present invention are: The present invention is a dynamic vision sensor designed by using a MOS tube to control the local on and off of the circuit, thereby using a photodiode and a capacitor to sense the change in light intensity and generate a pulse signal. It is highly compatible with the pulse neural network, improves the processing speed and accuracy of the dynamic vision sensor, realizes the rapid and accurate detection and tracking of emergencies, and has a wide range of application prospects. Specifically, when light shines on the photodiode, the capacitor begins to discharge, causing the transistor to turn on, and the output circuit generates a gradually rising voltage signal. When the voltage signal reaches the threshold voltage, the transistor is cut off, and the output circuit generates a pulse signal, completing a pulse output. BRIEF DESCRIPTION OF THE DRAWINGS
[0009] Figure 1It is a schematic diagram of the overall structure of the present invention; the figure shows only one hexagonal pyramid optical lens 5, and in fact, a plurality of hexagonal pyramid optical lenses 5 are arranged in a dot matrix on the top of the honeycomb unit frame 1; Figure 2 It is a flow chart of sensor operation involved in the present invention; Figure 3 is a schematic cross-sectional view of a sensor according to the present invention; Figure 4 It is a schematic diagram of the appearance of the AA layer of the sensor involved in the present invention; Figure 5 It is a schematic diagram of the BB layer output group end of the sensor involved in the present invention; Figure 6 It is a schematic diagram of the CC layer unit identification area of the sensor involved in the present invention; Figure 7 It is a schematic diagram of the principle of the honeycomb unit of the sensor involved in the present invention; Figure 8 It is a structural schematic diagram of a honeycomb unit of a sensor involved in the present invention; Fig. 9 The present invention relates to an event generator circuit diagram; Fig.10 It is a circuit structure diagram of a honeycomb unit involved in the present invention; Fig.11 It is a schematic diagram of the CC layer unit circuit structure of the present invention; In the figure, 1 is a honeycomb unit frame, 2 is a honeycomb unit, 3 is a circular optical lens, 4 is a bottom plate, 5 is a hexagonal pyramid optical lens, 6 is an event generator, 7 is a gap layer, 8 is a self-identification area, and 9 is a temporary identification area. DETAILED DESCRIPTION
[0010] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention. Specific implementation method one:
[0012] A dynamic visual sensor suitable for detecting sudden events in a spiking neural network is provided in this embodiment. Figure 1 , Figure 3 , Fig.10 and Fig.11 As shown, it comprises a honeycomb unit frame 1, a group of honeycomb units 2, a circular optical lens 3 and a bottom plate 4; A honeycomb unit frame 1, used for mounting and fixing the honeycomb unit 2; Honeycomb unit 2, used for converting the sensed light signal into a pulse signal; A circular optical lens 3, used to install and fix some components of the honeycomb unit 2; The bottom plate 4 is used to install and fix some components of the honeycomb unit 2; The honeycomb unit frame 1 is cylindrical, and the circular optical lens 3 is arranged in the middle and lower part of the honeycomb unit frame 1. The two are arranged coaxially, and the plane where the circular optical lens 3 is located is parallel to the cross section of the honeycomb unit frame 1. A group of honeycomb units 2 are arranged in an array manner in the honeycomb unit frame 1; Each honeycomb unit 2 includes a hexagonal pyramid optical lens 5, a group of event generators 6, and a gap layer 7; Hexagonal pyramid optical lens 5, used for selecting local light; An event generator 6, used to sense changes in light intensity; A part of the event generators 6 constitutes an emergency event detection and location layer, which is used to detect the occurrence of emergency events and locate the event occurrence area, marked as BB layer; Another part of the event generators 6 constitutes the emergency event tracking layer, which is used to track the displacement changes of the emergency event movement, marked as CC layer; The top layer of the honeycomb unit frame 1 is marked as the AA layer to show the integrity of the entire device; The gap layer 7 is used to separate the space between the emergency detection and positioning layer and the emergency tracking layer, and is used to separate the six sides of the hexagonal pyramid optical lens 5 and the space between the honeycomb units 2; In the middle and upper part of the honeycomb unit 2, a hexagonal prism optical lens 5 is arranged above the circular optical lens 3, and a sudden event detection and positioning layer composed of a part of the event generator 6 is arranged on the upper surface of the circular optical lens 3, below the sudden event detection and positioning layer is a gap layer 7, below the gap layer 7 is a bottom plate 4, and a sudden event tracking layer composed of another part of the event generator 6 is arranged on the inner surface of the bottom plate 4; the sudden event detection and positioning layer and the sudden event tracking layer are in a corresponding positional relationship up and down. Specific implementation method 2:
[0014] Different from the first specific implementation mode, the present implementation mode is a dynamic visual sensor suitable for pulse neural network sudden event detection. The number of event generators 6 in the sudden event detection and positioning layer is 3. The centers of the 3 event generators 6 are located on the same circumference, and the 3 event generators 6 are distributed at an angle of 120° between their centers and the centers of adjacent event generators 6. The 3 event generators 6 form output groups in 3 directions, and each group has 12 outputs, forming a 12-12-12 output format. The three directional output groups are connected in series to form an event detection output group. Whenever a signal occurs, at least one of the directional output groups has a pulse, and the event detection output group has a pulse synchronously. The pulse signal of the event detection output group is used to prompt the occurrence of an event and activate the subsequent pulse neural network. The 36 outputs of the three directional output groups are simultaneously input into the pulse neural network to determine the event occurrence area. The number of event generators 6 in the emergency event tracking layer is 9, of which 3 event generators 6 are arranged in the middle, and the remaining 6 event generators 6 are arranged in the outer circle; the centers of the 3 event generators 6 in the middle are located on the same circumference, and the 3 event generators 6 are distributed at an angle of 120° between their centers and the centers of adjacent event generators 6, and the 3 event generators 6 form output groups in 3 directions; the 6 event generators 6 in the outer circle are located on the same circumference, and the 6 event generators 6 are distributed at an angle of 30° between their centers and the centers of adjacent event generators 6, and the 6 event generators 6 form output groups in 6 directions; Each event generator 6 includes a photodiode D1, a capacitor C1, a transistor Q1, a transistor Q2, a transistor Q3, a transistor Q4, a transistor Q5, a transistor Q6, a transistor Q7, a transistor Q8 and an output circuit; The P pole of the photodiode D1 is connected to the D pole of the transistor Q3, and the N pole is connected to the S pole of the transistor Q1; the positive pole of the capacitor C1 is connected to the S pole of the transistor Q2, and the negative pole is grounded; the G pole of the transistor Q1 is connected to RES, the D pole is connected to VDD, and the S pole is connected to the G pole of the transistor Q7; the G pole of the transistor Q2 is connected to the S pole of the transistor Q6, the D pole is connected to the S pole of the transistor Q1, and the S pole is connected to the positive pole of the capacitor C1; the G pole of the transistor Q3 is connected to the S pole of the transistor Q6, the D pole is connected to the P pole of the photodiode D1, and the S pole is grounded; the G pole of the transistor Q4 is connected to the S pole of the transistor Q7, the D pole is connected to the S pole of the transistor Q1, and the S pole is connected to the D pole of the transistor Q5; the G pole of the transistor Q5 is connected to the S pole of the transistor Q6, the D pole is connected to the S pole of the transistor Q4, and the S pole is grounded; the transistor Q6 The G pole of transistor Q7 is connected to the S pole of transistor Q1, the D pole is connected to VDD, and the S pole is connected to Vout; the G pole of transistor Q7 is connected to the S pole of transistor Q1, the D pole is connected to the S pole of transistor Q6, and the S pole is connected to the G pole of transistor Q4; the G pole of transistor Q8 is connected to the S pole of transistor Q7, the D pole is connected to the S pole of transistor Q7, and the S pole is grounded; The MOS tube controls the local on and off of the circuit, so as to use the photodiode and capacitor to sense the change of light intensity and generate a pulse signal. When light shines on the photosensitive diode D1, the capacitor C1 starts to discharge, and the event generator 6 circuit generates a pulse signal through the conduction and cutoff of the transistor; wherein, the principle of a pulse output is: when light shines on the photosensitive diode D1, the capacitor C1 starts to discharge, Vin decreases, the transistor Q6 starts to conduct, and current flows through Q7 and Q8; the current flowing through Q7 and Q8 increases first and then decreases, the Vout voltage gradually increases, Q4 and Q5 are turned on, and Vin continues to decrease. When Vin drops to the threshold voltage Vth, Q7 and Q8 are turned off, and there is no current in the Q6, Q7, and Q8 branches; Q4 is turned off, the output voltage Vout approaches the power supply voltage VDD, and Q2 and Q3 are turned off. After receiving the RES signal, Q1 is turned on, Vin is reset, Q7 is turned off, Vout is reset, Q5 is turned off, and a pulse output is completed. Specific implementation method three:
[0016] Different from the second specific implementation mode, in this implementation mode, a dynamic visual sensor suitable for pulse neural network emergency event detection has the same functions as the event generators 6 in the emergency event tracking layer. The three event generators 6 are arranged clockwise at 0°, 120°, and 240° as the self-identification zone 8 of the current A honeycomb unit 2. The three event generators 6 in the self-identification zone 8 form an output group. The six event generators 6 in the outer circle are arranged clockwise at 0°, 60°, 120°, 180°, 240°, and 300° as six temporary identification zones 9 corresponding to the six adjacent B, C, D, E, F, and G honeycomb units 2, respectively. The event generators 6 in the temporary identification zones 9 form an output group. At the same time, the area adjacent to the current A honeycomb unit 2 in the adjacent six B, C, D, E, F, and G honeycomb units 2 is the temporary identification zone 9, and the event generators 6 in the temporary identification zone 9 form an output group. When there is a continuous sudden event of movement, continuous pulse signals will appear in the adjacent honeycomb units 2, and the signal direction between two adjacent units is the adjacent unit's immediate identification area 9, the current unit's immediate identification area 9, the current unit's self-identification area 8, the current unit's immediate identification area 9, the adjacent unit's immediate identification area 9... Specific implementation method four:
[0018] Different from the third specific implementation mode, the present implementation mode is a dynamic visual sensor suitable for pulse neural network sudden event detection, and all components are installed in the honeycomb unit frame 1. 112 hexagonal prism honeycomb units 2 are arranged in the honeycomb unit frame 1, and each honeycomb unit 2 is from top to bottom: the top layer of the honeycomb unit frame, marked as AA layer, used to maintain the integrity of the entire device, the next layer is the upper surface of the hexagonal prism type optical lens 5; the next layer is the hexagonal prism type optical lens 5; the next layer is the lower surface of the hexagonal prism type optical lens 5; the next layer is the BB layer, which is a sudden event detection and positioning layer including 3 event generators 6; the next layer is the overall circular optical lens 3; the next layer is the gap layer 7; the next layer is the CC layer, which is a sudden event tracking layer including 9 event generators 6; the next layer is the overall circular black bottom plate 4.
[0019] Each honeycomb unit 2 of the emergency event tracking layer has a total of 1 self-identification area and 6 temporary identification areas 9, with a total of 9 event generators 6. 112 honeycomb units 2 have a total of 784 self-identification areas 8, including 112 self-identification areas 8 and 672 temporary identification areas 9. Example
[0020] 1. Manufacturing and Optimization 1. Production and assembly High-strength, corrosion-resistant alloy materials are selected as the manufacturing materials of the honeycomb unit frame. The honeycomb unit frame is manufactured using precision machining technology, and its dimensional accuracy and shape consistency are ensured to ensure that its structure is stable and easy to process. The honeycomb unit frame is manufactured using advanced additive manufacturing 3D printing technology.
[0021] Select optical glass or optical plastic with high light transmittance and low dispersion to manufacture hexagonal prism optical lens to ensure that light can be accurately focused. The prism lens is manufactured by injection molding or grinding process, and its optical performance is tested to ensure that it meets the design requirements. The hexagonal prism optical lens is installed in the honeycomb unit frame. Its structure and appearance are as follows Figure 3 , Figure 4 , Figure 7 and Figure 8 shown.
[0022] 2. Electronic component selection and circuit optimization High-performance, low-power photodiodes, capacitors and transistor electronic components are selected to improve the performance and stability of the visual sensor. The present invention improves the response speed and accuracy of the event generator circuit by optimizing the design.
[0023] 3. Sensor assembly and calibration Use surface mount technology (SMT) to mount electronic components such as photodiodes, capacitors, transistors, etc. Fig. 9The circuit diagram is precisely soldered on the event generator circuit board. Figure 2 , Figure 5 , Fig.10 and Fig.11 It is required to connect the circuits of the event generators in the BB layer and the CC layer to form a complete sensor circuit and perform calibration to ensure that the sensor can work normally.
[0024] Install the hexagonal pyramid optical lens in the honeycomb unit frame and adjust its position and angle. Check the correctness and stability of the circuit connection to ensure that the sensor can work properly. Debug the sensor and test its performance indicators such as response speed, accuracy and stability.
[0025] 2. Application and Verification 1. Expansion of application scenarios Apply sensors to a wider range of scenarios, such as deep space exploration, autonomous driving, etc.
[0026] 2. System integration and debugging Deeply integrate sensors with pulse neural networks to build a complete intelligent detection system. Debug and optimize the system to ensure good coordination between sensors and pulse neural networks, so as to achieve rapid and accurate detection of emergencies in different scenarios.
[0027] 3. Long-term performance verification In actual application scenarios, the sensor is verified for long-term performance to evaluate its stability and reliability. Based on the verification results, the sensor is maintained and updated as necessary to ensure its long-term stable operation.
[0028] The embodiments of the present invention disclose preferred embodiments, but are not limited thereto. A person skilled in the art can easily understand the spirit of the present invention based on the above embodiments and make different extensions and changes. However, as long as they do not deviate from the spirit of the present invention, they are all within the protection scope of the present invention.
Claims
1. A dynamic visual sensor suitable for detecting sudden events in a spiking neural network, comprising a honeycomb unit frame, a group of honeycomb units, a circular optical lens and a bottom plate; A honeycomb unit frame, used for mounting and fixing the honeycomb unit; A honeycomb unit, used to convert the sensed light signal into a pulse signal; A circular optical lens used to mount and fix some parts of the honeycomb unit; A bottom plate, used to mount and fix some components of the honeycomb unit; The honeycomb unit frame is cylindrical, and the circular optical lens is arranged in the middle and lower part of the honeycomb unit frame, the two are arranged coaxially, and the plane where the circular optical lens is located is parallel to the cross section of the honeycomb unit frame, and a group of honeycomb units are arranged in an array manner in the honeycomb unit frame; Each honeycomb unit includes a hexagonal pyramid optical lens, a group of event generators, and a gap layer; Hexagonal pyramid optical lens, used to select local light; An event generator, used to sense changes in light intensity; A part of the event generators constitutes an emergency detection and location layer, which is used to detect the occurrence of emergencies and locate the area where the events occur; Another part of the event generators constitutes the emergency event tracking layer, which is used to track the displacement changes of the emergency event movement; A gap layer, used to separate the space between the emergency detection and positioning layer and the emergency tracking layer, and used to separate the six sides of the hexagonal pyramid optical lens and the space between the honeycomb units; In the middle and upper part of the honeycomb unit, a hexagonal pyramid optical lens is arranged above the circular optical lens, an emergency detection and positioning layer composed of a part of event generators is arranged on the upper surface of the circular optical lens, below the emergency detection and positioning layer is a gap layer, below the gap layer is a bottom plate, and an emergency tracking layer composed of another part of event generators is arranged on the inner surface of the bottom plate; the emergency detection and positioning layer and the emergency tracking layer are in a corresponding upper and lower positional relationship.
2. A dynamic visual sensor suitable for pulse neural network sudden event detection according to claim 1, characterized in that: The number of event generators in the emergency detection and positioning layer is 3. The centers of the 3 event generators are located on the same circumference, and the 3 event generators are distributed at an angle of 120° between their centers and the centers of adjacent event generators. The 3 event generators form output groups in 3 directions, and each group has 12 outputs. There are 9 event generators in the emergency event tracking layer, of which 3 are set in the middle and the remaining 6 are set in the outer circle; the centers of the 3 event generators in the middle are located on the same circumference, and the 3 event generators are distributed at an angle of 120° between their centers and the centers of adjacent event generators, and the 3 event generators form output groups in 3 directions; the 6 event generators in the outer circle are located on the same circumference, and the 6 event generators are distributed at an angle of 30° with the adjacent event generators, and the 6 event generators form output groups in 6 directions; Each event generator includes a photosensitive diode D1, a capacitor C1, a transistor Q1, a transistor Q2, a transistor Q3, a transistor Q4, a transistor Q5, a transistor Q6, a transistor Q7, a transistor Q8 and an output circuit; The P pole of the photodiode D1 is connected to the D pole of the transistor Q3, and the N pole is connected to the S pole of the transistor Q1; the positive pole of the capacitor C1 is connected to the S pole of the transistor Q2, and the negative pole is grounded; the G pole of the transistor Q1 is connected to RES, the D pole is connected to VDD, and the S pole is connected to the G pole of the transistor Q7; the G pole of the transistor Q2 is connected to the S pole of the transistor Q6, the D pole is connected to the S pole of the transistor Q1, and the S pole is connected to the positive pole of the capacitor C1; the G pole of the transistor Q3 is connected to the S pole of the transistor Q6, the D pole is connected to the P pole of the photodiode D1, and the S pole is grounded; the G pole of the transistor Q4 is connected to the S pole of the transistor Q7, the D pole is connected to the S pole of the transistor Q1, and the S pole is connected to the D pole of the transistor Q5; The G pole of transistor Q5 is connected to the S pole of transistor Q6, the D pole is connected to the S pole of transistor Q4, and the S pole is grounded; the G pole of transistor Q6 is connected to the S pole of transistor Q1, the D pole is connected to VDD, and the S pole is connected to Vout; the G pole of transistor Q7 is connected to the S pole of transistor Q1, the D pole is connected to the S pole of transistor Q6, and the S pole is connected to the G pole of transistor Q4; The G electrode of the transistor Q8 is connected to the S electrode of the transistor Q7 , the D electrode is connected to the S electrode of the transistor Q7 , and the S electrode is grounded.
3. A dynamic visual sensor suitable for pulse neural network sudden event detection according to claim 2, characterized in that: The three event generators in the middle of the emergency event tracking layer are arranged clockwise at 0°, 120°, and 240° as the self-identification area of the current A honeycomb unit 2, and the three event generators in the self-identification area constitute an output group. The six event generators in the outer circle are arranged clockwise at 0°, 60°, 120°, 180°, 240°, and 300°, respectively, as six temporary identification areas corresponding to adjacent honeycomb units, and the event generators in the temporary identification areas constitute an output group.
4. A dynamic visual sensor suitable for pulse neural network sudden event detection according to claim 3, characterized in that: Each honeycomb unit of the emergency event tracking layer has a total of 1 self-identification area and 9 temporary identification areas, a total of 9 event generators; 112 honeycomb units 2 have a total of 784 self-identification areas, including 112 self-identification areas and 672 temporary identification areas.