A billion-pixel compound eye camera integrated by 40-mesh single-lens dual-channel module

Abstract

The invention relates to the technical field of multi-view compound eye cameras, in particular to a billion-pixel compound eye camera integrated by a 40-view single-lens dual-channel module, which comprises a shell part, a plurality of layers of mounting plates are assembled in the shell part through connecting rods, and a plurality of groups of lens frames are fixedly mounted on each group of mounting plates through mounting frames; a lens shell is fixedly mounted at one end of the outer side of the lens frame, a plurality of groups of lenses are nested in the lens shell, a refraction sheet is mounted at the light output ends of the plurality of groups of lenses, sensors are mounted at the output ends of the two sides of the refraction sheet, and the two groups of sensors are used for respectively acquiring a color image and an infrared image; according to the billion-pixel compound eye camera integrated by the 40-mesh single-lens dual-channel module, the defects that in the prior art, two sets of independent cameras are needed to collect different image information, and consequently the occupied space is large are overcome.

Description

Technical Field

[0001] The present invention relates to the technical field of multi-eye compound eye cameras, and in particular to a billion-pixel compound eye camera integrated by a 40-eye single-lens dual-channel module. Background Art

[0002] A multi-eye compound eye camera refers to a camera with multiple groups of lenses regularly arranged in the hope of obtaining a larger viewing angle. For example, a highway compound eye camera monitoring system based on sound source positioning proposed in Chinese patent application number: "201821570649.7" is a multi-eye compound eye camera.

[0003] In the existing multi-eye compound eye cameras, each camera has only one set of CMOS modules. When color information and infrared light information need to be collected, two sets of cameras are required to cooperate to complete the task, which reduces the number of cameras in different directions that can be accommodated in the same space. In addition, there is an angle between the two sets of cameras, and the image collection angles are different, which makes the image prone to problems such as ghosting.

[0004] Based on the above problems, a new structure in which each camera has a dual CMOS module is proposed below to improve the image quality while reducing the space occupation. Summary of the invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art that two sets of independent cameras are needed to collect different image information respectively, resulting in a large space occupation, and a billion-pixel compound eye camera integrated by a 40-mesh single-lens dual-channel module is proposed.

[0006] In order to achieve the above object, the present invention adopts the following technical solutions: A billion-pixel compound eye camera integrated with a 40-mesh single-lens dual-channel module is designed, comprising a housing, wherein a plurality of mounting plates are assembled inside the housing through a connecting rod, and each group of mounting plates is fixedly mounted with a plurality of lens frames through a mounting frame; A lens shell is fixedly installed at one end of the outer side of the lens frame, and multiple groups of lenses are nested inside the lens shell. Refraction plates are installed at the light output ends of the multiple groups of lenses. The refraction plates can copy the light and refract it to both sides. Sensors are installed at the output ends on both sides of the refraction plates. The two groups of sensors collect color images and infrared images respectively.

[0007] Preferably, a base member is rotatably mounted at the lower end of the outer shell member, a gear ring is mounted at the connection between the outer shell member and the base member, a motor is mounted in the base member, a gear is fixedly mounted on the output end of the motor, and the gear at the output end of the motor meshes with the gear ring for transmission.

[0008] Preferably, an expansion part is fixedly mounted on the upper end of the outer shell, and the expansion part includes an infrared rangefinder and a millimeter wave radar.

[0009] Preferably, the mounting plate is vertically arranged with four layers, and ten sets of lens holders and their combined components are horizontally installed in each layer of the mounting plate and arranged in a specification of 4×10.

[0010] Preferably, two adjacent layers of the lens frames and their assembly components are arranged in a staggered manner.

[0011] Preferably, the mounting plate and the mounting frame are connected via a driving guide rail, so that multiple groups of lens frames and their combined components located on the same layer can be unfolded to angles of 100°, 180°, and 220°.

[0012] Preferably, a circuit board is mounted on the upper end surface of the mounting plate, the circuit board is electrically connected to the plurality of groups of sensors, and a heat sink is integrally formed on the outer end surface of the mounting plate.

[0013] Preferably, the two groups of sensors respectively collect color images and infrared images, and the images collected by the two groups of sensors are spliced ​​through AI image fusion. The images after the AI ​​image fusion and the original color images and infrared images are optimized through the LSP algorithm to improve the display effect of the light and dark parts.

[0014] Preferably, AI image fusion marks images frame by frame, confirms and aligns the timeline of the image, combines the timeline with the frame-by-frame marked images so that each frame can be accurately aligned, and finally fuses the aligned images. When marking images frame by frame, the corners of each frame are marked in a cycle according to seven groups: red, orange, yellow, green, cyan, blue, and purple.

[0015] Preferably, the LSP algorithm optimization first obtains the information of the initial image, and combines the obtained information with the image after AI image fusion to achieve fused image enhancement, so as to correct the light errors in the bright and dark parts generated during the AI ​​image fusion.

[0016] The invention proposes a billion-pixel compound eye camera integrated with a 40-mesh single-lens dual-channel module, which has the following beneficial effects: Through the refraction plate, the image can be refracted to both sides, so that the two groups of sensors in one lens housing can synchronously receive the same image, and the two groups of sensors have different focuses. Then the two groups of images are fused to improve the image effect. Compared with the existing two groups of cameras that focus on collection respectively, this case can be achieved by only one group of cameras, so as to reduce the number of cameras. It is expected that more cameras with effective angles can be accommodated in a unit space, thereby realizing the layout of 40-mesh cameras. At the same time, the two groups of sensors in the same lens housing collect the same image angle, thereby reducing the difficulty of post-processing and improving the image effect. Through the proposed frame-by-frame labeling, the images can be better aligned frame by frame, the fusion difficulty can be reduced, and the optimization difficulty of the LSP algorithm can be reduced, that is, the original image information can be used to complete the bright and dark parts of the fused image. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 A schematic diagram of the three-dimensional structure of the main body; Figure 2 This is a schematic diagram of the layout structure of the lens housing; Figure 3 This is a schematic diagram of the explosion disassembly structure of the lens housing; Figure 4 Schematic diagram of the three-dimensional structure of the lens housing; Figure 5 It is a schematic diagram of the main flow of image fusion processing; Figure 6 for Figure 5 Schematic diagram of the process of AI image fusion; Figure 7 This is a flowchart of the LSP algorithm optimization process in Figure 5.

[0018] In the figure: 1. Shell component; 101. Base component; 102. Extension part; 2. Mounting plate; 201. Connecting rod; 202. Heat sink; 203. Mounting frame; 3. Lens holder; 301. Lens housing; 302. Lens; 303. Refractor; 304. Sensor. DETAILED DESCRIPTION

[0019] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0020] Reference Figure 1-4 , here is an example of the implementation of the main structure: A billion-pixel compound eye camera integrated with a 40-mesh single-lens dual-channel module includes a housing 1 for peripheral protection and installation. The interior of the housing 1 is assembled with a multi-layer mounting plate 2 through a connecting rod 201. Each group of mounting plates 2 is fixedly mounted with a plurality of lens frames 3 through a mounting frame 203. A lens shell 301 is fixedly mounted on one end of the outer side of the lens frame 3. A plurality of lens groups 302 are nested inside the lens shell 301. In order to enable a single lens to output two groups of images synchronously, reference Figure 3A refraction plate 303 is installed at the light output end of the multiple groups of lenses 302. The refraction plate 303 can copy the light and refract it to both sides to form a single-in and double-out effect. Sensors 304 are installed at the output ends on both sides of the refraction plate 303. The sensors 304 are CMOS sensors 304, and the two groups of sensors 304 collect color images and infrared images respectively. The above content describes a compound eye camera structure formed by a plurality of lens frames 3 and an array of their combined components; each lens housing 301 divides the light beam into two parts through a refraction plate 303 and is provided with two groups of sensors 304 for sensing and receiving light, and the two groups of sensors 304 collect the color and bright details of the image respectively, so that the final imaging effect is richer and clearer.

[0021] In order to enable the outer shell 1 to adjust its direction, a base member 101 is rotatably installed at the lower end of the outer shell 1, a gear ring is installed at the connection between the outer shell 1 and the base member 101, a motor is installed in the base member 101, and a gear is fixedly sleeved at the output end of the motor. When the motor is started, the gear at the output end of the motor meshes with the gear ring to achieve a driving effect.

[0022] In order to improve the detection effect, an expansion part 102 is fixedly installed on the upper end of the outer shell 1. The expansion part 102 includes an infrared rangefinder and a millimeter wave radar, so that the distance of the object can be detected and the accuracy of the object contour detection can be improved.

[0023] refer to Figure 2 In the figure, the mounting plate 2 is vertically arranged with four layers, and ten sets of lens frames 3 and their combined components are horizontally installed in each layer of the mounting plate 2, so that the compound eye camera is arranged in a specification of 4×10, and the two adjacent layers of lens frames 3 and their combined components are staggered to improve the image acquisition effect of the compound eye camera; Furthermore, the mounting plate 2 is connected to the mounting frame 203 via a driving rail, so that multiple groups of lens frames 3 and their combined components located on the same layer can be controlled to unfold to angles of 100°, 180°, 220°, etc., to achieve instant adjustment.

[0024] A circuit board is mounted on the upper end surface of the mounting plate 2, and the circuit board is electrically connected to the multiple sets of sensors 304 to play a master control role. A heat sink 202 is integrally formed on the outer end surface of the mounting plate 2 to improve the heat dissipation effect of the circuit board.

[0025] Reference Figure 5-7 , here is an example of how to implement image fusion and optimization: The two groups of sensors 304 collect color images and infrared images respectively; then the images collected by the two groups of sensors 304 are spliced ​​through AI image fusion; then the images after AI image fusion and splicing as well as the original color images and infrared images are optimized through the LSP algorithm to improve the display effect of light and dark parts, and finally the image is output.

[0026] AI image fusion uses frame-by-frame image marking, and then confirms and aligns the time axis of the image. Confirming and aligning the time axis of the image can determine the approximate range, and combining the range with the frame-by-frame marked image allows each frame to be accurately aligned, so as to further confirm the image alignment through the time axis and with the marking, and finally fuse the aligned images. The above approach enables better frame-by-frame alignment of images, and can be quickly identified using markings, reducing the hardware burden generated by the AI ​​algorithm; Furthermore, when marking images frame by frame, the corners of each frame can be marked in a cycle according to seven groups: red, orange, yellow, green, cyan, blue, and purple. Since most of the existing surveillance images are collected in 14-24 frames, there are at least two groups of cyclic marks per second. When the time axis is aligned to within 0.5S, the marking method can be used to assist in aligning the image frames. The comparison only needs to identify whether the mark colors are the same, without comparing the entire picture, to achieve rapid error correction in a small range.

[0027] The LSP algorithm optimization first obtains the information of the initial image, such as the bright and dark areas of infrared images and the color information of color images, and then combines this information with the image after AI image fusion to achieve fusion image enhancement, so as to correct the light errors in the bright and dark areas caused by AI image fusion. Furthermore, LSP can also mark images frame by frame, so that the picture frames can be aligned more conveniently when performing LSP algorithm optimization, thereby improving the optimization effect.

[0028] Based on the above implementation examples, when used: S1, the light is copied into two parts through the refraction plate 303 and transmitted to two sets of sensors 304 respectively, and the two sets of sensors 304 receive the color image and the infrared image respectively; S2, transmitting the color images and infrared images received by the two sets of sensors 304 to AI image fusion and LSP algorithm optimization respectively; S3, AI image fusion fuses the two sets of images and passes them to LSP algorithm optimization; S4. The LSP algorithm optimizes the light and dark features of the original image as an optimization benchmark, and then performs algorithmic enhancement of the light and dark features of the image after AI image fusion based on this benchmark.

[0029] By marking images frame by frame, the difficulty of AI image fusion and LSP algorithm-optimized picture alignment is reduced, thereby reducing the demand for algorithms and hardware.

[0030] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A billion-pixel compound eye camera integrated with a 40-mesh single-lens dual-channel module, characterized by: It comprises a housing (1), wherein a plurality of mounting plates (2) are assembled inside the housing (1) via a connecting rod (201), and each group of mounting plates (2) is fixedly mounted with a plurality of lens frames (3) via a mounting frame (203); A lens housing (301) is fixedly mounted on one end of the outer side of the lens frame (3), a plurality of groups of lenses (302) are nested inside the lens housing (301), refraction plates (303) are mounted at the light output ends of the plurality of groups of lenses (302), the refraction plates (303) are capable of replicating the light and refracting it to two sides respectively, sensors (304) are mounted at the output ends on both sides of the refraction plates (303), and the two groups of sensors (304) collect color images and infrared images respectively.

2. The billion-pixel compound eye camera integrated with a 40-mesh single-lens dual-channel module according to claim 1, characterized in that: A base member (101) is rotatably mounted on the lower end of the housing member (1), a gear ring is mounted at the connection between the housing member (1) and the base member (101), a motor is mounted in the base member (101), a gear is fixedly mounted on the output end of the motor, and the gear at the output end of the motor meshes with the gear ring for transmission.

3. The billion-pixel compound eye camera integrated with a 40-mesh single-lens dual-channel module according to claim 1, characterized in that: An expansion part (102) is fixedly mounted on the upper end of the outer shell (1), and the expansion part (102) comprises an infrared rangefinder and a millimeter wave radar.

4. The billion-pixel compound eye camera integrated with a 40-mesh single-lens dual-channel module according to claim 1, characterized in that: The mounting plate (2) is vertically arranged with four layers, and ten groups of lens frames (3) and their combined components are horizontally installed in each layer of the mounting plate (2) and arranged in a 4×10 specification.

5. The billion-pixel compound eye camera integrated with a 40-mesh single-lens dual-channel module according to claim 4, characterized in that: The lens frames (3) and their combined components in two adjacent layers are arranged in a staggered manner.

6. The billion-pixel compound eye camera integrated with a 40-mesh single-lens dual-channel module according to claim 1, characterized in that: The mounting plate (2) and the mounting frame (203) are connected via a driving guide rail, so that multiple groups of lens frames (3) and their combined components located on the same layer can be unfolded to angles of 100°, 180°, and 220°.

7. The billion-pixel compound eye camera integrated with a 40-mesh single-lens dual-channel module according to claim 1, characterized in that: A circuit board is mounted on the upper end surface of the mounting plate (2), the circuit board is electrically connected to the plurality of groups of sensors (304), and a heat sink (202) is integrally formed on the outer end surface of the mounting plate (2).

8. The billion-pixel compound eye camera integrated with a 40-mesh single-lens dual-channel module according to claim 1, characterized in that: The two groups of sensors (304) respectively collect color images and infrared images. The images collected by the two groups of sensors (304) are spliced ​​through AI image fusion. The images obtained by the AI ​​image fusion and the original color images and infrared images are optimized through the LSP algorithm to improve the display effect of light and dark parts.

9. The billion-pixel compound eye camera integrated with a 40-mesh single-lens dual-channel module according to claim 8, characterized in that: AI image fusion marks images frame by frame, confirms and aligns the timeline of the image, combines the timeline with the frame-by-frame marked images so that each frame can be accurately aligned, and finally fuses the aligned images. When marking images frame by frame, the corners of each frame are marked in a cycle according to seven groups: red, orange, yellow, green, cyan, blue, and purple.

10. The billion-pixel compound eye camera integrated with a 40-mesh single-lens dual-channel module according to claim 8, characterized in that: The LSP algorithm optimization first obtains the information of the initial image, and then combines the obtained information with the image after AI image fusion to achieve fused image enhancement, so as to correct the light errors in the bright and dark parts generated during AI image fusion.

Images