Large-view-field high-resolution compound eye camera array system based on bionic human eye vision

Abstract

The invention provides a large-view-field high-resolution compound eye camera array system based on bionic human eye vision. The system is composed of a curved surface support and an industrial camera array arranged on the curved surface support. Reverse extension lines of optical axes of all the cameras intersect at the same point to form a concave surface structure similar to the retina, and the fixed view field overlapping rate of the adjacent cameras is kept. The industrial camera array is used for collecting a target scene image, large-view-field information is obtained by fusing and splicing all camera images, and a high-resolution image is obtained by conducting super-resolution reconstruction on a plurality of related images in a view field center interested area. The system provides three working modes, namely a low-resolution mode, a conventional-resolution mode and a super-resolution mode, which can be switched by adjusting camera parameters. Finally, the system can perceive the dynamic change of the edge of the field of view in a high-frame-rate and low-resolution mode, magnify the details in the center of the field of view in a super-resolution mode, realize efficient non-uniform imaging, and optimize the information acquisition and processing efficiency and the structure compactness.

Description

Technical Field

[0001] This invention belongs to the field of optoelectronic imaging technology, specifically relating to a large field-of-view, high-resolution compound eye camera array system based on bionic human eye vision, which is suitable for scenarios requiring large field-of-view, high-resolution imaging, such as target surveillance, traffic monitoring, military reconnaissance, and intelligent security. Background Technology

[0002] Target surveillance systems based on photoelectric imaging technology have been widely used in traffic monitoring, public safety, counter-terrorism and bomb disposal, and military reconnaissance due to their excellent visual information acquisition capabilities, making outstanding contributions to socio-economic development, scientific and technological progress, and national security. However, most traditional target surveillance systems consist of a single-camera platform, which is usually limited by the perfect imaging theory, making it impossible to simultaneously achieve a large field of view, high resolution, and fast imaging. The system's effective distance and range are relatively limited, reducing the image quality of the equipment.

[0003] With the rapid development of imaging technology in recent years, compound eye camera array systems have become one of the main solutions to this contradiction. Array-type multi-camera systems can stitch together the fields of view of multiple cameras by using mathematical models and algorithms to reconstruct and obtain images with a large field of view and high resolution. They are gradually being applied to light field information acquisition and continuous surveillance of key areas by various UAV platforms and reconnaissance platforms. However, compound eye camera array systems also have certain drawbacks: First, the cameras in the system usually operate in parallel, resulting in a large amount of data being generated simultaneously. The time required for continuous reading, writing, transmission, storage, and processing of data will increase significantly. Second, the complex structure and massive amount of data require the system to have a large structure and a large number of data computing units, resulting in very high operating and maintenance costs.

[0004] The core reason for the aforementioned problems lies in the fact that traditional compound eye camera array systems always acquire information about the target scene in a uniform manner, and allocate the same system resources to each pixel during the data processing stage. However, in real-world scenes, the distribution of targets is usually sparse and non-uniform, and the region of interest often exists only in the center of the field of view. Therefore, the system always acquires a large amount of redundant information that is not of interest, reducing efficiency and causing a significant waste of resource allocation. Current solutions for optimizing compound eye camera array systems mostly focus on improving the system's resolution and field of view. Some increase the number of cameras to improve resolution or use curved lens arrays to optimize the field of view, but they do not address the fundamental contradiction of data redundancy at the hardware structure level. The system still suffers from heavy data processing load and high hardware complexity. Summary of the Invention

[0005] To address the problems of complex structure, large amount of redundant information, and high processing time in compound eye camera array systems, this invention proposes a large field-of-view, high-resolution compound eye camera array system based on bionic human eye vision. The system consists of a curved support and an industrial camera array, wherein the industrial camera array comprises an industrial automation lens (FA lens) and an industrial area array camera, as shown in the attached figure. Figure 1 As shown in the attached diagram, an industrial camera array is arranged on a curved support. The backward extensions of the optical axes of each camera converge at the same focal point, forming a concave imaging structure similar to the human retina. Adjacent cameras maintain a fixed field-of-view overlap rate. Figure 2 As shown, a curved support is used to support the camera array, ensuring that the backward extensions of the optical axes of all cameras intersect at the same point, while also ensuring a stable overlap rate of the fields of view of adjacent cameras. The industrial camera array is used to acquire images of the target scene. Large field-of-view information is obtained by fusing and stitching the images captured by each camera, and high-resolution images are obtained by super-resolution reconstruction of multiple highly correlated images of the region of interest in the center of the field of view. The system offers three operating modes: low-resolution mode, normal resolution mode, and super-resolution mode. Low-resolution mode is suitable for monitoring dynamic changes across the entire field of view and quickly capturing moving objects in the scene; normal resolution mode is used for target recognition and tracking under normal circumstances; and super-resolution mode provides high-resolution detail information for specific targets in the region of interest in the center of the field of view, meeting the needs for precise analysis of key areas and targets. Different modes can be selected to meet different monitoring priorities by adjusting the parameters of the industrial cameras. Ultimately, the system can perceive dynamic changes at the edge of the field of view in a high frame rate, low resolution mode, and magnify local details of the region of interest in the center of the field of view in a super-resolution mode, thereby achieving non-uniform imaging, taking into account both a large field of view and high resolution information acquisition, significantly improving information acquisition and processing efficiency, and optimizing structural compactness.

[0006] The curved support has a concave structure, causing the backward extensions of the optical axes of the n×n cameras arranged on it to converge at the same point, forming a concave imaging layout similar to the human retina. This stabilizes the overlap rate of the fields of view of adjacent cameras, providing a hardware foundation for image stitching. The curvature of the surface naturally optimizes edge field-of-view aberrations, ensuring stitching stability and avoiding edge field-of-view aberrations and stitching misalignment caused by planar arrays.

[0007] Preferably, the curved support is precision machined to control the spatial position and orientation deviation of each camera mounted on the support to within a range of no more than 1°, thereby better ensuring the stability of the field of view overlap rate of adjacent cameras. The use of lightweight alloy materials reduces the hardware volume and weight compared to traditional planar arrays.

[0008] The industrial camera array consists of industrial automation lenses and industrial area array cameras. The system contains a total of n×n cameras, arranged on a curved support, forming a concave imaging structure similar to the human retina. (See attached image.) Figure 2 As shown, the field of view of a single camera is A×B, and there is α% overlap between adjacent cameras. This is used to ensure the accuracy of camera calibration and image stitching processes, and to provide sufficient reference pixels for super-resolution reconstruction of the region of interest.

[0009] Preferably, the industrial camera array uses a small field-of-view lens for each camera, which can achieve high imaging spatial resolution while avoiding large off-axis aberrations.

[0010] Furthermore, to obtain a wide field-of-view panoramic image, a weighted fusion algorithm is used to fuse corresponding pixels in overlapping areas according to a set weight ratio for images captured by two adjacent cameras. For non-overlapping areas, a stitching operation is performed, as shown in the attached diagram. Figure 3 As shown, the field of view of each camera in the system occupies 4 / n of the total field of view. 2 The overlap ratio of the fields of view of adjacent cameras is α%. This system is set up to achieve full field of view target information acquisition. The final full field of view is C×D, where the horizontal field of view C is obtained from formula (1), and the vertical field of view D is obtained from formula (2): The vertical field of view D is obtained from formula (2):

[0011] C = [A×nA×α%×(n-1)] (1)

[0012] D = [B×nB×α%×(n-1)] (2)

[0013] Furthermore, to simulate the unique attention mechanism of the human eye, a super-resolution reconstruction algorithm is used to reconstruct local details in the central region of interest (ROI) of the visual field. The central ROI of the system's full field of view is covered by the fields of view of m cameras, which is used to simulate the densely distributed photoreceptor cells in the macular region of the human retina, achieving m-fold resolution enhancement. The peripheral ROI of the system is covered by the fields of view of a small number of cameras, aiming to complete dynamic scene monitoring with lower data acquisition costs.

[0014] The conventional resolution mode is used for target recognition and tracking under normal circumstances, and is suitable for scenarios where the target position is relatively fixed or the target is moving at a low speed, as shown in the attached diagram. Figure 4 As shown.

[0015] Preferably, a small field-of-view lens with adjustable focal length is used. By adjusting the focal length in real time, the target is always kept within the depth of field, which can improve the recognition accuracy.

[0016] The low-resolution mode is used for dynamic monitoring of the entire field of view. By acquiring information through low resolution and high frame rate, it improves the accuracy of dynamic perception and reduces the data acquisition and transmission load. When the system switches to low-resolution mode, it uses the Binning mode of the industrial camera to accumulate the signal current of adjacent p pixels and read it out at once, thus achieving low-resolution, high-frame-rate image output with a resolution of 1 / p and a frame rate p times that of the original camera.

[0017] The super-resolution mode is used to magnify the details of an image in a region of interest. By fusing images from adjacent cameras, a super-resolution reconstruction algorithm is used to overcome the physical resolution limits of the cameras, providing clear local detail information of the target of interest. The central region of the field of view is covered by the overlapping fields of view of m cameras. Through a multi-camera synchronous exposure hardware design, multi-angle information is acquired, thereby enhancing image resolution and improving the detail acquisition capability by m times compared to a single camera. The details of the image after super-resolution reconstruction are clearly reproduced.

[0018] This invention proposes a large field-of-view, high-resolution compound eye camera array system based on biomimetic human eye vision. Combining non-uniform imaging theory, by designing and adjusting the field-of-view overlap and shooting modes between adjacent cameras, it can acquire images with different sampling frame rates and resolutions across the entire field of view. This enables high-resolution detail magnification of the central region of interest (ROI) and dynamic perception of the edge regions of the field of view. The system uses a curved support to support the camera array, ensuring that the backward extensions of the optical axes of all cameras intersect at the same point, stabilizing the overlap rate of the adjacent camera fields of view. An industrial camera array is used to acquire images of the target scene. Large field-of-view information is obtained by fusing and stitching the images captured by each camera. High-resolution images are obtained by super-resolution reconstruction of multiple highly correlated images of the central ROI. Ultimately, the system can perceive dynamic changes at the edge of the field of view in a high frame rate, low-resolution mode, and perform local detail magnification of the central ROI in a super-resolution mode combined with image super-resolution algorithms. This achieves non-uniform imaging, balancing large field of view and high-resolution information acquisition, significantly improving information acquisition and processing efficiency, and optimizing structural compactness. Attached Figure Description

[0019] Appendix Figure 1 This is a diagram of the device for the large field-of-view, high-resolution compound eye camera array system based on bionic human eye vision proposed in this invention.

[0020] Appendix Figure 2 This is a schematic diagram of a one-dimensional array of the system of the present invention.

[0021] Appendix Figure 3 This is a field-of-view distribution diagram of the system of the present invention.

[0022] Appendix Figure 4 This is a schematic diagram of the working mode of the system of the present invention.

[0023] The figure labels in the above figures are:

[0024] 100 Two-dimensional n×n industrial camera array, 101 One-dimensional 1×n industrial camera array, 102 Industrial automation lens, 103 Industrial area array camera, 2 Curved support, 3 Optical axis intersection, 400 Full field of view, 401 Field of view of a single camera, 500 Central region of interest in the field of view, 501 Edge field of view, 6 Panoramic image, 7 Normal resolution mode, 8 Low resolution mode, 9 Super resolution mode.

[0025] It should be understood that the above figures are only schematic and are not drawn to scale. Detailed Implementation

[0026] The following detailed description of a typical embodiment of the large field-of-view, high-resolution compound eye camera array system based on bionic human eye vision provides further specific information about the present invention. It is important to note that the following embodiments are for illustrative purposes only and should not be construed as limiting the scope of protection of the present invention. Any non-essential improvements and adjustments made to the present invention by those skilled in the art based on the above description are still within the scope of protection of the present invention.

[0027] Please refer to the appendix for the structure of the large field-of-view, high-resolution compound eye camera array system based on bionic human eye vision in this example. Figure 1 and attached Figure 2 The system consists of a curved support and an industrial camera array. The industrial camera array is arranged on the curved support, with the backward extensions of the optical axes of each camera converging at the same focal point, forming a concave imaging structure similar to the human retina. Adjacent cameras maintain a certain overlap in their fields of view. The curved support supports the camera array, ensuring that the backward extensions of the optical axes of all cameras intersect at the same point, while also ensuring a stable overlap rate between the fields of view of adjacent cameras. The industrial camera array is used to acquire images of the target scene. A large field of view is obtained by fusing and stitching the images captured by each camera. High-resolution images are obtained by super-resolution reconstruction of multiple highly relevant images of the region of interest in the center of the field of view. The system offers three operating modes: low-resolution mode, normal resolution mode, and super-resolution mode. By adjusting the parameters of the industrial cameras, the system can perceive dynamic changes at the edge of the field of view in high frame rate, low-resolution mode, and magnify local details in the region of interest in the center of the field of view in super-resolution mode. This achieves non-uniform imaging, balancing a large field of view with high-resolution information acquisition, significantly improving information acquisition and processing efficiency, and optimizing structural compactness.

[0028] In this example, the industrial camera array consists of FA lenses and industrial area array cameras. The system contains a total of 5×5 cameras, arranged on a curved support to form a concave imaging structure similar to the human retina.

[0029] In this example, the MVL-HF1624M-10MP FA lens is used, with a fixed focal length of 16mm and supporting 10 megapixels. This meets the system's requirements for high-resolution image acquisition, capturing clear details of targets and providing high-quality raw data for target recognition in standard resolution mode, dynamic monitoring in low-resolution mode (maintaining effective resolution even at high frame rates), and reconstruction in super-resolution mode. Its optical distortion is as low as -0.02%, greatly reducing geometric distortion in imaging. The minimum focusing distance is 0.1m, supporting imaging of close-range targets. The aperture range is F2.4-F16; the large aperture increases light intake in low-light environments, ensuring image brightness, while the small aperture increases depth of field, enabling clear imaging of targets at different distances, adapting to the system's dynamic monitoring and recognition needs in complex environments. Its dimensions are Φ29×38.36mm and its weight is 70.2g. It is small and lightweight, which is conducive to compact layout in compound eye camera arrays, reducing the load pressure on curved support structures, and at the same time, it is easy to integrate into the overall mechanical structure of the system, ensuring the rationality of the spatial layout and the pose accuracy of the compound eye camera array.

[0030] In this example, the MV-CU120-10UMUC industrial area scan camera is used. This camera employs a CMOS rolling shutter sensor, specifically the Sony IMX226. In industrial vision scenarios, this sensor balances image quality and data readout speed, meeting the system's requirements for rapid dynamic capture and clear imaging. The camera's pixel size is 1.85μm × 1.85μm. This smaller pixel size helps achieve high-resolution imaging within a limited target area, providing a hardware foundation for the system to acquire richer image details, especially in super-resolution reconstruction, where it enhances the fineness of the reconstructed image. The camera's target surface size is 1 / 1.7", which determines the size of the camera's imaging area and its compatibility with the lens in the system. The camera has a resolution of 4024×3036, a high resolution of 12 megapixels, which meets the high requirements of the invention system for image detail. Whether for target recognition in regular resolution mode or as a raw image data source in super-resolution mode, it provides sufficient pixel information to ensure recognition accuracy and reconstruction results. Its monochrome camera (MV-CU120-10UM) can achieve a frame rate of 29.2fps in 4024×3036 Mono 8 mode; the color camera (MV-CU120-10UC) can achieve a frame rate of 4024×3036 Bayer 8 mode. In RG8 mode, the frame rate is also 29.2fps. This frame rate is suitable for the system's need for high frame rate, low resolution image acquisition in dynamic monitoring mode, and can also complete target recognition and tracking image acquisition at a suitable speed in normal resolution mode. The camera supports custom regions of interest, allowing selection of specific areas in the image for high-resolution acquisition, which is suitable for the system's accurate recognition and tracking of target areas in normal resolution mode. It also supports horizontal and vertical mirror output, which allows the system to flexibly adjust the image orientation according to the camera's installation position, ensuring the convenience of image stitching and processing.

[0031] The specifications of the industrial area scan camera are shown in Table 1.

[0032] Table 1. Specifications of the industrial area scan camera in this example system.

[0033]

[0034] In this example, as shown in the appendix Figure 2As shown, five industrial area array cameras are equidistantly arranged on a curved aluminum support. The backward extensions of the optical axes of each camera intersect at a single point, forming a 1×5 horizontal array. Each camera has a field of view of 20°×15°. Using a small field-of-view lens for each camera achieves high spatial resolution while avoiding significant off-axis aberrations. Adjacent cameras have 50% field-of-view overlap to ensure accuracy during camera calibration and image stitching, and to provide sufficient reference pixels for super-resolution reconstruction of the region of interest. This system configuration enables full-field target information acquisition, with a full field of view of 60°×45°.

[0035] In this example, after the system acquires images of the target scene, it inputs the captured images into the computer. For images acquired by two adjacent cameras, a weighted fusion algorithm is used to fuse corresponding pixels in the overlapping area according to a set weight ratio. For non-overlapping areas, a stitching operation is performed, realizing the acquisition of full field-of-view information in the bionic human eye visual imaging theory, as shown in the attached diagram. Figure 3 As shown, the field of view of each individual camera in the system occupies 4 / 25 of the entire field of view, and the overlap ratio of the fields of view of adjacent cameras is 50%. Next, to further simulate the unique attention mechanism of the human eye, a super-resolution reconstruction algorithm was used to reconstruct local details of the region of interest in the center of the field of view. The central region of interest in the system's entire field of view is covered by the fields of view of the four cameras, used to simulate the densely distributed photoreceptor cells in the macula of the human retina, achieving a 4-fold resolution enhancement; the peripheral field of view of the system is covered by the fields of view of a small number of cameras, aiming to complete scene dynamic monitoring with lower data acquisition costs. The system offers three modes during operation: low-resolution mode, normal resolution mode, and super-resolution mode. Low-resolution mode is suitable for monitoring dynamic changes across the entire field of view and quickly capturing moving objects in the scene; normal resolution mode is used for target recognition and tracking under normal circumstances; and super-resolution mode provides high-resolution detail information for specific targets of interest in the center of the field of view, meeting the needs for accurate analysis of key areas and targets. Different modes can be selected to meet different monitoring priorities according to different needs.

[0036] In this example, the normal resolution mode is used for target recognition and tracking under normal circumstances, and is suitable for scenarios where the target position is relatively fixed or moves at low speed (such as static object detection and low-speed vehicle tracking).

[0037] In this example, the low-resolution mode is used for dynamic monitoring of the entire field of view. By acquiring information at a low resolution and high frame rate, the accuracy of dynamic perception is improved, while reducing the data acquisition and transmission load. When the system switches to low-resolution mode, the signal current of four adjacent pixels is accumulated and read out at once using the industrial camera's Binning mode. This achieves low-resolution, high-frame-rate image output with a resolution of 1 / 4 that of the original camera and a frame rate four times that of the original camera.

[0038] In this example, the super-resolution mode is used to magnify the details of the region of interest (ROI) image. By fusing images from adjacent cameras, a super-resolution reconstruction algorithm is used to overcome the physical resolution limit of the cameras, providing clear local detail information of the target of interest. In the proposed system, the central region of the field of view is covered by the overlapping fields of view of four cameras. Through a multi-camera synchronous exposure hardware design, multi-angle information is acquired, achieving image resolution enhancement. The detail acquisition capability is improved by 4 times compared to a single camera, and the details of the reconstructed image are clearly reproduced.

[0039] This invention proposes a large field-of-view, high-resolution compound eye camera array system based on biomimetic human eye vision. Combining non-uniform imaging theory, by designing field-of-view overlap and shooting modes between adjacent cameras, it can acquire images with different sampling frame rates and resolutions across the entire field of view, achieving high-resolution detail magnification of the region of interest (ROI) at the center of the field of view and dynamic perception of the edge regions. The system uses a curved support to support the camera array, ensuring that the backward extensions of the optical axes of all cameras intersect at the same point, maintaining a certain degree of field-of-view overlap between adjacent cameras. An industrial camera array is used to acquire images of the target scene. Large field-of-view information is obtained by fusing and stitching images captured by various cameras, and high-resolution images are obtained by super-resolution reconstruction of multiple highly correlated low-resolution images of the ROI at the center of the field of view. Ultimately, the system can perceive dynamic changes at the edge of the field of view in a high frame rate, low-resolution mode, and perform local detail magnification of the ROI at the center of the field of view using a super-resolution mode combined with image super-resolution algorithms, thereby achieving non-uniform imaging, balancing large field of view and high-resolution information acquisition, significantly improving information acquisition and processing efficiency, and optimizing structural compactness.

Claims

1. A large field-of-view, high-resolution compound eye camera array system based on bionic human eye vision, characterized in that, The system comprises a curved support and an industrial camera array. The industrial camera array consists of industrial automation lenses and industrial area array cameras. The industrial camera array is arranged on the curved support, with the backward extensions of the optical axes of each camera converging at the same focal point, forming a concave imaging structure similar to the human retina. Adjacent cameras maintain a fixed field-of-view overlap rate. The curved support supports the camera array, ensuring that the backward extensions of the optical axes of all cameras intersect at the same point, while also guaranteeing a stable overlap rate between the fields of view of adjacent cameras. The industrial camera array is used to acquire images of the target scene. A large field-of-view information is obtained by fusing and stitching the images captured by each camera. High-resolution images are obtained by super-resolution reconstruction of multiple related images of the region of interest in the center of the field of view. The system offers three operating modes: low-resolution mode, standard resolution mode, and high-resolution mode. The system employs a combination of modes, including a low-resolution mode suitable for monitoring dynamic changes across the entire field of view and quickly capturing moving objects in the scene. A standard resolution mode is used for target recognition and tracking in general situations. The super-resolution mode provides high-resolution detail information for specific targets in the central region of interest, meeting the need for precise analysis of key areas and targets. Different modes can be selected to meet different monitoring priorities by adjusting the parameters of the industrial camera. Ultimately, the system can perceive dynamic changes at the edge of the field of view in a high frame rate, low-resolution mode, and magnify local details in the central region of interest in the super-resolution mode, thereby achieving non-uniform imaging. This balances a large field of view with high-resolution information acquisition, significantly improving information acquisition and processing efficiency while optimizing structural compactness.

2. The large field-of-view, high-resolution compound eye camera array system based on bionic human eye vision according to claim 1, characterized in that, The curved support has a concave structure, which makes the backward extensions of the optical axes of the n×n cameras arranged on it converge at the same point, forming a concave imaging layout similar to the human retina. This stabilizes the overlap rate of the fields of view of adjacent cameras and provides a hardware foundation for image stitching. The curvature of the curved surface naturally optimizes the edge field of view aberrations, ensuring stitching stability and avoiding edge field of view aberrations and stitching misalignment caused by planar arrays.

3. The large field-of-view, high-resolution compound eye camera array system based on bionic human eye vision according to claim 1, characterized in that, The curved support is precision-machined to control the spatial position and orientation deviation of each camera mounted on the support to within 1°, thus better ensuring the stability of the field of view overlap rate of adjacent cameras. The curved support uses lightweight alloy materials, which reduces the hardware volume and weight compared to traditional planar arrays.

4. The large field-of-view, high-resolution compound eye camera array system based on bionic human eye vision according to claim 1, characterized in that, The industrial camera array consists of industrial automation lenses and industrial area array cameras. The system contains a total of n×n cameras arranged on a curved support to form a concave imaging structure similar to the human retina. The field of view of a single camera is A×B, and there is α% overlap in the field of view between adjacent cameras. This is used to ensure the accuracy of camera calibration and image stitching processes, and to provide sufficient reference pixels for super-resolution reconstruction of the region of interest.

5. The large field-of-view, high-resolution compound eye camera array system based on bionic human eye vision according to claim 1, characterized in that, The industrial camera array uses small field-of-view lenses for each camera, which can achieve high imaging spatial resolution while avoiding large off-axis aberrations.

6. The large field-of-view, high-resolution compound eye camera array system based on bionic human eye vision according to claim 1, characterized in that, To obtain a wide field-of-view panoramic image, a weighted fusion algorithm is used to fuse corresponding pixels in overlapping areas according to a set weight ratio for images captured by two adjacent cameras. For non-overlapping areas, a stitching operation is performed. In this system, the field of view of each camera occupies 4 / n of the full field of view. 2 The overlap ratio of the fields of view of adjacent cameras is α%. The system is set to achieve full field of view target information acquisition. The full field of view is C×D, where the horizontal field of view C can be obtained from formula (1), and the vertical field of view D can be obtained from formula (2): C = [A×nA×α%×(n-1)] (1) D = [B×nB×α%×(n-1)] (2) 7. The large field-of-view, high-resolution compound eye camera array system based on bionic human eye vision according to claim 1, characterized in that, To simulate the unique attention mechanism of the human eye, a super-resolution reconstruction algorithm is used to reconstruct local details of the central region of interest in the field of view. The central region of interest in the full field of view of the system is covered by the field of view of m cameras, which is used to simulate the densely distributed photoreceptor cells in the macular region of the human retina, and can achieve m times resolution enhancement. The peripheral field of view of the system is covered by the field of view of a small number of cameras, which aims to complete scene dynamic monitoring with a low data acquisition cost.

8. The large field-of-view, high-resolution compound eye camera array system based on bionic human eye vision according to claim 1, characterized in that, The conventional resolution mode is used for target recognition and tracking under normal circumstances and is suitable for scenarios where the target position is relatively fixed or the target is moving at a low speed.

9. The large field-of-view, high-resolution compound eye camera array system based on bionic human eye vision according to claim 1, characterized in that, By using a small field-of-view lens with adjustable focal length, the target can always be kept within the depth of field by adjusting the focal length in real time, which can improve recognition accuracy.

10. The large field-of-view, high-resolution compound eye camera array system based on bionic human eye vision according to claim 1, characterized in that, The low-resolution mode is used for dynamic monitoring of the entire field of view. Information is acquired through low resolution and high frame rate, thereby improving the accuracy of dynamic perception and reducing the data acquisition and transmission load. When the system switches to low-resolution mode, the signal current of adjacent p pixels is accumulated and read out at once using the Binning mode of the industrial camera. This achieves low-resolution, high-frame-rate image output with a resolution of 1 / p and a frame rate of p times that of the original camera.

11. The large field-of-view, high-resolution compound eye camera array system based on bionic human eye vision according to claim 1, characterized in that, The super-resolution mode is used to magnify the details of the region of interest image. By fusing images from adjacent cameras, a super-resolution reconstruction algorithm is used to break through the physical resolution limit of the camera and provide clear local detail information of the target in the region of interest. The central region of the field of view is covered by the overlapping fields of view of m cameras. Multi-angle information is obtained through a multi-camera synchronous exposure hardware design, thereby enhancing the image resolution and improving the detail acquisition capability by m times compared to a single camera. The details of the image after super-resolution reconstruction are clearly reproduced.

Images