Image acquisition system of camera and method for constructing characteristics and position of visual object

Abstract

The invention discloses an image acquisition system of a camera and a method for constructing characteristics and positions of a visual object. The image acquisition system comprises an image acquisition system. The method has the beneficial effects that by utilizing the pixel change rule caused by the displacement of the camera and combining with the plumb line direction, the image region is dynamically divided, so that efficient and low-cost three-dimensional modeling is realized; a deep learning model depending on object attribute recognition is abandoned, and high convex, low concave and planar objects are quickly distinguished through direct comparison of pixel variation; a dynamic region division rule based on a camera attitude (such as an inclination angle) is designed to ensure that an image region can be effectively segmented in different motion states, and the dynamic adaptability is enhanced; through segmented focusing and lightweight pixel difference calculation, the data processing delay is reduced, and the real-time navigation requirement is met.

Description

Technical Field

[0001] The present invention relates to an image acquisition system, in particular to an image acquisition system of a camera and a method for constructing characteristics and positions of visual objects, belonging to the technical field of computer vision and three-dimensional space perception. Background Art

[0002] With the rapid development of intelligent devices, environmental perception and obstacle avoidance technologies have become core technologies in areas such as navigation for the blind, autonomous robot movement, and autonomous driving. Current mainstream obstacle avoidance solutions rely primarily on ultrasonic sensors, laser radar (LiDAR), and infrared ranging technologies. However, these methods have significant limitations in 3D modeling and object feature recognition. Ultrasonic technology calculates object distance by emitting high-frequency sound waves and receiving echoes. However, it only provides single-dimensional depth information and cannot construct 3D spatial models. It is also susceptible to interference from environmental noise and has low measurement accuracy. For example, in complex indoor scenes, multipath reflections can distort data, making it difficult to distinguish the outlines of overlapping objects. While LiDAR can generate point cloud data through laser beam scanning, enabling high-precision 3D modeling, its core limitation is that the actively emitted laser beam is electromagnetic radiation. Long-term use may pose a potential threat to human health, limiting its application in medical or household applications. Furthermore, LiDAR hardware is expensive (a single device can cost thousands to tens of thousands of dollars), making it difficult to popularize in consumer products. More importantly, all of these technologies rely on the principle of "active detection," emitting energy waves (sound waves, lasers) to obtain feedback data. This model not only poses a radiation risk but also can cause signal interference in multi-device collaboration scenarios. For example, when multiple LiDARs operate simultaneously, intersecting laser beams can cause data conflicts, requiring complex filtering algorithms to correct them, further increasing system complexity.

[0003] Camera-based visual perception technology is considered an ideal alternative due to its advantages such as no radiation, low cost, and high information density. For example, a multi-camera unmanned vehicle visual perception system is disclosed in publication number CN102819263A. Its purpose is to overcome the problems of the existing image processing systems of micro unmanned vehicles and autonomous mobile robots, such as single perception algorithms and weak real-time performance, poor system scalability, inconvenient debugging, and limitations in field of view control. A multi-threaded optimized visual perception method is adopted, including a multi-camera video acquisition thread, a lane line perception thread, and an obstacle, cone, and signal light perception thread. Compared with traditional single-threaded visual perception methods, it has higher real-time performance, stability, and scalability. Traditional cameras capture environmental information through two-dimensional images and infer three-dimensional structures with the help of computer vision algorithms (such as stereo matching and optical flow methods). However, existing camera technology faces the following challenges in the recognition of three-dimensional object features:

[0004] First, the algorithm complexity is too high. Mainstream methods rely on convolutional neural networks (CNNs) or deep learning models, which require pre-training of object attributes (such as category and texture), resulting in high computing power requirements and poor real-time performance. For example, in real-time navigation scenarios, the system needs to complete object recognition and distance estimation within milliseconds, but CNN-based models have difficulty meeting low-latency requirements due to their large number of parameters. In addition, such algorithms are highly dependent on hardware platforms (such as GPUs), limiting their application in embedded devices.

[0005] Secondly, existing methods do not fully utilize the law of image changes under continuous displacement. In dynamic scenes, when the camera moves with the carrier (such as the robot body), the continuous frame images contain rich spatial information. For example, when the camera moves forward, the pixel change rate of nearby objects is much faster than that of distant objects, and the pixel offset directions of highly convex and concave objects are different. However, traditional optical flow methods or stereo vision technologies only focus on the pixel displacement amount, and do not associate it with the three-dimensional characteristics of the object (convexity, concavity, flatness), resulting in insufficient modeling accuracy;

[0006] Finally, it suffers from poor adaptability to dynamic environments. Existing region segmentation methods are often based on fixed thresholds or static templates, and are unable to dynamically adjust to camera posture (e.g., tilt angle). For example, when the camera rotates horizontally or vertically due to camera motion, fixed region segmentation can lead to loss of key object information (e.g., edge objects being mistakenly classified as adjacent regions), thus affecting feature recognition accuracy. Summary of the Invention

[0007] The purpose of the present invention is to provide an image acquisition system of a camera and a method for constructing the characteristics and positions of visual objects in order to solve at least one of the above technical problems.

[0008] The present invention achieves the above-mentioned object through the following technical solutions: A camera image acquisition system includes a camera, a rotating connector and a body, the camera is movably connected to the upper end of the body through the rotating connector, and the camera tail end of the camera is connected to a gravity sensor and a gyroscope;

[0009] The rotating connector includes a rotating motor unit and a bracket unit. The camera and the rotating motor unit, the body and the rotating motor unit, and the motors of the rotating motor unit are all connected through the bracket unit.

[0010] As a further solution of the present invention: the rotating motor unit includes an upper and lower rotating motor and a left and right rotating motor, the bracket unit includes a connecting seat, a connecting rod bracket and an arc plate bracket, the body of the left and right rotating motor is fixedly connected to the upper end of the body through the connecting seat, the body of the upper and lower rotating motor is fixedly connected to the rotating shaft of the left and right rotating motor through the arc plate bracket, and the upper and lower rotating motors and the left and right rotating motors are staggered and distributed up and down, and the rotating shaft of the upper and lower rotating motors is fixedly connected to the side of the camera shell through the connecting rod bracket.

[0011] As a further solution of the present invention: the camera includes but is not limited to a monocular or binocular camera, and the camera 1 continuously captures images through displacement, and compares the pixel changes between the current frame and the historical frame to analyze the three-dimensional characteristics of the object.

[0012] As a further solution of the present invention: the camera collects image data through a segmented focusing method during mobile shooting.

[0013] As a further solution of the present invention: the image pixel picture captured by the camera is extracted into different layers in color difference segments, and the pixel difference analysis is completed by the MCU or GPU.

[0014] As a further solution of the present invention: the image acquisition system can be integrated into blind navigation equipment, robots and unmanned driving systems.

[0015] A method for constructing the characteristics and positions of visible objects using a camera includes an image acquisition system. The construction method includes the following steps:

[0016] S1, with the camera body as the center point and the straight forward direction as the extension of the straight line, the left and right rotation motor can drive the camera to rotate horizontally, and the up and down rotation motor can drive the camera to rotate up and down, with the rotation center being the camera center;

[0017] S2. Continuous image acquisition begins at the fuselage point. Image data acquired from the second frame onward is invalid data if it is identical to the previous frame or the previous frame. When the fuselage point is moved forward, each frame of image data is continuously acquired and compared with the previous frame or the previous frame. Horizontal objects and concave-convex objects exhibit their own unique characteristics and deflect or shift in their respective characteristic directions.

[0018] S3. The camera continuously captures image pixels and compares the pixel change between the current frame and the previous frame or a previous frame. The camera then analyzes whether each object in the current three-dimensional space is convex, concave, or flat based on the increase or decrease in the pixel change between the current frame and the previous frame.

[0019] S4, during the continuous image acquisition process when the body and the camera are moving, the focus is continuously focused on each corresponding position in each time period according to the current environmental characteristics. The position can be a certain point or a certain area;

[0020] S5. Divide the image pixels collected at a certain position into three areas, namely, a left area, a middle area, and a right area. The three areas are divided into the left area, the middle area, and the right area with the fuselage as the center and the vertical line as the straight line;

[0021] S6. Analyze the 3D space in the left area. If the object is highly convex, the 2D pixel image generated as the camera lens moves forward will be highly convex, and the vertical object will always maintain a constant vertical line. If the object is flat, the 2D pixel image generated as the lens moves forward will form an angle with the previously captured vertical image. This angle changes by a variable amount as the camera and camera body move forward, and this line is to the right of the previously captured line.

[0022] S7. Analyze the three-dimensional space of the right area. If the object is highly convex, the two-dimensional pixel image generated as the camera lens moves forward will be highly convex, and the vertical object will always maintain a constant vertical line. If the object is flat, the two-dimensional pixel image generated as the lens moves forward will form an angle with the previously captured vertical image. This angle changes by a variable amount as the camera and camera body move forward, and this line is to the left of the previously captured line.

[0023] S8. Analyze the three-dimensional space of the center area. If it is a highly convex object, the two-dimensional pixel image generated by the camera lens moving forward will be highly convex and the vertical object will always keep the vertical line unchanged, but the height of the two-dimensional pixel image generated by the collection will be relatively smaller. If it is a flat object, the two-dimensional pixel image generated by the camera lens moving forward will keep the vertical line unchanged, but the height of the two-dimensional pixel image generated by the collection will be relatively larger. The increase or decrease will be greater as the distance between the camera and the object gets closer.

[0024] As a further solution of the present invention: when the camera captures image pixels and the camera body moves forward continuously, convex objects, concave objects and flat objects all change according to their own unique objective laws, and the objective laws of change of convex objects, concave objects and flat objects are different from each other.

[0025] As a further solution of the present invention, the size division of the left, center, and right areas is based on the objective laws of optical projection principles. Specifically, the center area occupies a very small width, while the left and right areas occupy relatively large widths. Because the Y-axis plane of the horizontal rotation of the camera forms an angle with the X-axis plane of the body, the size division of the three areas, the left, center, and right areas, may change.

[0026] If the angle between the Y-axis and the X-axis is less than 90 degrees, the right zone will become larger and the left zone will become smaller. As the angle between the Y-axis and the X-axis gradually decreases, the left zone may disappear, and the middle zone remains almost unchanged.

[0027] If the angle between the Y-axis plane and the X-axis plane is greater than 90 degrees, the left zone division area will become larger, and the right zone division area will become smaller. As the angle between the Y-axis plane and the X-axis plane gradually decreases, the left zone division area may also disappear, and the middle zone division remains almost unchanged.

[0028] The beneficial effects of the present invention are:

[0029] 1. This invention utilizes the pixel variation caused by camera displacement and dynamically divides the image region in the direction of the vertical line to achieve efficient and low-cost 3D modeling. It abandons deep learning models that rely on object attribute recognition and instead uses direct comparison of pixel variation to quickly distinguish convex, concave, and flat objects. It designs dynamic region division rules based on camera posture (such as tilt angle) to ensure effective image region segmentation under different motion states and enhance dynamic adaptability. Through segmented focus and lightweight pixel difference calculation, it reduces data processing latency and meets real-time navigation requirements.

[0030] 2. This invention analyzes pixel changes in consecutive frames: By comparing the pixel differences between the current frame and the previous frames, it captures the characteristic deviation patterns of objects due to displacement. For example, the edges of highly convex objects in the left and right areas will shift laterally as the camera moves, while the edges of flat objects will show angular changes. Such patterns can be directly used for feature classification.

[0031] 3. The present invention uses gravity sensor and gyroscope data as a reference to implement dynamic area segmentation based on the vertical line of gravity. The image is divided into three areas: left, center, and right. The area width is dynamically adjusted according to the camera tilt angle. For example, when the camera is tilted to the right, the right area is reduced to avoid redundant data, thereby focusing on the key area.

[0032] 4. The present invention adopts segmented focus and variation association. During the camera displacement process, the focus point is adjusted by the motor to ensure the comparability of the corresponding areas of consecutive frames. At the same time, an inverse relationship is established between the pixel variation and the object distance (the greater the variation, the closer the distance), realizing distance estimation without the need for a depth sensor. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 Schematic diagram of the structure of the image acquisition system of the present invention;

[0034] Figure 2 Schematic diagram of image data segmentation area of ​​the present invention;

[0035] Figure 3Schematic diagram of the offset relationship of convex objects in the present invention;

[0036] Figure 4 This is a schematic diagram of the offset relationship of a planar object according to the present invention;

[0037] Figure 5 Schematic diagram of the offset relationship of concave objects in the present invention;

[0038] In the figure: 1. Camera; 2. Up and down rotation motor; 3. Left and right rotation motor; 4. Body; 5. Gravity sensor; 6. Gyroscope; 7. Connecting seat; 8. Connecting rod bracket; 9. Arc plate bracket. DETAILED DESCRIPTION

[0039] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only 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 making creative efforts are within the scope of protection of the present invention.

[0040] Example 1

[0041] like Figure 1 As shown, a camera image acquisition system includes a camera 1, a rotating connector and a body 4. The camera 1 is movably connected to the upper end of the body 4 through the rotating connector. The camera tail end of the camera 1 is connected to a gravity sensor 5 and a gyroscope 6.

[0042] The rotating connector includes a rotating motor unit and a bracket unit. The camera 1 and the rotating motor unit, the body 4 and the rotating motor unit, and the motors of the rotating motor unit are all connected via the bracket unit.

[0043] Example 2

[0044] In addition to all the technical features of the first embodiment, this embodiment also includes:

[0045] Furthermore, the rotating motor unit includes an upper and lower rotating motor 2 and a left and right rotating motor 3, and the bracket unit includes a connecting seat 7, a connecting rod bracket 8 and an arc plate bracket 9. The body of the left and right rotating motor 3 is fixedly connected to the upper end of the body 4 through the connecting seat 7, and the body of the upper and lower rotating motor 2 is fixedly connected to the rotating shaft of the left and right rotating motor 3 through the arc plate bracket 9, and the upper and lower rotating motor 2 and the left and right rotating motor 3 are staggered and distributed up and down, and the rotating shaft of the upper and lower rotating motor 2 is fixedly connected to the side of the shell of the camera 1 through the connecting rod bracket 8.

[0046] Furthermore, the camera 1 includes but is not limited to a monocular or binocular camera. The camera 1 continuously captures images through displacement, and compares the pixel changes between the current frame and the historical frame to analyze the three-dimensional characteristics of the object.

[0047] Furthermore, the camera 1 collects image data by a segmented focus method during mobile shooting, and dynamically adjusts the focus point during the displacement process to ensure the effectiveness of continuous data collection.

[0048] Furthermore, the image pixel picture captured by the camera 1 is divided into different layers using color difference segments, and the pixel difference analysis is performed by the MCU or GPU.

[0049] Furthermore, the image acquisition system can be integrated into navigation equipment for the blind, robots, and unmanned driving systems.

[0050] Example 3

[0051] A method for constructing the characteristics and positions of visible objects using a camera includes an image acquisition system. The construction method includes the following steps:

[0052] S1, with the four points of the fuselage as the center starting point and the straight forward direction as the extension of the straight line, the left and right rotation motor 3 can drive the camera 1 to do horizontal rotation movement, and the up and down rotation motor 2 can drive the camera 1 to do up and down rotation movement, and the rotation center is the camera center;

[0053] S2. Continuously capture images starting from the four points on the fuselage. Image data captured starting from the second frame is invalid data if it is identical to the previous frame or the previous frame. When the fuselage moves forward starting from the four points on the fuselage, each frame of image data is continuously captured and compared with the previous frame or the previous frame. Horizontal objects and concave and convex objects will have their own unique characteristics and will deviate or shift in their respective characteristic directions.

[0054] S3, camera 1 continuously captures image pixels, compares the pixel change between the current frame and the previous frame or a previous frame, and analyzes whether each object in the current three-dimensional space is convex, concave, or flat by increasing or decreasing the pixel change between the current frame and the previous frame;

[0055] S4, during the continuous image acquisition process when the body 4 and the camera 1 are moving, each corresponding position is continuously focused in each time period according to the current environmental characteristics. The position can be a certain point or a certain area;

[0056] S5. Divide the image pixels captured at a certain position into three areas, namely, a left area, a middle area, and a right area. The three areas are divided into the left area, the middle area, and the right area with the fuselage 4 as the center and the vertical line as the straight line;

[0057] S6. Analyze the 3D space in the left area. If the object is highly convex, the 2D pixel image generated by the forward movement of camera 1 will be highly convex, and the vertical object will always maintain a constant vertical line. If the object is flat, the 2D pixel image generated by the forward movement of the lens will form an angle with the previously captured vertical image. This angle changes by a certain amount as camera 1 and body 4 move forward, and this line is to the right of the previously captured line.

[0058] S7. Analyze the three-dimensional space of the right area. If the object is highly convex, the two-dimensional pixel image captured by camera 1 as it moves forward will be highly convex, and the vertical object will always maintain a constant vertical line. If the object is flat, the two-dimensional pixel image captured by camera 1 as it moves forward will form an angle with the previously captured vertical image. This angle changes by a variable amount as camera 1 and body 4 move forward, and this line is to the left of the previously captured line.

[0059] S8. Analyze the three-dimensional space of the center area. If it is a highly convex object, as the camera 1 lens moves forward, the two-dimensional pixel image generated by the capture is highly convex and the vertical object always keeps the vertical line unchanged, but the height of the two-dimensional pixel image generated by the capture will be relatively smaller. If it is a flat object, as the lens moves forward, the two-dimensional pixel image generated by the capture is highly convex and the vertical object always keeps the vertical line unchanged, but the height of the two-dimensional pixel image generated by the capture will be relatively larger, and the increase or decrease will be greater as the distance between the camera 1 and the object gets closer.

[0060] Furthermore, when the camera 1 captures image pixels and the camera body moves forward continuously, convex objects, concave objects and flat objects all change according to their own unique objective laws, and the objective laws of change of convex objects, concave objects and flat objects are different from each other.

[0061] Furthermore, the size division of the left, center, and right areas is based on objective laws of optical projection principles. Specifically, the center area occupies a very small width, while the left and right areas occupy relatively large widths. Because the Y-axis plane of the horizontal rotation of the camera 1 forms an angle with the X-axis plane of the body 4, the size division of the three areas, the left, center, and right areas, may change.

[0062] If the angle between the Y-axis and the X-axis is less than 90 degrees, the right zone will become larger and the left zone will become smaller. As the angle between the Y-axis and the X-axis gradually decreases, the left zone may disappear, and the middle zone remains almost unchanged.

[0063] If the angle between the Y-axis plane and the X-axis plane is greater than 90 degrees, the left zone division area will become larger, and the right zone division area will become smaller. As the angle between the Y-axis plane and the X-axis plane gradually decreases, the left zone division area may also disappear, and the middle zone division remains almost unchanged.

[0064] Example 4

[0065] like Figures 2 to 5 As shown, a method for constructing the characteristics and positions of visible objects by a camera, the vertical dotted lines of all views are regarded as parallel to the perpendicular line, and the relationship between the three areas of left area, middle area and right area and the three surfaces of convex surface, flat surface and concave surface is analyzed.

[0066] 1. Relationship between Change and Depth

[0067] Focusing method within split time periods:

[0068] like Figure 2 As shown in Figure 1, as time progresses, the camera continuously focuses on a certain position in the current time period and another position in the next time period according to the environmental characteristics during the continuous image acquisition process. The focus position is determined based on the environmental characteristics.

[0069] When the position of the body 4 and the camera 1 moves, such as from point O to point O1, the original focus position D of the body 4 and the camera 1 will inevitably be offset. In order to keep the camera 1 focused on the original position D, the left and right rotating motor 3 and the up and down rotating motor 2 rotate together to keep the camera focused on the original position D.

[0070] Image data three-area division technology

[0071] like Figure 2 As shown, according to the angle a between the Y-axis plane of the camera and the X-axis plane of the fuselage, the image data collected by the camera is divided into several corresponding areas in the direction of the gravity perpendicular line of the gravity sensor.

[0072] 1) When the angle a is 90 degrees, the image data is divided into three corresponding areas, namely the left area, the middle area and the right area;

[0073] 2) When the angle a is less than 90 degrees, the image data is divided into three corresponding areas, namely the left area, the middle area and the right area; however, as the angle a gradually decreases, the division of the left area will gradually decrease. As the angle a gradually decreases, the left area may disappear, and the middle area may also become smaller or disappear;

[0074] 3) When the angle a is greater than 90 degrees, the image data is divided into three corresponding areas, namely the left area, the middle area and the right area; however, as the angle a gradually increases, the division of the right area will gradually decrease. As the angle a gradually increases, the right area may disappear, and the middle area may also become smaller or disappear.

[0075] II. Offset Relationship of Image Frame Data after Displacement

[0076] As Figures 3 to 5 shown, when a set of continuous two-dimensional data of planar images is collected and the focal point is at position D. At this time, when comparing the current frame data with the data of a certain previous frame, the fuselage is at position O1. If the current object is a planar object:

[0077] If it is a planar object and a left-region object, for the image frame data currently collected, the side line of the plane offsets towards the center with a change amount of the angle b with respect to the perpendicular bisector; if it is a planar object and a central-plane object, for the image frame data currently collected, the side line of the plane is linearly stretched by a change amount value h; if it is a planar object and a right-region object, for the image frame data currently collected, the side line of the plane offsets towards the center with a change amount of the angle b with respect to the perpendicular bisector.

[0078] When a set of continuous two-dimensional data of high-convex images is collected and the focal point is at position D. At this time, when comparing the current frame data with the data of a certain previous frame, the fuselage is at position O1. If the current object is a high-convex object relative to the plane: If it is a high-convex object and a left-region object, for the image frame data currently collected, the perpendicular line of the vertex of the high-convex object still remains parallel to the perpendicular line, but the side line that was originally parallel to the perpendicular line will have a leftward parallel offset, and the height of the high-convex object retracts and shrinks by a length less than that of the left-region planar object at the same position; if it is a high-convex object and a central high-convex object, for the image frame data currently collected, the distance from the vertex to the focal point will be stretched, and the perpendicular line of the vertex still coincides and there is no offset, with no offset amount, and the height of the high-convex object retracts and shrinks by a length h1 < h relative to the planar object; if it is a high-convex object and a right-region object, for the image frame data currently collected, the perpendicular line of the vertex of the high-convex object still remains parallel to the perpendicular line, but the side line that was originally parallel to the perpendicular line will have a rightward parallel offset, and the height of the high-convex object retracts and shrinks by a length less than that of the right-region planar object at the same position.

[0079] When a set of continuous two-dimensional data of low-concave images is collected and the focal point is at position D. At this time, when comparing the current frame data with the data of a certain previous frame, the fuselage is at position O1. If the current object is a low-concave object relative to the plane: If it is a low-concave object and a left-region object, for the image frame data currently collected, the side line of the plane has a linear offset of the right side line with a change amount or an angle b; if there is line-of-sight occlusion, when the camera moves to position O1, a new plane will be added; if it is a low-concave object and a central-region object, for the image frame data currently collected, the side line of the plane is linearly stretched by a change amount value, and if there is line-of-sight occlusion, when the camera moves to position O1, a new plane will be added; if it is a low-concave object and a right-region object, for the image frame data currently collected, the side line of the plane has a linear offset of the left side line with a change amount or an angle b;

[0080] 3. Relationship between Change and Depth

[0081] For planar objects, when the camera and the camera move from the starting position O to the target position O1, a continuous set of two-dimensional planar image data is collected. When the current frame data is compared with the previous frame data, the height of the central object will increase. The greater the change in the height of the planar object, the closer the planar object is to the camera and the camera.

[0082] For convex objects, when the camera body and camera move from the starting position O to the target position O1, a set of continuous convex image two-dimensional data is collected. At this time, when the current frame data is compared with the previous frame data, the vertical line of the outer edge of the convex object will be offset. The greater the change in the vertical line of the outer edge is, the closer the convex object is to the camera body and camera.

[0083] For concave objects, when the camera body and camera move from the starting position O to the target position O1, a set of continuous concave image 2D data is collected. At this time, when the current frame data is compared with the previous frame data, the distance from any point on the inner surface of the concave object to point O1 will become larger than the flat image 2D data. The larger the numerical change of the concave surface change, the closer the concave object is to the camera body and camera.

[0084] Working Principle: Camera 1 continuously shifts and captures the image of the current area to generate pixel data. This pixel data is generated using a three-region method, which divides the currently captured two-dimensional pixel image into three regions: left, center, and right. By comparing the characteristic changes of each region of pixels in the current image with the corresponding region of pixels in the previous frame or a previous frame, the three-dimensional characteristics of each object in the current region can be analyzed. This method requires fewer algorithms, requiring a simple algorithm to determine whether the current object is convex, concave, or flat. Object characteristics are not analyzed based on object attributes, names, convolutional neural networks, or object appearance. Instead, color layering extracts characteristics to analyze the spatial characteristics of visible objects. The pixel data captured by the camera also facilitates the continued use of other algorithms, such as convolutional neural networks. This is because any object has two surface characteristics: a flat surface is a continuation of a horizontal line, and a vertical surface is a continuation of a vertical line. An inclined or curved surface is a combination of the continuation of a horizontal and vertical line.

[0085] Based on the different characteristic changes of the plane and vertical surfaces of the two-dimensional image pixels continuously captured by the camera 1, and based on the objective regularity change analysis of the different characteristics of the image pixels formed by the plane and vertical surfaces, a method for demonstrating the characteristics of three-dimensional space objects is constructed.

[0086] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above and that the invention can be embodied in other specific forms without departing from the spirit or essential characteristics of the invention. Therefore, the embodiments should be considered in all respects as illustrative and non-restrictive, and the scope of the invention is defined by the appended claims, not the foregoing description, and all variations within the meaning and range of equivalents of the claims are intended to be included therein. Any reference sign in a claim should not be construed as limiting the claim to which it relates.

[0087] In addition, it should be understood that although this specification is described in terms of implementation methods, not every implementation method contains only one independent technical solution. This narrative method of the specification is only for the sake of clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each embodiment can also be appropriately combined to form other implementation methods that can be understood by those skilled in the art.

Claims

1. A camera image acquisition system, comprising a camera (1), a rotating connector and a body (4), characterized in that: The camera (1) is movably connected to the upper end of the fuselage (4) via a rotating connector, and the camera tail end of the camera (1) is connected to a gravity sensor (5) and a gyroscope (6); The rotating connection member comprises a rotating motor unit and a bracket unit, and the camera (1) and the rotating motor unit, the body (4) and the rotating motor unit, and the motors of the rotating motor unit are all connected via the bracket unit.

2. The image acquisition system of a camera according to claim 1, characterized in that: The rotary motor unit comprises an upper and lower rotary motor (2) and a left and right rotary motor (3); the bracket unit comprises a connecting seat (7), a connecting rod bracket (8) and an arc plate bracket (9); the body of the left and right rotary motor (3) is fixedly connected to the upper end of the body (4) through the connecting seat (7); the body of the upper and lower rotary motor (2) is fixedly connected to the rotating shaft of the left and right rotary motor (3) through the arc plate bracket (9); and the upper and lower rotary motor (2) and the left and right rotary motor (3) are distributed in an offset manner up and down; the rotating shaft of the upper and lower rotary motor (2) is fixedly connected to the side of the housing of the camera (1) through the connecting rod bracket (8).

3. The camera image acquisition system according to claim 1, wherein: The camera (1) includes but is not limited to a monocular or binocular camera. The camera (1) continuously captures images through displacement, and compares the pixel changes between the current frame and the historical frame to analyze the three-dimensional characteristics of the object.

4. The image acquisition system of a camera according to claim 1, characterized in that: The camera (1) collects image data by a segmented focusing method during mobile shooting.

5. The camera image acquisition system according to claim 1, wherein: The image pixel picture collected by the camera (1) is divided into different layers using color difference segments, and pixel difference analysis is performed by the MCU or GPU.

6. The camera image acquisition system according to claim 1, characterized in that: The image acquisition system can be integrated into navigation equipment for the blind, robots and unmanned driving systems.

7. A method for constructing visual object characteristics and positions using a camera, comprising the image acquisition system according to any one of claims 1 to 6, characterized in that: The construction method comprises the following steps: S1, with the body (4) as the center starting point and the front direction as the extension of the straight line, the left and right rotation motor (3) can drive the camera (1) to do horizontal rotation movement, and the up and down rotation motor (2) can drive the camera (1) to do up and down rotation movement, and the rotation center is the camera center; S2, starting from the fuselage (4) point to continuously capture the image, at this time, the image data of each current frame and the image data of the previous frame or the previous frame collected from the second frame are the same and are invalid data; when starting from the fuselage (4) point to move in the forward direction, as the image data of each frame is continuously collected and compared with the image data of the previous frame or the previous frame collected, the horizontal objects and the concave and convex objects will have their own unique characteristics and will be offset or displaced in their own different characteristic directions; S3, the camera (1) continuously collects image pixels, compares the pixel change amount between the current frame and the previous frame or a previous frame, and analyzes whether each object in the current three-dimensional space is a convex, concave or flat object by increasing or decreasing the pixel change amount between the current frame and the previous frame; S4, during the continuous image acquisition process when the body (4) and the camera (1) are moving, each corresponding position is continuously focused on in each time period according to the current environmental characteristics, where the position may be a point or an area; S5, dividing the image pixel screen collected at a certain position into three areas, namely, the left area, the middle area and the right area, wherein the three areas are divided into the left area, the middle area and the right area with the fuselage (4) as the center and the vertical line as the straight line; S6. Analyze the three-dimensional space of the left area. If it is a highly convex object, the two-dimensional pixel image generated by the forward movement of the camera (1) lens is highly convex and the vertical object always maintains a constant vertical line. If it is a flat object, the two-dimensional pixel image generated by the forward movement of the lens is at an angle to the vertical image previously collected, and the angle changes by a variable amount as the camera (1) and the body (4) move forward, and the line is to the right of the previously collected line. S7. Analyze the three-dimensional space of the right area. If it is a highly convex object, the two-dimensional pixel image generated by the forward movement of the camera (1) lens is highly convex and the vertical object always maintains a constant vertical line. If it is a planar object, the two-dimensional pixel image generated by the forward movement of the lens is at an angle to the vertical image previously collected, and the angle changes by a variable amount as the camera (1) and the body (4) move forward, and the line is on the left side of the previously collected line. S8. Analyze the three-dimensional space of the center area. If it is a highly convex object, the two-dimensional pixel image generated by the camera (1) moves forward and the vertical object always keeps the vertical line unchanged, but the height of the two-dimensional pixel image generated by the capture will be relatively smaller. If it is a flat object, the two-dimensional pixel image generated by the camera (1) moves forward and the flat object always keeps the vertical line unchanged, but the height of the two-dimensional pixel image generated by the capture will be relatively larger. The increase or decrease will be greater as the distance between the camera (1) and the object gets closer.

8. The method for constructing visual object characteristics and positions according to claim 7, wherein: In said S5, when the camera (1) collects image pixels and the camera and the body continuously move forward, convex objects, concave objects and flat objects all change according to their own unique objective laws, and the objective laws of change of convex objects, concave objects and flat objects are different from each other.

9. The method for constructing visual object characteristics and positions according to claim 7, wherein: In said S5, the size division of the left area, the middle area and the right area is based on the objective law of the optical projection principle, specifically: the middle area occupies a very small width, and the left area and the right area occupy a relatively large width; because the Y-axis plane of the horizontal rotation of the camera (1) will form an angle with the X-axis plane of the body (4): at this time, the size division of the three areas, the left area, the middle area and the right area, will change; If the angle between the Y-axis and the X-axis is less than 90 degrees, the right zone will become larger and the left zone will become smaller. As the angle between the Y-axis and the X-axis gradually decreases, the left zone may disappear, and the middle zone remains almost unchanged. If the angle between the Y-axis plane and the X-axis plane is greater than 90 degrees, the left zone division area will become larger, and the right zone division area will become smaller. As the angle between the Y-axis plane and the X-axis plane gradually decreases, the left zone division area may also disappear, and the middle zone division remains almost unchanged.

Images