TWI784451B - Image conversion system and image conversion method - Google Patents

Abstract

An image conversion system includes a camera and a processor. The camera is used to capture an image. The image includes a plurality of positioning label patterns. The processor is used for receiving the image, by identifying the positioning label patterns, generating an image specific area according to the positioning label patterns, and converting the image specific area to a plane specific area of a plan map.

Description

Image conversion system and image conversion method

The invention relates to an application of image conversion, in particular to an image conversion system and an image conversion method.

When an object is detected by a traditional camera image, it is necessary to use camera positioning, calibration, etc. to convert the spatial position coordinates of the object to the plan. If the camera type changes, the model is updated, the placement position is changed, or the internal parameters of the camera are changed, then Positioning and alignment will need to be redone. However, positioning and calibration often requires personnel to replace the camera on site and measure the position of the camera on the spot, which is very inconvenient.

In addition, when the coordinates of multiple cameras are converted to the same plan, the problem of coordinating coordinates on the plan will also occur. The traditional coordinate conversion method needs to perform appropriate image conversion on the camera first, such as converting the image taken by the fisheye camera into a square plane to correspond to the real map, and then measure the installation height of the camera itself and the installation position corresponding to the world map, namely The coverage area corresponding to the floor plan of the image can be obtained, and then the detected position of the object in the camera image can be converted to the floor plan. When there are a large number of cameras, it is necessary to accurately measure the coverage of the cameras to ensure that the coverage of the images does not interfere with each other. Otherwise, the same object will appear in two cameras, and it is necessary to deal with the problem of which coordinates the object belongs to.

Therefore, how to set a new image conversion method has become one of the problems to be solved in this field, regardless of the type, parameters, installation location, etc. of the camera itself. It only needs to be able to directly set the correlation with the floor plan from the image after the camera is installed. one.

In order to solve the above problems, an aspect of the present disclosure provides an image conversion system. The image conversion system includes a camera and a processor. The camera is used to shoot an image; wherein, the image includes a plurality of positioning label patterns. The processor is used for receiving the image, by identifying the positioning tag patterns, generating a specific image area according to the positioning tag patterns, and converting the specific image area into a plane specific area of a plan view.

In one embodiment, the image conversion system further includes an indoor positioning device. The indoor positioning device is used to transmit a signal with a plurality of base stations through a plurality of positioning tags. After receiving the signals, the base stations transmit the strength of the signals to the processor, and the processor performs a triangulation calculation method to calculate the corresponding relative positions of these positioning tags in a space.

In one embodiment, the processor obtains three different relative positions of the positioning tags in space, or obtains one of the positioning tags and moves to different positions three times in space, and obtains three Different relative positions, connect these three different relative positions to obtain a triangular plane transformed into space.

In one embodiment, after the processor corresponds each vertex of the triangular plane to the position of these positioning label patterns, the corresponding position of the triangular plane in the image is obtained, and then the corresponding position of the triangular plane in the image is converted into a plan view superior.

In one embodiment, a specific area of a complete plane is composed of a plurality of triangular planes. When the specific area of the complete plane is established, the processor inputs the image into a trained deep learning network, and the deep learning network outputs a frame The processor calculates a center point position of the bottommost line of the frame selection range, and tracks a movement path of the center point position.

Another aspect of the disclosure provides an image conversion method. The image conversion method includes the following steps: taking an image; wherein, the image includes a plurality of positioning label patterns; receiving the image, identifying the positioning label patterns, and generating a specific area of the image according to the positioning label patterns; and transforming the image The specific area is converted to a plane specific area of a plan.

In one embodiment, the image conversion method further includes: transmitting a signal to a plurality of base stations through a plurality of positioning tags, and each of the base stations transmits the strength of the signals to a processor after receiving the signals The processor uses a triangulation algorithm to calculate the corresponding relative positions of the positioning tags in a space.

In one embodiment, the image conversion method further includes: obtaining three different relative positions of the positioning tags in space, or obtaining one of the positioning tags to move three times to different positions in space, Three different relative positions are obtained; and connecting the three different relative positions to obtain a triangular plane transformed into space.

In one embodiment, the image conversion method further includes: after corresponding the vertices of the triangular planes to the positions of the positioning label patterns, the corresponding positions of the triangular planes in the image are obtained, and then the corresponding positions of the triangular planes in the image are obtained. Location converted to floor plan.

In one embodiment, a specific area of a complete plane is composed of a plurality of triangular planes. When the specific area of the complete plane is established, the image is input to a deep learning network that has been trained, and the deep learning network outputs a frame selection range ; Calculating a center point position of the bottommost line of the marquee range; and converting a specific area of the image to a specific area of a plane of a plan view.

The image conversion system and image conversion method shown in the present invention, no matter the images taken by the fisheye camera or the bullet camera, can convert these images to the same plan to know the coverage of the camera on the plan , each coverage area is composed of multiple triangular planes. The image conversion system and image conversion method shown in the present invention do not need to know any parameters and positions of multiple cameras, and the coverage areas of different lenses do not need to be positioned or corrected. Convert the shooting range of multiple identical or different cameras (that is, a specific area of a complete plane) to the same floor plan, greatly reducing the time spent on manpower when deploying cameras.

In addition, after the processor automatically converts the range captured by the camera to the plane map, the processor can automatically track the human body displacement in a specific area of the complete plane through image processing methods. This can be applied to long-term care centers, because the elderly generally have different Due to the habit of wearing electronic devices, the image conversion system and image conversion method shown in the present invention can use the center point position (that is, the position of the human body) marked in the specific area of the complete plane to track the moving path of the elderly, which is helpful It is used to judge whether the elderly have dementia.

The following description is a preferred implementation of the invention, and its purpose is to describe the basic spirit of the invention, but not to limit the invention. For the actual content of the invention, reference must be made to the scope of the claims that follow.

It must be understood that words such as "comprising" and "comprising" used in this specification are used to indicate the existence of specific technical features, values, method steps, operations, components and/or components, but do not exclude possible Add more technical characteristics, values, method steps, operation processes, components, components, or any combination of the above.

Words such as "first", "second", and "third" used in the claims are used to modify the elements in the claims, and are not used to indicate that there is an order of priority, an antecedent relationship, or an element An element preceding another element, or a chronological order in performing method steps, is only used to distinguish elements with the same name.

Please refer to FIGS. 1-5. FIG. 1 is a block diagram of an image conversion system 100 according to an embodiment of the present invention. FIG. 2 is a block diagram of an image conversion system 200 according to an embodiment of the present invention. FIG. 3 is a schematic diagram illustrating an image conversion method 300 according to an embodiment of the present invention. FIG. 4 is a schematic diagram illustrating an indoor positioning method according to an embodiment of the present invention. FIG. 5 is a schematic diagram illustrating an indoor positioning method according to an embodiment of the present invention.

As shown in FIG. 1 , the image conversion system 100 includes a camera 10 and a processor 20 . As shown in FIG. 2 , the image conversion system 200 includes a plurality of cameras 10 , 12 , a processor 20 , a plurality of base stations AP1 - AP4 and a positioning tag TAG0 .

The cameras 10 and 12 in FIGS. 1-2 may be fisheye cameras, bullet cameras or other types of cameras. In one embodiment, the cameras 10 and 12 are communicatively coupled to the processor 20 so that the cameras 10 and 12 can transmit captured images to the processor 20. The communication coupling can be through wired or wireless transmission.

In one embodiment, the processor 20 in Figures 1-2 can be composed of an integrated circuit such as a micro controller, a microprocessor, a digital signal processor, or an application-specific integrated circuit. circuit (application specific integrated circuit, ASIC) or a logic circuit to implement. In an embodiment, the processor 20 in FIGS. 1-2 may be located in a server, a desktop computer, a laptop or other electronic devices with computing functions.

In one embodiment, the base stations AP1-AP4 may be wireless base stations (also called "wireless AP", where AP refers to "Access Point").

Please also refer to Figures 2-4. Indoor positioning can be realized by using Beacon wireless signal system technology or radio frequency identification (Radio Frequency Identification, RFID). The following uses Beacon wireless signal system technology as an example. Generally, knowledgeable persons should be able to understand that any method that can perform indoor positioning can be applied here.

In step 310, the camera 10 captures an image; wherein, the image includes a plurality of positioning label patterns.

In one embodiment, the indoor positioning device of the image conversion system 200 is used to transmit a signal with one or more base stations (such as base stations AP1~AP4) respectively through a plurality of positioning tags, and each of these base stations receives these After the signals are sent, the strengths of these signals are sent to the processor 20, and the processor 20 uses a triangulation algorithm to calculate the respective relative positions of the positioning tags in a space. Since the triangulation algorithm is a known spatial positioning method, it is not repeated here.

In one embodiment, as shown in FIG. 4 , the base stations AP1 - AP4 can be deployed in four different positions in the space SP, and the positioning tag TAG0 can be placed in any position in the space SP. The implementation of Beacon wireless signal system technology requires a positioning tag TAG0 (a tag capable of actively transmitting signals, Beacon tag) and base stations AP1~AP4 (applying at least three base stations to realize the triangulation algorithm).

In one embodiment, the positioning tag TAG0 sends a signal through Beacon's wireless signal transmission technology, so that the base stations AP1~AP4 receive the signal, and the base station AP1~AP4 can indicate the received signal strength (Received Signal Strength Indication) of the received signal , RSSI) to the processor 20 or the positioning server coupled with the communication of the base stations AP1~AP4, so that the processor 20 or the positioning server receiving these received signal strength indications can locate based on the signal field strength indication principle, according to Each base station AP1-AP4 receives the strength of the signal, and calculates the position of the positioning tag TAG0 in the space SP. In one embodiment, if the positioning server calculates the position of the positioning tag TAG0 in the space SP, the positioning server will send the position to the processor 20 .

In other words, after the base stations AP1-AP4 are set up, the processor 20 or the positioning server can calculate the position of the positioning tag TAG0 in the space SP based on the signal field strength indication positioning principle and the triangulation positioning algorithm.

In one embodiment, as shown in FIG. 5 , there may be three or more positioning tags TAG1 - TAG3 in the space SP. For convenience of description, three positioning tags TAG1 - TAG3 are taken as an example below.

In one embodiment, the processor 20 or the positioning server can respectively calculate the positions of the three positioning tags TAG1-TAG3 in the space SP according to the above-mentioned method and the triangulation positioning algorithm, and calculate the respective positions of the three positioning tags TAG1-TAG3 in the space SP. The respective positions of TAG3 in space SP are connected by line segments to form a triangular plane TR. Next, the processor 20 needs to actually map the triangular plane TR to the range of images captured by the camera 10 or 12 .

In one embodiment, as shown in FIG. 5, the processor 20 obtains three different relative positions of the three positioning tags TAG1-TAG3 in the space SP, or obtains one of the positioning tags TAG1-TAG3 in the space SP Move to different positions three times (for example, move the positioning tag TAG1 to different positions three times), to obtain three different relative positions, and connect these three different relative positions to obtain a triangular plane in space SP TR, corresponding to the three vertices of the triangular plane TR and the positioning label patterns in the image 600 captured by the camera 10, so as to obtain the triangular plane TR' in the image 600, and then convert the triangular plane TR' into a plan view 700 A triangular plane TR'' in . This part will be described in detail later.

In step 320, the processor 20 receives the image 600, recognizes the positioning tag patterns, and generates an image specific region (such as a triangular plane TR') according to the positioning tag patterns.

Please refer to FIG. 6 . FIG. 6 is a schematic diagram illustrating an image 600 captured by the camera 10 according to an embodiment of the present invention. In one embodiment, in order to make it easy for the camera 10 to find the location tags TAG1~TAG3, the numbers "1", "2" and "3" are respectively drawn on the location tags TAG1~TAG3 as the location tag patterns in advance. The image 600 includes these positioning label patterns.

In one embodiment, numbers are not necessarily drawn on the positioning tags TAG1~TAG3 as the positioning tag pattern, and any identifiable pattern can also be drawn on the positioning tags TAG1~TAG3 as the positioning tag pattern, such as red circle, green square, yellow Triangles... and so on.

In one embodiment, the processor 20 can train a deep learning network in advance, such as a convolutional neural network (Convolutional Neural Network, CNN), and the processor 20 can pre-train a plurality of images captured by the camera 10. The image is input into the deep learning network to train the deep learning network to recognize the position of the human body. After the pre-training of the deep learning network is completed, the processor 20 inputs the currently captured image into the deep learning network, and the deep learning network outputs an image 600 including the frame selection ranges F1 and F2 of human body positions.

In one embodiment, the processor 20 calculates a center point position P1, P2 of the bottom line of the frame selection range F1, F2, and then by tracking the center point position P1, P2, the movement of the center point position P1, P2 can be obtained. path, which represents the movement path of the human body.

In one embodiment, the processor 20 corresponds each vertex of the triangular plane (such as the triangular plane TR in FIG. 5 ) to the positions of these positioning label patterns (such as the positioning label pattern "1" shown in FIG. , "2" and "3"), the corresponding position of the triangular plane TR in the image 600 (ie, the triangular plane TR'') is obtained, and then the corresponding position of the triangular plane TR in the image 600 (ie, the triangular plane TR'') is obtained Plane TR'') is transformed onto the plan view 700.

In one embodiment, the camera 10 transmits the image 600 to the processor 20, and the processor 20 searches for the corresponding positioning tag patterns of the positioning tags TAG1 - TAG3 in the image 600 according to a known image recognition method.

In one embodiment, the positioning tags TAG1~TAG3 can be attached to the table, chairs, pillars, cabinets, etc., which are relatively easy to be identified at the beginning, and the processor 20 can be added to identify the corresponding positioning tags TAG1~TAG3 in the image 600. The accuracy of positioning label patterns.

The processor 20 corresponds the identified position of the positioning tag TAG1 in the space SP to the position of the positioning tag pattern "1", and corresponds the position of the positioning tag TAG2 in the space SP to the position of the positioning tag pattern "2", The position of the positioning tag TAG3 in the space SP corresponds to the position of the positioning tag pattern "3", in the image 600, it can be seen that the position of the positioning tag pattern is "1", the position of the positioning tag pattern is "2" and The positions where the positioning label pattern is "3" can be connected to form a triangular plane TR', and this triangular plane TR' is regarded as a specific area of the image.

Next, the processor 20 needs to actually map the triangular plane TR' to a plan view 700 in the room.

In step 330, the processor 20 converts the specific area of the image (such as the triangular plane TR') into a specific area of the plane of the plan view 700 (such as the triangular plane TR'').

Please refer to FIG. 7 , which is a schematic diagram illustrating a plan view 700 according to an embodiment of the present invention. In one embodiment, the objects in the plan view 700 are known, such as: all positioning label patterns, objects OBJ1 - OBJ5 , center point positions P1, P2.

Since the coordinates of all positioning label patterns are known in the plan view 700 (it can be a pre-set map data), the position of the positioning label pattern "1" in the image 600 can be corresponding to the position of the positioning label pattern "1" in the plan view 700 Position, the position of the positioning label pattern "2" in the image 600 corresponds to the position of the positioning label pattern "2" in the plan view 700, and the position of the positioning label pattern "3" in the image 600 corresponds to the positioning label in the plan view 700 The position where the pattern is "3".

In one embodiment, in the plan view 700, the position of the positioning label pattern "1", the position of the positioning label pattern "2" and the position of the positioning label pattern "3" can be connected into a triangular plane TR'', This means that the triangular plane TR'' is located within the range that the camera 10 can capture, and it is also possible to convert the specific area of the image captured by the camera 10 (such as the triangular plane TR') into a specific area of the plane on the plan view 700 (such as the triangular plane TR'). type plane TR'').

By repeating the above steps several times, moving the positions of multiple positioning tags TAG1~TAG3, or increasing the distance between the positioning tags TAG1~TAG3, the processor 20 can generate multiple triangular planes TR'', thereby establishing A complete plane specific area Ra is drawn, and the plane specific area Ra represents the position that can be photographed by the camera 10 in the plane view 700 .

In one embodiment, the complete plane specific region Ra is composed of a plurality of triangular planes TR''.

In one embodiment, the processor 20 converts a plurality of triangular planes TR' of different sizes and/or different positions into a plurality of triangular planes TR'' of different sizes and/or different positions of the plan view 700 to generate a complete Plane specific area Ra.

From the above, it can be seen that the triangular plane TR in Figure 5 is formed by connecting the three-dimensional coordinates of the three positioning tags TAG1~TAG3 located in the space SP, and the triangular plane TR in Figure 5 corresponds to that in Figure 6. A triangular plane TR' with two-dimensional coordinates can be generated in the image 600, and each vertex of the triangular plane TR' in FIG. TR''.

In other words, as long as the position of at least one of the positioning tags TAG1-TAG3 is moved, the processor 20 can obtain triangular planes TR'' of different sizes and/or positions, and repeat the above steps to establish a complete plane specific area Ra.

In one embodiment, when the complete plane specific area Ra is established, the processor 20 inputs the image 600 into a trained deep learning network, and the deep learning network is used to frame the range of the human body in the image. Then, the deep learning network outputs a frame selection range (such as frame selection range F1, F2), and the processor 20 calculates a center point position (such as center point position P1, P2) of the bottom line of the frame selection range, and tracks the center A moving path of the point position.

In one embodiment, the processor 20 can mark the triangular plane TR'' with color blocks on the plan view 700, so the entire color block can be regarded as a complete specific area Ra, and the complete specific area Ra represents the shooting of the camera 10 The range is converted onto the floor plan 700 .

It can be seen that the image conversion method 300 does not need to know any parameters and positions of multiple cameras, and the coverage of different lenses (such as fisheye cameras or bullet cameras) can be applied to establish multiple triangular planes without positioning or correction. TR'', present the shooting ranges of multiple identical or different cameras (that is, the complete plane specific area Ra) on the same plan view 700 .

In one embodiment, by repeating the above steps until the positioning tag pattern does not exist in the image 600, it means that at least a part of the position of the positioning tags TAG1-TAG3 has exceeded the shooting range of the camera 10, and there can only be ones that are not beyond the shooting range. Portions of a range can be converted to partial plan specific areas on the plan view 700 and represented by color patches.

When the boundaries of the shooting range of the camera 10 have been converted to the plan view 700, the entirety of the color blocks on the plan view 700 at this time (by a plurality of triangular planes TR'' of different sizes and/or positions after being colored) ) is regarded as a complete plane specific area Ra.

It can be seen that the position outside the specific area Ra of the complete plane is the position that the camera 10 has not photographed, and those positions that have not been photographed can be filled with one or more cameras (such as the camera 12) to make up for the entire plan view. 700 were photographed. Through the image conversion method in the above steps 310-340, only after the camera is installed, the correlation between the image captured by the camera and the floor plan 700 can be set from the image captured by the camera.

In one embodiment, since the complete plane specific area Ra has been obtained, multiple base stations AP1-AP4 and a positioning tag TAG can be removed, and then the processor 20 applies a known image recognition method, such as a convolutional neural network Identification algorithm, region-based convolutional neural network (Regions with CNN, R-CNN) algorithm, region-based fast convolutional neural network (Fast R-CNN), region-based faster convolutional neural network ( Faster R-CNN)... etc. to track the central point positions P1 and P2 in the image captured by the camera 10, and the movement of the central point positions P1 and P2, so as to know the moving path of the human body.

Please refer to FIG. 8 . FIG. 8 is a schematic diagram illustrating a moving path 800 of a human body according to an embodiment of the present invention. As mentioned above, the processor 20 uses the known image recognition method to track the central point positions P1 and P2 in the image captured by the camera 10 , and the movement of the central point positions P1 and P2 represents the moving path of the human body.

This situation can be applied to long-term care centers. Some elderly people with dementia sometimes know that they are going to the object OBJ1, but after walking for a while, they will forget where they are going, and they may walk randomly or It is walking around a circular path non-stop, this is the wandering state.

It can be seen in Figure 8 that the moving track L1 of the central point position P1 is a straight line, which means that the human body corresponding to the central point position P1 walks directly to the object OBJ1 (for example, the object OBJ1 is a dining table), and the moving track L2 of the central point position P2 is The curved and circuitous line segment means that the human body corresponding to the center point P2 walks to the object OBJ1 in a circuitous manner, and the human body corresponding to the center point P2 may have a wandering symptom of dementia.

The processor 20 can record the moving paths of the central point positions P1 and P2 by tracking the central point positions P1 and P2 in the image captured by the camera 10 and store them in a storage device (the storage device can be composed of a read-only memory, fast Flash memory, floppy disk, hard disk, optical disk, flash drive, magnetic tape, database accessible from the network or those familiar with this technology can easily think of storage media with the same function to realize it).

In this way, the caregiver can find that the elderly person corresponding to the center point P2 is in a state of wandering by observing the moving paths of the center point positions P1 and P2. The caregiver can choose whether to take further treatment according to the situation. For example, the caregiver can pay special attention to the elderly or notify the family.

The image conversion system and image conversion method shown in the present invention, no matter the images taken by the fisheye camera or the bullet camera, can convert these images to the same plan to know the coverage of the camera on the plan , each coverage area is composed of multiple triangular planes. The image conversion system and image conversion method shown in the present invention do not need to know any parameters and positions of multiple cameras, and the coverage areas of different lenses do not need to be positioned or corrected. Convert the shooting range of multiple identical or different cameras (that is, a specific area of a complete plane) to the same floor plan, greatly reducing the time spent on manpower when deploying cameras.

In addition, after the processor automatically converts the range captured by the camera to the plane map, the processor can automatically track the human body displacement in a specific area of the complete plane through image processing methods. This can be applied to long-term care centers, because the elderly generally have different Due to the habit of wearing electronic devices, the image conversion system and image conversion method shown in the present invention can use the center point position (that is, the position of the human body) marked in the specific area of the complete plane to track the moving path of the elderly, which is helpful It is used to judge whether the elderly have dementia.

The methods of the present invention, or specific forms or parts thereof, may exist in the form of program codes. The code may be contained in a physical medium, such as a floppy disk, compact disc, hard disk, or any other machine-readable (such as computer-readable) storage medium, or a computer program product without limitation in external form, wherein, When the program code is loaded and executed by a machine, such as a computer, the machine becomes a device for participating in the present invention. Code may also be sent via some transmission medium, such as wire or cable, optical fiber, or any type of transmission in which when the code is received, loaded, and executed by a machine, such as a computer, that machine becomes the Invented device. When implemented on a general-purpose processing unit, the code combines with the processing unit to provide a unique device that operates similarly to application-specific logic circuits.

Although the present invention has been disclosed above in terms of implementation, it is not intended to limit the present invention. Anyone skilled in this art can make various changes and modifications without departing from the spirit and scope of the present invention. Therefore, the protection of the present invention The scope shall be defined by the appended patent application scope.

100: Image conversion system 10, 12: Camera 20: Processor AP1~AP4: base station TAG0~TAG3: positioning tag 200: Image conversion system 300: Image conversion method 310~330: Steps SP: space TR, TR’, TR’’: triangular plane F1, F2: frame selection range P1, P2: center point position OBJ1~OBJ5: Objects Ra: Specific areas of the plane 700: floor plan 800:Movement path L1, L2:Movement track

FIG. 1 is a block diagram of an image conversion system according to an embodiment of the present invention. FIG. 2 is a block diagram illustrating an image conversion system according to an embodiment of the present invention. FIG. 3 is a schematic diagram illustrating an image conversion method according to an embodiment of the present invention. FIG. 4 is a schematic diagram illustrating an indoor positioning method according to an embodiment of the present invention. FIG. 5 is a schematic diagram illustrating an indoor positioning method according to an embodiment of the present invention. FIG. 6 is a schematic diagram illustrating an image captured by a camera according to an embodiment of the present invention. Fig. 7 is a schematic diagram showing a plan view according to an embodiment of the present invention. FIG. 8 is a schematic diagram illustrating a moving path of a human body according to an embodiment of the present invention.

300: Image conversion method

310~330: Steps

Claims (4)

An image conversion system, comprising: a camera, used to shoot an image; wherein, the image includes a plurality of positioning label patterns; a processor, used to receive the image, and identify the positioning label patterns according to the positioning The tag pattern generates a specific area of an image, and converts the specific area of the image into a specific area of a plane of a plan; and an indoor positioning device, which is used to transmit a signal with a plurality of base stations respectively through a plurality of positioning tags, and the base stations After receiving the signals, the strengths of the signals are sent to the processor, and the processor uses a triangulation algorithm to calculate the corresponding relative positions of the positioning tags in a space; wherein, the processor obtains The three positioning tags are in three different relative positions in the space, or the person who obtains one of the positioning tags moves to different positions in the space three times to obtain three different relative positions. Three different relative positions are connected to obtain a triangular plane transformed into the space; wherein, the processor corresponds to each vertex of the triangular plane to the positions of the positioning label patterns to obtain the triangular plane The plane is at the corresponding position in the image, and the corresponding position of the triangular plane in the image is converted to the plan view. Such as the image conversion system of claim 1, wherein a specific area of a complete plane is composed of a plurality of triangular planes, and when the specific area of the complete plane is established, the processor inputs the image into a trained deep learning network , the deep learning network outputs a frame selection range, the processor calculates a center point position of the bottom line of the frame selection range, and tracks a movement path of the center point position. An image conversion method, comprising: shooting an image; wherein, the image includes a plurality of positioning label patterns; receiving the image, identifying the positioning label patterns, and generating a specific area of the image according to the positioning label patterns; The specific area of the image is converted to a specific area of the plane; through a plurality of positioning tags, a signal is transmitted to a plurality of base stations, and after receiving the signals, the base stations transmit the strength of these signals to a A processor, the processor uses a triangulation algorithm to calculate the respective relative positions of the positioning tags in a space; obtain three different relative positions of the positioning tags in the space, or obtain the One of the positioning tags is moved three times to different positions in the space to obtain three different relative positions; connecting the three different relative positions to obtain a triangular plane transformed into the space; And after corresponding the vertices of the triangular plane to the positions of the positioning label patterns, the corresponding position of the triangular plane in the image is obtained, and then the corresponding position of the triangular plane in the image is converted to the plane view superior. Such as the image conversion method of claim 3, wherein a specific area of a complete plane is composed of a plurality of triangular planes, and when the specific area of the complete plane is established, the image is input into a trained deep learning network, the depth The learning network outputs a frame selection range; calculates a center point position of the bottom line of the frame selection range; and tracks a movement path of the center point position.

Images