Peer-to-peer pre-negotiation adaptive image transmission system and method

Abstract

The application discloses a point-to-point pre-negotiation adaptive image transmission system and method, and the method steps of the image transmission are as follows: S1, a request stage, a sending end A sends a communication request, and a receiving end B carries parameters to give a response; S2, a test stage, the sending end A sends test packets in groups to all sending ends; S3, a negotiation stage, the receiving end B obtains communication parameters through algorithm processing and informs the sending end A; and S4, a communication stage, the sending end A sends image data to the receiving end B. The application optimizes the performance and robustness of the image transmission.

Description

Technical Field

[0001] This invention relates to the field of image transmission technology, and more particularly to a point-to-point pre-negotiation adaptive image transmission system and method. Background Technology

[0002] In the field of point-to-point image transmission, existing industrial camera transmission protocols such as GigEVision, CoaXPress, and CameraLink, while performing well in specific application scenarios, generally suffer from problems such as fixed transmission parameters, lack of adaptability, fixed packet header format, and insufficient buffer congestion control. This results in significant differences in transmission performance under different network environments, and a large number of retransmissions are prone to occur when the buffer is congested, affecting transmission efficiency and reliability. Summary of the Invention

[0003] Based on the technical problems existing in the background technology, this invention proposes a point-to-point pre-negotiation adaptive image transmission system and method, which optimizes the performance and robustness of image transmission.

[0004] The point-to-point pre-negotiation adaptive image transmission method proposed in this invention comprises the following steps:

[0005] S1: Request phase, sender A sends a communication request, receiver B responds with parameters;

[0006] S2: During the testing phase, sender A sends test packets until all packets have been sent.

[0007] S3: During the negotiation phase, receiver B obtains communication parameters through algorithm processing and informs sender A.

[0008] S4: During the communication phase, the sending end A sends image data to the receiving end B.

[0009] Preferably, the algorithm processing steps in S3 are as follows:

[0010] S31: Abstracting comprehensive performance indicators Y 0-model and equivalent dimension total index Y 1. Model;

[0011] S32: Calculating the overall index based on the entropy weight method Y 1;

[0012] S33: Comprehensive performance index of least squares fitting method Y 0;

[0013] S34: Gradient Ascent Method for Obtaining Comprehensive Performance Indicators Y The optimal value is 0, and communication is conducted using the MTU and IFG values ​​at this time.

[0014] Preferably, the comprehensive performance indicators in S31Y Model 0 is:

[0015]

[0016]

[0017]

[0018]

[0019] In the formula, Weighted by transmission efficiency; For delayed weights; Weights based on processing capacity; For transmission efficiency; The total amount of data to be transmitted; This represents the total amount of data transmitted. This represents the total number of Ethernet frames. The maximum transmission unit for application layer data packets; This represents the total overhead of Ethernet protocols outside the application layer. This refers to the size of the application layer packet header. Total delay; Total transmission time; Total processing time; The total Ethernet frame interval (IFG) time; Transmission time for each Ethernet frame; Processing time for each Ethernet frame; This is the minimum time interval between Ethernet frames; This represents the number of Ethernet frames that can be processed per second.

[0020] Preferably, the overall index of the price dimension in S31 Y Model 1 is:

[0021]

[0022] In the formula, For throughput; Packet loss rate; Weights for throughput; The weight for packet loss rate.

[0023] Preferably, in step S32, the total index is calculated based on the entropy weight method. Y The steps for method 1 are as follows:

[0024] S321: Using the initial frame interval IFG as the initial value, increment the value by bytes successively, and perform several value tests to obtain the throughput and packet loss rate;

[0025] S322: For n samples and m indicators, let This represents the value of the j-th indicator for the i-th sample, and the value is normalized.

[0026] S323: Calculate the weight of the j-th indicator sample value for that indicator in the normalized data.

[0027] S324: Calculate the entropy value of the j-th index. ;

[0028]

[0029] In the formula, ,satisfy ;

[0030] S325: Calculate the weight of each indicator by using its entropy value, thereby determining the overall indicator. Y The expression for 1.

[0031]

[0032] Preferably, the fitting comprehensive performance index in S33 Y The steps for method 0 are as follows:

[0033] S331: Based on the overall indicators Y The expression for 1 is calculated for a given sample. Y The value of 1 will Y 1 Assigned to Y 0;

[0034] S332: Based on comprehensive performance indicators Y 0 got , , and Y The mapping relationship between 0 and 0 is constructed from the statistical samples to form a data matrix. and target vector Y 0;

[0035] S333: When At that time, the specific values ​​of the weights are obtained, and the comprehensive performance index is achieved. Y Fit to 0.

[0036] Preferably, the comprehensive performance indicators in S34 Y The method for finding the optimal value of 0 is as follows: Select an initial MTU value. and initial IFG time Set the learning rate Perform gradient ascent updates to maximize Y 0:

[0037]

[0038]

[0039]

[0040]

[0041] Repeat the above steps until... Y 0. Convergence or the maximum number of iterations is reached, and communication is conducted using the MTU and IFG values ​​at that time.

[0042] Preferably, the communication phase in S4 employs a sliding blocking window, adaptive variable frame storage, and cumulative response.

[0043] Preferably, the workflow of the sliding blocking window is as follows:

[0044] Sending end A continuously sends data packets according to the negotiated frame interval IFG. Receiving end B returns an acknowledgment signal for each data packet received. The acknowledgment signal contains the secondary sequence number of the expected next data packet.

[0045] When sender A receives the expected response signal from receiver B, it continuously pushes the sliding blocking window forward according to the sequence number, thereby allowing sender A to send new data packets.

[0046] When sender A receives two identical expected response packets, it enters the retransmission logic for data packet readback or reassembly. At this time, the sliding blocking window stops sliding and performs the blocking function. The sequential transmission logic of data packets automatically enters PAUSE until the retransmission ends, and then continues sequential transmission.

[0047] Preferably, the workflow for cumulative responses is as follows:

[0048] A frame of image is divided into several data payload packets. First, the header packet of the frame is transmitted, then several data payload packets are transmitted in sequence, and finally the tail packet is transmitted. Each packet has its own secondary sequence number. The first-level sequence number indicates the current data block to which it belongs, and the second-level sequence number indicates which data packet it belongs to in the current data packet.

[0049] The data payload packets are acquired by the FPGA from the front-end image sensor and stored in the DDR SDRAM. The header and tail packets are generated by the FPGA according to the current transmitted data blocks. The sending end A receives the acknowledgment signal from the receiving end B. The acknowledgment signal contains the secondary sequence number that the receiving end B expects to receive. If the sending end A receives two identical expected acknowledgment packets in a row, it reads back the data at the specific address from the DDR SDRAM, reassembles the packets, and sends them. If the expected acknowledgment packet is an expected retransmission of the header or tail packet, it reassembles the packets and retransmits them.

[0050] Beneficial technical effects of the present invention:

[0051] (1) This invention can automatically adjust the values ​​of MTU and IFG according to the network conditions and the responsiveness of the terminal device during the communication. Under high latency or unstable network conditions, the protocol adopts a more conservative data transmission strategy; while in a high-speed and stable network environment, it adopts more aggressive optimization measures. This improves the adaptability and robustness of the image transmission protocol.

[0052] (2) The present invention can also avoid misjudgment leading to unnecessary retransmission in the event of lost response packets, and perform selective retransmission to improve transmission efficiency in the event of packet loss. Attached Figure Description

[0053] Figure 1 This is a flowchart of the point-to-point pre-negotiation adaptive image transmission method proposed in this invention;

[0054] Figure 2 This invention presents an application layer protocol based on a five-layer communication model.

[0055] Figure 3 This invention provides an FPGA implementation of the Ethernet protocol stack data flow direction.

[0056] Figure 4 This is a diagram of the application layer data packet structure proposed in this invention. Detailed Implementation

[0057] The present invention will be further explained below with reference to specific embodiments.

[0058] Reference Figure 2-4The image transmission protocol of this invention supports pre-negotiation between communicating parties before formal transmission to determine optimal transmission parameters (such as the application layer maximum transmission unit (MTU) and frame interval (IFG)). This process takes into account the actual network conditions and the capabilities of terminal devices, thus achieving more flexible and adaptable communication. Furthermore, the header information of application layer data packets can be dynamically adjusted according to different network conditions and transmission requirements. For example, when the MTU is small, a complex verification mechanism may not be necessary; while when the MTU is large, a more robust verification function, such as 16-bit or 32-bit CRC checksum, is added to ensure data accuracy and reliability. The protocol also employs a sliding blocking window to control the data flow, adjusting the transmission rate through acknowledgment signals from the receiving end. This not only effectively avoids network congestion but also ensures that data packets arrive in order, reducing the possibility of error retransmissions. Multiple DDR SDRAM chips are used to store image data, and efficient data transmission is achieved through window blocking and automatic pause mechanisms. When a data packet loss is detected, it can be quickly retransmitted while maintaining the integrity and real-time performance of the image data. Furthermore, the adaptive variable frame storage mechanism stores each image frame as a data block in a single DDR SDRAM. The FPGA, acting as the camera (transmitter A), has multiple DDR SDRAMs and sends image data to the host (receiver B). Transmitter A receives acknowledgment signals from receiver B and continuously pushes the sliding blocking window forward according to the sequence number, allowing the transmitter to send new data packets. When transmitter A receives two identical expected acknowledgment packets, the sliding blocking window stops sliding, automatically pausing the transmission logic and entering the retransmission logic for data packet readback or reassembly. This avoids the significant bandwidth waste caused by cumulative packet loss due to continuous data output, improving bandwidth utilization. The protocol utilizes entropy weighting, least squares, and gradient descent methods for test value design, weight calculation, and parameter selection, ensuring optimal performance under various network conditions.

[0059] Specifically, refer to Figure 1 The point-to-point pre-negotiation adaptive image transmission method proposed in this invention has the following steps:

[0060] S1: Request phase, sender A sends a communication request, receiver B responds with parameters;

[0061] S2: During the testing phase, sender A sends test packets until all packets have been sent.

[0062] S3: During the negotiation phase, receiver B obtains communication parameters through algorithm processing and informs sender A.

[0063] S4: During the communication phase, the sending end A sends image data to the receiving end B.

[0064] The steps of the algorithm in S3 are as follows:

[0065] Abstract comprehensive performance indicators Y 0-model and equivalent dimension total index Y Model 1

[0066] Overall performance indicators Y The 0 model is as follows:

[0067]

[0068]

[0069]

[0070]

[0071] In the formula, Weighted by transmission efficiency; For delayed weights; Weights based on processing capacity; For transmission efficiency; The total amount of data to be transmitted; This represents the total amount of data transmitted. This represents the total number of Ethernet frames. The maximum transmission unit for application layer data packets; This represents the total overhead of Ethernet protocols outside the application layer. This refers to the size of the application layer packet header. Total delay; Total transmission time; Total processing time; The total Ethernet frame interval (IFG) time; Transmission time for each Ethernet frame; Processing time for each Ethernet frame; This is the minimum time interval between Ethernet frames; This represents the number of Ethernet frames that can be processed per second.

[0072] Equivalence Dimension Overall Indicators Y The model is as follows:

[0073]

[0074] In the formula, For throughput; Packet loss rate; Weights for throughput; The weight for packet loss rate.

[0075] The overall index is calculated based on the weights derived from the entropy weight method. Y 1

[0076] In this embodiment, the test value strategy (MTU and IFG) is as follows: The initial test value of the application layer maximum transmission unit (MTU) is set to a multiple of 128 bytes, which is half the maximum value of the receiver's buffer. The MTU is gradually decreased in 128-byte increments, with the minimum value being 128 bytes. (The receiver's buffer is normally a maximum of 4096 bytes, ensuring that receiver B can buffer at least two data packets, avoiding waste of the receiver buffer and maximizing its utilization). The initial MTU value is calculated by receiver B at the beginning, determining the maximum multiple of 128 bytes contained in half the maximum value of the receiver's buffer, and sent to sender A. The sender uses this value for subsequent tests.

[0077] Choosing a data packet size that is a multiple of 128 bytes allows for better utilization of memory alignment and cache optimization, reducing memory access frequency and improving memory access speed. It also better utilizes cache lines, reducing cache miss rates. DMA controllers typically transmit data in fixed-size blocks, and 128 bytes is a common DMA transfer block size, which optimizes DMA transfers, reduces the number of transfers, improves transfer efficiency, and reduces CPU load. It also simplifies header overhead management and segmentation reassembly processes in the network protocol stack, reducing processing complexity and time. Since the maximum value that can be stored in the registers of the hardware circuit is a power of 2, it better utilizes the characteristics of the network processor and FPGA hardware accelerator, improving data processing speed and overall performance, thereby significantly reducing data processing latency and improving data transmission bandwidth utilization.

[0078] The Ethernet standard IEEE 802.3 specifies a minimum frame interval of 12 bytes (96 bits of time). This is to ensure that network devices have sufficient time to process the previous frame and prepare to receive the next. Modern network devices are typically designed with this standard in mind, with optimized processing time and buffer management, enabling efficient operation within the 12-20 byte range. Multiple studies and practical tests have shown that frame intervals within this range have a significant impact on network throughput and latency. Smaller frame intervals can improve throughput, but too small an interval increases the probability of collisions; larger frame intervals increase latency but can reduce device processing load and packet loss rate. 20 bytes, as a relatively large value, provides enough data points to evaluate the impact of different frame intervals on network performance, while avoiding excessively large intervals that lead to resource waste. Therefore, choosing a range of 12 to 20 bytes for testing is scientifically reasonable. Starting with a 12-byte frame interval (IFG), the interval is increased in increments of 1 byte, up to a maximum of 20 bytes. Two values ​​are tested for each group (M, g). 16 This is used to obtain throughput and packet loss rate. For n samples and m metrics, let... This represents the value of the j-th indicator for the i-th sample.

[0079] Because the dimensions of the various indicators are not consistent, a range transformation method is used for normalization before calculation to homogenize the heterogeneous indicators. Positive indicators such as throughput T and negative indicators such as packet loss rate L are treated differently. For convenience, the normalized data is still denoted as... .

[0080] Positive indicators:

[0081] Negative indicators:

[0082] After normalizing the test data, the weight of the j-th indicator sample value relative to that indicator is calculated sequentially:

[0083]

[0084] Calculate the entropy value of the j-th index, where, ,satisfy :

[0085]

[0086] The weights of each indicator are calculated using the information entropy of each indicator:

[0087]

[0088] From this, we can obtain The specific formula.

[0089] Least squares fitting comprehensive performance index Y 0

[0090] Within a certain range Y 0 and Y 1 can be considered equivalent, and based on the test results, it can be derived from... Calculate the value of a sample The value will Assign to Based on comprehensive performance indicators , can be obtained , , and The mapping relationship between them. A data matrix is ​​constructed from the statistical samples. and target vector .

[0091] when At that time, the specific value of the weight is obtained. The expression can make the sum of squared errors minimize.

[0092] Gradient Ascent Method for Obtaining Comprehensive Performance Indicators Y The optimal value of 0

[0093] At this point, the objective function has a specific form:

[0094]

[0095] Calculate the objective function with respect to and gradient:

[0096] in:

[0097]

[0098]

[0099] Select initial MTU value and initial IFG time Set the learning rate Perform gradient ascent updates to maximize :

[0100]

[0101]

[0102] Repeat the above steps until... Once convergence or the maximum number of iterations is reached, the MTU and IFG values ​​at that point are used for communication.

[0103] The image transmission protocol of this invention can automatically adjust its MTU and IFG values ​​based on the network conditions and the responsiveness of the terminal device during the communication. Under high-latency or unstable network conditions, the protocol employs a more conservative data transmission strategy; while in high-speed, stable network environments, it adopts more aggressive optimization measures. This improves the adaptability and robustness of the image transmission protocol.

[0104] Furthermore, the communication phase in S4 of this invention employs a sliding blocking window, adaptive variable frame storage, and cumulative response.

[0105] For a sliding blocking window, the window size is communicated to the receiver B during the negotiation phase. The window size is equal to the size of the receiver B's buffer. Within the maintained window capacity, the sender A can continuously send data packets according to the previously negotiated frame interval (IFG). Each time the receiver B receives a data packet, it returns an acknowledgment signal containing the secondary sequence number of the expected next data packet. The sender A accepts these acknowledgments and continuously pushes the sliding blocking window forward according to the sequence number, thus allowing the sender to send new data packets. When the sender A receives two identical expected acknowledgment packets, it enters the retransmission logic for data packet readback or reassembly. At this point, the sliding blocking window stops sliding and performs its blocking function. The sequential data packet transmission logic automatically enters PAUSE until the retransmission ends, then resumes sequential transmission. The blocking window limits the sender A from uncontrolled burst data transmission when the receiver B is congested, avoiding cumulative congestion in the flow channel.

[0106] For adaptive variable frame storage, an image frame is stored as a data block in a single DDR SDRAM. The FPGA, acting as the camera (transmitter A), has multiple DDR SDRAMs and sends image data to the host (receiver B). Transmitter A receives acknowledgment signals from receiver B and continuously pushes the sliding blocking window forward according to the sequence number, thus allowing the transmitter to send new data packets. When transmitter A receives two identical expected acknowledgment packets, it enters the retransmission logic for data packet readback or reassembly. At this point, the sliding blocking window stops sliding and performs its blocking function. The sequential data packet transmission logic automatically enters PAUSE until the retransmission ends, and then sequential transmission continues.

[0107] Sending end A transmits images to receiving end B. Frame-changing storage is implemented using window blocking (response with sequence number pushes the window forward) plus automatic pause, allowing for the retransmission of missing data. When sending end A receives the header packet ACK of the next frame of image data block from receiving end B, it indicates that receiving end B has completely received this image data block. The DDR SDRAM storing this data block can then be overwritten by new image sensor input data. During the retransmission logic at the camera end, due to the pause in sequential transmission logic, the data to be transmitted from the upstream image sensor is stored in the overwriteable DDR SDRAM. Multiple onboard DDR SDRAMs employ a sequential overwrite cyclic read / write strategy, avoiding image frame loss in the event of network transmission congestion, smoothing the image data flow from the front-end acquisition sensor to the receiving host, and ensuring the accuracy and integrity of the data acquired at the camera end.

[0108] For cumulative responses, a single image frame is stored as a data block in a DDR SDRAM. To transmit this complete data block, it is divided into several data payload packets. When transmitting a data block, the header packet is transmitted first, followed by the subsequent data payload packets, and finally the tail packet. Each of these three types of packets has its own secondary sequence number. The first-level sequence number indicates the current data block it belongs to, and the second-level sequence number indicates which data packet it belongs to within the current data packet.

[0109] The data payload packets are acquired by the FPGA from the front-end image sensor and stored in the DDR SDRAM. The header and tail packets are generated by the FPGA based on the current transmitted data blocks. The sending end A will receive the acknowledgment signal from the receiving end B. The acknowledgment signal contains the secondary sequence number that the receiving end B expects to receive. Once the sending end A receives two identical expected acknowledgment packets in a row, it reads back the data at the specific address from the DDR SDRAM, reassembles the packets, and resends them. If the expected acknowledgment packet is an expectation to retransmit the header or tail packet, it will reassemble the packets and retransmit them.

[0110] The cumulative acknowledgment mechanism ensures successful data packet reception at the receiving end B. In the event of packet loss, selective retransmission can be performed based on logical judgment, transmitting only the lost data packets to reduce unnecessary retransmissions. In the event of unexpected loss of the returned acknowledgment packet, the probability of false retransmissions is significantly reduced, avoiding unnecessary waste of network resources and improving bandwidth utilization. The receiving end B maintains an Expect Response Serial Number (ERSNREGISTER) register. Whenever the receiving end B receives a data packet, the ERSN register enters a check to determine whether to increment. Based on the ERSN register, the receiving end B returns an acknowledgment signal containing the secondary sequence number of the expected next data packet.

[0111] For receiver B, only if the data packets received by receiver B are received sequentially (e.g., data packets with sequence numbers T, T+1, and T+2 are received sequentially) will the expected acknowledgment sequence number maintained by register ERSN increment sequentially to T+3. The expected sequence numbers contained in the returned acknowledgment packet will be T+1, T+2, and T+3 sequentially. If data packets with sequence numbers T, T+2, and T+3 are received sequentially, and the data packet with sequence number T+1 is lost, the receiver's judgment process is analyzed as follows: When the expected acknowledgment sequence number maintained by register ERSN is received, it increments to T+1, and an acknowledgment packet with expected sequence number T+1 is returned. Immediately afterwards, receiver B receives the data packet with sequence number T+2, which is inconsistent with the expected data packet. Register ERSN does not increment, and an expected acknowledgment packet with expected sequence number T+1 is sent again. At the same time, receiver B stores the T+2 data packet. During the entire data block transmission process, the sender only needs to send the lost packet. At this point, receiver A will receive two identical response packets, determine that the data packet with sequence number T+1 has been lost, and therefore organize a retransmission.

[0112] For sender A, if a response packet returned by receiver B to sender A is unexpectedly lost, and the expected sequence number contained in the received response packet is not continuous, only the maximum sequence number needs to be considered. The packet loss mechanism for judging two identical response packets perfectly avoids the situation where a response packet is lost.

Claims

1. A point-to-point pre-negotiation adaptive image transmission method, characterized in that, The steps are as follows: S1: Request phase, sender A sends a communication request, receiver B responds with parameters; S2: During the testing phase, sender A sends test packets until all packets have been sent. S3: During the negotiation phase, receiver B obtains communication parameters through algorithm processing and informs sender A. S4: During the communication phase, sender A sends image data to receiver B; The steps of the algorithm in S3 are as follows: S31: Abstract the comprehensive performance index Y0 model and the equivalent dimension total index Y1 model; S32: Calculate the overall index Y1 based on the entropy weight method; S33: Least squares fitting comprehensive performance index Y0; S34: The gradient ascent method is used to obtain the optimal value of the comprehensive performance index Y0, and the MTU value and IFG value at this time are used for communication. The comprehensive performance index Y0 model in S31 is: In the formula, Weighted by transmission efficiency; For delayed weights; Weights based on processing capacity; For transmission efficiency; The total amount of data to be transmitted; This represents the total amount of data transmitted. This represents the total number of Ethernet frames. The maximum transmission unit for application layer data packets; This represents the total overhead of Ethernet protocols outside the application layer. This refers to the size of the application layer packet header. Total delay; Total transmission time; Total processing time; The total Ethernet frame interval (IFG) time; Transmission time for each Ethernet frame; Processing time for each Ethernet frame; This is the minimum time interval between Ethernet frames; The number of Ethernet frames that can be processed per second; The Y1 model for the overall price dimension in S31 is as follows: In the formula, For throughput; Packet loss rate; Weights for throughput; The weight for packet loss rate.

2. The point-to-point pre-negotiation adaptive image transmission method according to claim 1, characterized in that, The steps for calculating the overall index Y1 based on the entropy weight method in S32 are as follows: S321: Using the initial frame interval IFG as the initial value, increment the value by bytes successively, and perform several value tests to obtain the throughput and packet loss rate; S322: For n samples and m indicators, let This represents the value of the j-th indicator for the i-th sample, and the value is normalized. S323: Calculate the weight of the j-th indicator sample value for that indicator in the normalized data. S324: Calculate the entropy value of the j-th index. ; In the formula, ,satisfy ; S325: Calculate the weight of each indicator by the entropy value of each indicator, thereby determining the expression of the total indicator Y1; 。 3. The point-to-point pre-negotiation adaptive image transmission method according to claim 2, characterized in that, The steps for fitting the comprehensive performance index Y0 in S33 are as follows: S331: Calculate the value of Y1 for a certain sample based on the expression of the total index Y1, and assign Y1 to Y0; S332: Obtained from the comprehensive performance index Y0 , , The mapping relationship between Y0 and Y0 is constructed from the statistical samples to form a data matrix. and target vector Y0; S333: When At that time, the specific values ​​of the weights are obtained, and the comprehensive performance index Y0 is fitted.

4. The point-to-point pre-negotiation adaptive image transmission method according to claim 1, characterized in that, The method for finding the optimal value of the comprehensive performance index Y0 in S34 is as follows: Select an initial MTU value. and initial IFG time Set the learning rate Perform gradient ascent updates to maximize Y0: Repeat the above steps until Y0 converges or the maximum number of iterations is reached, and use the MTU value and IFG value at this time for communication.

5. The point-to-point pre-negotiation adaptive image transmission method according to claim 1, characterized in that, In the S4 communication phase, sliding blocking window, adaptive variable frame storage, and cumulative response are used.

6. The point-to-point pre-negotiation adaptive image transmission method according to claim 5, characterized in that, The workflow of a sliding blocking window is as follows: Sending end A continuously sends data packets according to the negotiated frame interval IFG. Receiving end B returns an acknowledgment signal for each data packet received. The acknowledgment signal contains the secondary sequence number of the expected next data packet. When sender A receives the expected response signal from receiver B, it continuously pushes the sliding blocking window forward according to the sequence number, thereby allowing sender A to send new data packets. When sender A receives two identical expected response packets, it enters the retransmission logic for data packet readback or reassembly. At this time, the sliding blocking window stops sliding and performs the blocking function. The sequential transmission logic of data packets automatically enters PAUSE until the retransmission ends, and then continues sequential transmission.

7. The point-to-point pre-negotiation adaptive image transmission method according to claim 5, characterized in that, The workflow for cumulative responses is as follows: A frame of image is divided into several data payload packets. First, the header packet of the frame is transmitted, then several data payload packets are transmitted in sequence, and finally the tail packet is transmitted. Each packet has its own secondary sequence number. The first-level sequence number indicates the current data block to which it belongs, and the second-level sequence number indicates which data packet it belongs to in the current data packet. The data payload packets are acquired by the FPGA from the front-end image sensor and stored in the DDR SDRAM. The header and tail packets are generated by the FPGA based on the current transmitted data blocks. The sending end A receives the acknowledgment signal from the receiving end B. The acknowledgment signal contains the secondary sequence number that the receiving end B expects to receive. If the sending end A receives two identical expected acknowledgment packets in succession, it reads back the data at the specific address from the DDR SDRAM, reassembles the packets, and resends them. If the expected acknowledgment packet is an expected retransmission of the header or tail packet, it reassembles the packets and retransmits them.

Images