Dedicated font design quality judgment method and system oriented to long-distance scene
The construction of a special font design quality evaluation model through deep learning network models solves the problem of difficult to guarantee font visual recognition in long-distance complex environments, and realizes intelligent judgment and scientific guidance on the quality of special font design in long-distance scenarios, improving the accuracy and reliability of information transmission.
Patent Information
- Application Number
- CN202411755426.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-03
- Publication Date
- 2025-03-21
AI Technical Summary
In complex environments from a long distance, traditional font design methods are difficult to ensure that fonts maintain high visual recognition under different light, weather and perspective conditions, resulting in an impact on traffic efficiency and safety.
The deep learning network model is adopted to build a special font design quality evaluation model and use preprocessed historical special font images for training to achieve intelligent judgment on the quality of special font design in long-distance scenarios.
It realizes accurate and objective evaluation of the quality of special font design in long-distance scenarios, provides scientific design guidance, and significantly improves the accuracy and reliability of special fonts in long-distance scenarios.
Smart Images
Figure CN119672732A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of computer vision and font design, and more specifically, to a method for determining the quality of a special font design in a long-distance scene and a system for determining the quality of a special font design in a long-distance scene. Background Art
[0002] In long-distance scenarios, special font design is crucial. With the continuous expansion and improvement of transportation infrastructure, such as the extension of highways and urban expressways, the demand for long-distance information communication has increased greatly. The license plate recognition of road vehicle intelligent monitoring and recording systems and the guidance of long-distance traffic signs all rely on clear and recognizable fonts. However, traditional fonts are greatly affected by environmental factors at long distances, such as strong light, rainy and foggy weather, and different viewing angles, which can easily cause font blur and reduced recognition. This not only affects traffic efficiency, but may also cause traffic safety accidents. Therefore, special fonts designed for long-distance visual recognition scenarios can improve their visibility and stability through special glyphs, structures, styles, and specific strokes, and functional designs that serve long-distance visual recognition from the whole to the details.
[0003] Currently, there are major challenges in designing special fonts required for long-distance scenes. The key problem is the lack of effective design guidance methods based on long-distance visual effect evaluation. For a long time, the traditional font design process has mainly relied on the personal experience and subjective judgment of designers. Designers rely on their own artistic qualities and past design practices to determine the strokes, structure, style and other elements of the font. However, this approach exposes obvious limitations when facing the quality issues of font design in long-distance and complex environments.
[0004] The long-distance environment is extremely complex and contains a variety of influencing factors. Light conditions vary, from strong direct sunlight to dim cloudy environments, to night lighting. Different light intensities and angles will cause fonts to produce different degrees of reflection, shadows, and contrast changes. Weather conditions also have a significant impact on font visibility. For example, in rainy and foggy weather, water vapor will blur the line of sight and reduce the clarity of the font; on snowy days, accumulated snowflakes may block the font. In addition, the difference in viewing angles cannot be ignored. Drivers, law enforcement officers, and license plate automatic reading equipment may observe the font from different angles such as the front and oblique side, which requires the font to maintain good visibility from various viewing angles. Under the combined effect of these complex factors, it is difficult for traditional design methods to accurately optimize font design and ensure that the font still has high visibility in complex environments at long distances.
[0005] At present, deep learning technology is developing rapidly in the field of image analysis, which has brought new ways to solve font design for visibility problems. Deep learning has a powerful feature learning ability. It can process massive amounts of image data and explore the relationship between hidden font features and long-distance visibility. Using this technology to build an intelligent judgment model can open up new and effective ways for special font design to deal with long-distance visibility problems. Summary of the invention
[0006] In response to the above-mentioned problems existing in the prior art, the present invention provides a method and system for determining the quality of special font design in long-distance scenarios. By utilizing a deep learning network model, the quality of special font design in long-distance scenarios can be accurately and objectively determined intelligently, and comprehensive and scientific guidance can be provided for special font design based on the determination results.
[0007] As a first aspect of the present invention, a method for determining the quality of a special font design in a long-distance scenario is provided, and the method for determining the quality of a special font design in a long-distance scenario comprises the following steps: Step S1: Acquire a historical special font image in a long-distance scene; Step S2: preprocessing the historical special font image to obtain a preprocessed historical special font image; Step S3: constructing a special font design quality assessment model, and using the pre-processed historical special font images to train the special font design quality assessment model to obtain a trained special font design quality assessment model; Step S4: obtaining a special font image to be determined in a long-distance scene, and inputting the special font image to be determined into the trained special font design quality assessment model for assessment, so as to output a font visibility quality assessment result of the special font image to be determined.
[0008] Furthermore, the acquiring of the historical special font image in the long-distance scene further includes: Collecting a plurality of images containing special fonts from various actual long-distance scenes by means of high-resolution photographic equipment, and using the collected plurality of images containing special fonts as historical special font images in the long-distance scenes; Among them, various actual long-distance scenarios include different distance ranges, different weather conditions, different lighting conditions, and different observation perspectives.
[0009] Furthermore, the preprocessing of the historical special font image to obtain the preprocessed historical special font image further includes: graying the historical special font image to obtain a grayed historical special font image, while highlighting the contrast between the font and the background in the grayed historical special font image; Performing size normalization processing on the grayscale historical special font image to obtain a normalized historical special font image; The normalized historical special font image is enhanced and denoised to obtain the pre-processed historical special font image.
[0010] Furthermore, the method of training the special font design quality assessment model using the preprocessed historical special font image to obtain a trained special font design quality assessment model further includes: Dividing the preprocessed historical special font images into a training set, a validation set and a test set; The special font design quality assessment model is trained using the training set. During the training process of the special font design quality assessment model, the performance index of the special font design quality assessment model is monitored using the validation set. When the performance index of the special font design quality assessment model reaches stability or begins to decrease, the training of the special font design quality assessment model is stopped to obtain the trained special font design quality assessment model. After the special font design quality assessment model is trained, the test set is used to test and evaluate the performance of the trained special font design quality assessment model.
[0011] Furthermore, the training of the special font design quality assessment model using the training set further includes: The dedicated font design quality assessment model includes multiple convolutional layers, multiple pooling layers and multiple fully connected layers, wherein a ReLU activation function is used in the convolutional layers and the fully connected layers, and Dropout regularization is applied in the fully connected layers; Inputting the preprocessed historical special font image in the training set into the convolution layer to extract the font multi-dimensional features of the preprocessed historical special font image; Downsampling the font multi-dimensional features extracted by the convolution layer through the pooling layer to output the downsampled font multi-dimensional features; The downsampled font multi-dimensional features output by the pooling layer are integrated and mapped through the fully connected layer to output a font visibility quality assessment result of the preprocessed historical special font image.
[0012] Furthermore, the method for determining the quality of a special font design in a long-distance scenario further includes: According to the font visibility quality evaluation result of the special font image to be determined, specific guidance suggestions for special font design are generated.
[0013] As a second aspect of the present invention, a system for determining the quality of a special font design in a long-distance scenario is provided, and the system for determining the quality of a special font design in a long-distance scenario comprises the following steps: An acquisition module, used to acquire historical special font images in long-distance scenes; A processing module, used for preprocessing the historical special font image to obtain a preprocessed historical special font image; A training module, used for constructing a special font design quality assessment model, and training the special font design quality assessment model using the preprocessed historical special font images to obtain a trained special font design quality assessment model; The evaluation module is used to obtain the special font image to be determined in a long-distance scene, and input the special font image to be determined into the trained special font design quality evaluation model for evaluation, so as to output the font visibility quality evaluation result of the special font image to be determined.
[0014] The method for determining the quality of special font design in long-distance scenarios provided by the present invention has the following beneficial effects: by using a deep learning network model, the quality of special font design in long-distance scenarios can be accurately and objectively determined intelligently, and comprehensive and scientific guidance can be provided for special font design based on the determination results; the method fully considers the influence of long-distance complex environmental factors on font visibility, which helps to significantly improve the accuracy and reliability of information transmission of special fonts in long-distance scenarios, effectively avoids information misunderstanding or omission caused by poor font visibility, and is widely applicable to fields such as traffic roads and checkpoints that require clear identification of fonts at a long distance. BRIEF DESCRIPTION OF THE DRAWINGS
[0015] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. Together with the following specific embodiments, they are used to explain the present invention, but do not constitute a limitation of the present invention.
[0016] Figure 1 The present invention provides a flowchart of a method for determining the quality of a special font design in a long-distance scenario.
[0017] Figure 2 This is a structural diagram of the special font design quality assessment model provided by the present invention.
[0018] Figure 3 A schematic diagram of a first historical special font image provided by the present invention.
[0019] Figure 4 This is a schematic diagram of a second historical special font image provided by the present invention. DETAILED DESCRIPTION
[0020] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the specific implementation method, structure, characteristics and effects of a special font design quality determination method for long-distance scenes proposed by the present invention are described in detail below in combination with the accompanying drawings and preferred embodiments. Obviously, the described embodiments are part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0021] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so as to describe the embodiments of the present invention described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.
[0022] In this embodiment, a method for determining the quality of a special font design in a long-distance scenario is provided. Figure 1 As shown, the method for determining the quality of a special font design in a long-distance scenario includes the following steps: Step S1: Acquire a historical special font image in a long-distance scene; Preferably, the step of acquiring the historical special font image in the long-distance scene further includes: Using high-resolution photography equipment to collect a number of images containing special fonts from various actual long-distance scenes, so as to ensure that the images can clearly reflect the presentation status of the fonts in a complex environment; and using the collected images containing special fonts as historical special font images in the long-distance scenes; Among them, various actual long-distance scenes include different distance ranges (5 meters, 10 meters, 20 meters, 30 meters, 40 meters), different weather conditions (sunny, cloudy, rainy, foggy), different lighting conditions (direct light, scattered light, backlight, side light) and different observation perspectives (looking straight, squinting, looking up, looking down).
[0023] It should be noted that high-performance image acquisition equipment, including but not limited to high-resolution telephoto cameras, variable-focus surveillance cameras, etc., can adapt to the image acquisition needs of different distances, angles, lighting and weather conditions in long-distance scenes. These devices have the function of automatically adjusting parameters, such as focal length, aperture, exposure time, white balance, etc., to ensure that the quality of the collected images meets the requirements of subsequent analysis and can accurately reflect the visual state of the special fonts in actual long-distance environments.
[0024] In an embodiment of the present invention, a telephoto camera installed at a high position is used to capture images of special fonts within a distance range of 5 to 40 meters. The captured images contain various font styles and designs.
[0025] Step S2: preprocessing the historical special font image to obtain a preprocessed historical special font image; Preferably, the preprocessing of the historical special font image to obtain the preprocessed historical special font image further includes: Graying the historical special font image to obtain a grayed historical special font image, thereby reducing data dimension and computational complexity; and highlighting the contrast between the font and the background in the grayed historical special font image; Normalizing the size of the grayscale historical special font image to obtain a normalized historical special font image to meet the input requirements of the deep learning network model; The normalized historical special font image is enhanced and denoised to obtain the pre-processed historical special font image, wherein image enhancement techniques such as histogram equalization are used to enhance the overall contrast of the image, sharpening is used to enhance the edge clarity of the font, and denoising is used to reduce noise interference in the image, making the font features more prominent.
[0026] In the embodiment of the present invention, the collected image is first grayed to adjust the pixel value range of the image to a gray value of 0 to 255. Then, the image size is normalized to 224x224 pixels to meet the model input requirements. Next, the histogram equalization algorithm is used to enhance the overall contrast of the image, the sharpening filter is used to enhance the font edge, and the denoising method such as the median filter is used to reduce the noise in the image, making the font features clearer.
[0027] Step S3: constructing a special font design quality assessment model, and using the pre-processed historical special font images to train the special font design quality assessment model to obtain a trained special font design quality assessment model; Preferably, the using the pre-processed historical special font images to train the special font design quality assessment model to obtain a trained special font design quality assessment model further includes: Dividing the preprocessed historical special font images into a training set, a validation set and a test set; The special font design quality assessment model is trained using the training set. During the training process of the special font design quality assessment model, the performance index of the special font design quality assessment model is monitored using the validation set. When the performance index of the special font design quality assessment model reaches stability or begins to decrease, the training of the special font design quality assessment model is stopped to obtain the trained special font design quality assessment model. After the special font design quality assessment model is trained, the test set is used to test and evaluate the performance of the trained special font design quality assessment model.
[0028] It should be noted that the preprocessed image dataset is divided into a training set, a validation set, and a test set. The dataset is divided in a ratio of 80% for training, 10% for validation, and 10% for testing. The constructed special font design quality assessment model is trained using the training set. The mean square error (MSE) is used as the loss function to measure the difference between the model prediction result and the actual annotated font visibility quality. The optimization algorithm used is Adam, and the parameters of the model are updated to minimize the loss function. During the training process, the validation set is used to monitor the performance of the model. The model's learning rate, number of iterations, and other hyperparameters are dynamically adjusted according to indicators such as the accuracy and loss value on the validation set to avoid overfitting of the model. When the performance indicators of the model reach stability or begin to decline, stop training.
[0029] Preferably, the step of training the special font design quality assessment model using the training set further includes: like Figure 2 As shown, the dedicated font design quality assessment model includes multiple convolutional layers, multiple pooling layers and multiple fully connected layers, wherein the ReLU activation function is used in the convolutional layers and the fully connected layers to increase the nonlinear processing capability of the model, so that the model can better fit the complex mapping relationship between font features and visual quality; and Dropout regularization is applied in the fully connected layers to reduce the overfitting risk of the model and improve the stability and robustness of the model; The preprocessed historical special font images in the training set are input into the convolution layer to extract the multi-dimensional font features of the preprocessed historical special font images; wherein, convolution kernels of different sizes are designed, wherein smaller convolution kernels (3x3, 5x5) are used to capture the detailed features of the font strokes, including the changes in the thickness of the strokes, the subtleties of the stroke shapes (such as the transition between straight lines and curves, the morphology of the stroke endpoints, etc.); larger convolution kernels (7x7, 9x9) are used to obtain the overall structural features of the font, such as the center of gravity position, the outline, and the relationship between the font and the background. Through multi-layer convolution operations, rich font features from local to overall are gradually extracted.
[0030] The multi-dimensional features of the font extracted by the convolution layer are downsampled by the pooling layer to output the downsampled multi-dimensional features of the font; wherein the pooling layer is set after the convolution layer, and the feature map is downsampled using a 2x2 pooling window using a maximum pooling method, thereby reducing the amount of data while retaining the main features of the font, improving the computational efficiency and generalization ability of the model, and preventing overfitting.
[0031] The downsampled font multi-dimensional features output by the pooling layer are integrated and mapped through the fully connected layer to output the font visibility quality assessment result of the preprocessed historical special font image. In the last fully connected layer of the special font design quality assessment model, a continuous value related to the font visibility quality is output as the font visibility quality assessment result. The font visibility quality assessment result is presented in the form of a score.
[0032] In an embodiment of the present invention, the special font design quality assessment model is a CNN model with 6 convolutional layers, 3 pooling layers and 3 fully connected layers. The convolution kernel sizes are 3x3, 3x3, 5x5, 5x5, 7x7, and 9x9, respectively, and the pooling layer adopts a maximum pooling method of 2x2. The ReLU activation function is used in the convolutional layer and the fully connected layer, and the Dropout regularization is applied in the fully connected layer, and the drop probability is set to 0.5. The 10,000 pre-processed images collected (8,000 of which are used for training, 1,000 for verification, and 1,000 for testing) are annotated, and the annotation method is a visual quality score of 0 to 100. The mean square error loss function and the Adam optimization algorithm are used for training, the training cycle is 200 epochs, the initial learning rate is set to 0.001, and the hyperparameters such as the learning rate are adjusted according to the loss value and accuracy of the verification set.
[0033] Step S4: obtaining a special font image to be determined in a long-distance scene, and inputting the special font image to be determined into the trained special font design quality assessment model for assessment, so as to output a font visibility quality assessment result of the special font image to be determined.
[0034] Preferably, the method for determining the quality of a special font design for a long-distance scene further includes: According to the font visibility quality assessment result of the special font image to be determined, specific guidance suggestions for special font design are generated. The specific guidance suggestions for special font design include the stroke thickness, stroke shape, center of gravity position, proportional relationship, outline, aspect ratio, etc. in guiding the special font design, which are presented to the font designer in the form of a report document to help optimize the special font design.
[0035] Specifically, for stroke thickness, if the design quality assessment score of thinner stroke fonts is low, the stroke thickness should be appropriately increased; for stroke shape, if it is found that certain complex curve shapes affect visibility, the stroke shape can be simplified; based on the evaluation results of the center of gravity position, the font design is adjusted to make the center of gravity more conducive to long-distance recognition, such as raising the center of gravity to increase the sense of lightness or lowering the center of gravity to enhance stability. For proportional relationships, such as the internal stroke ratio of characters and the spacing between characters and lines, optimization is performed based on model feedback to make the font clearer at a distance; based on the judgment of the outline and aspect ratio, a shape that is more conducive to long-distance recognition is selected, such as appropriately increasing the aspect ratio to make the font fuller and more prominent, thereby comprehensively improving the quality of visibility of special fonts in long-distance scenes.
[0036] In an embodiment of the present invention, the dedicated font design quality assessment model runs on a high-performance GPU cluster to ensure fast and efficient processing capabilities of the model.
[0037] In the actual evaluation of special font design, the newly designed special font image is obtained through the image acquisition module and processed by the data preprocessing module. Figure 3 The dedicated font image to be determined shown in the figure will Figure 3 The image of the special font to be judged is input into the trained special font design quality assessment model for assessment, and the output visual quality assessment score is 60 points. Based on the comparative analysis of the assessment score and historical data, it is found that Figure 3 The font strokes in the font are thick and some of the curves are complex, which affects the long-distance visual effect. After appropriately reducing the stroke thickness and simplifying the curve shape, we get Figure 4 The special font image shown is predicted by the model. Figure 4 The font visibility quality evaluation result of the dedicated font image shown is 92 points.
[0038] As another embodiment of the present invention, a system for determining the quality of a special font design in a long-distance scenario is provided, and the system for determining the quality of a special font design in a long-distance scenario comprises the following steps: An acquisition module, used to acquire historical special font images in long-distance scenes; A processing module, used for preprocessing the historical special font image to obtain a preprocessed historical special font image; A training module, used for constructing a special font design quality assessment model, and training the special font design quality assessment model using the preprocessed historical special font images to obtain a trained special font design quality assessment model; The evaluation module is used to obtain the special font image to be determined in a long-distance scene, and input the special font image to be determined into the trained special font design quality evaluation model for evaluation, so as to output the font visibility quality evaluation result of the special font image to be determined.
[0039] The method for determining the quality of special font design in long-distance scenarios provided by an embodiment of the present invention utilizes a deep learning network model to accurately and objectively perform intelligent determination on the quality of special font design in long-distance scenarios, and provide comprehensive and scientific guidance for special font design based on the determination results. The present invention fully considers the impact of long-distance complex environmental factors on font visibility, which helps to significantly improve the accuracy and reliability of information transmission of special fonts in long-distance scenarios, effectively avoids information misunderstanding or omission caused by poor font visibility, and is widely applicable to fields such as traffic roads and checkpoints that require clear identification of fonts at a long distance.
[0040] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any form. Although the present invention has been disclosed as a preferred embodiment as above, it is not used to limit the present invention. Any technician familiar with this profession can make some changes or modifications to equivalent embodiments of equivalent changes using the technical contents disclosed above without departing from the scope of the technical solution of the present invention. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention still falls within the scope of the technical solution of the present invention.
Claims
1. A method for determining the quality of a special font design in a long-distance scenario, characterized in that: The method for determining the quality of a special font design in a long-distance scenario comprises the following steps: Step S1: Acquire a historical special font image in a long-distance scene; Step S2: preprocessing the historical special font image to obtain a preprocessed historical special font image; Step S3: constructing a special font design quality assessment model, and using the pre-processed historical special font images to train the special font design quality assessment model to obtain a trained special font design quality assessment model; Step S4: obtaining a special font image to be determined in a long-distance scene, and inputting the special font image to be determined into the trained special font design quality assessment model for assessment, so as to output a font visibility quality assessment result of the special font image to be determined.
2. The method for determining the quality of a special font design in a long-distance scenario according to claim 1, characterized in that: The step of acquiring the historical special font image in the long-distance scene also includes: Collecting a plurality of images containing special fonts from various actual long-distance scenes by means of high-resolution photographic equipment, and using the collected plurality of images containing special fonts as historical special font images in the long-distance scenes; Among them, various actual long-distance scenarios include different distance ranges, different weather conditions, different lighting conditions, and different observation perspectives.
3. The method for determining the quality of a special font design in a long-distance scenario according to claim 1, characterized in that: The preprocessing of the historical special font image to obtain the preprocessed historical special font image further includes: graying the historical special font image to obtain a grayed historical special font image, while highlighting the contrast between the font and the background in the grayed historical special font image; Performing size normalization processing on the grayscale historical special font image to obtain a normalized historical special font image; The normalized historical special font image is enhanced and denoised to obtain the pre-processed historical special font image.
4. The method for determining the quality of a special font design in a long-distance scenario according to claim 1, characterized in that: The method of training the special font design quality assessment model by using the pre-processed historical special font image to obtain a trained special font design quality assessment model further includes: Dividing the preprocessed historical special font images into a training set, a validation set and a test set; The special font design quality assessment model is trained using the training set. During the training process of the special font design quality assessment model, the performance index of the special font design quality assessment model is monitored using the validation set. When the performance index of the special font design quality assessment model reaches stability or begins to decrease, the training of the special font design quality assessment model is stopped to obtain the trained special font design quality assessment model. After the special font design quality assessment model is trained, the test set is used to test and evaluate the performance of the trained special font design quality assessment model.
5. The method for determining the quality of a special font design in a long-distance scenario according to claim 4, characterized in that: The step of training the special font design quality assessment model using the training set further includes: The dedicated font design quality assessment model includes multiple convolutional layers, multiple pooling layers and multiple fully connected layers, wherein a ReLU activation function is used in the convolutional layers and the fully connected layers, and Dropout regularization is applied in the fully connected layers; Inputting the preprocessed historical special font image in the training set into the convolution layer to extract the font multi-dimensional features of the preprocessed historical special font image; Downsampling the font multi-dimensional features extracted by the convolution layer through the pooling layer to output the downsampled font multi-dimensional features; The downsampled font multi-dimensional features output by the pooling layer are integrated and mapped through the fully connected layer to output a font visibility quality assessment result of the preprocessed historical special font image.
6. The method for determining the quality of a special font design in a long-distance scenario according to claim 1, characterized in that: The method for determining the quality of a special font design in a long-distance scenario also includes: According to the font visibility quality evaluation result of the special font image to be determined, specific guidance suggestions for special font design are generated.
7. A system for determining the quality of a special font design for a long-distance scene, used to implement the method for determining the quality of a special font design for a long-distance scene as claimed in any one of claims 1 to 6, characterized in that: The dedicated font design quality determination system for long-distance scenarios includes: An acquisition module, used to acquire historical special font images in long-distance scenes; A processing module, used for preprocessing the historical special font image to obtain a preprocessed historical special font image; A training module, used for constructing a special font design quality assessment model, and training the special font design quality assessment model using the preprocessed historical special font images to obtain a trained special font design quality assessment model; The evaluation module is used to obtain the special font image to be determined in a long-distance scene, and input the special font image to be determined into the trained special font design quality evaluation model for evaluation, so as to output the font visibility quality evaluation result of the special font image to be determined.