A method of adjusting a visual feedback parameter for garment pressing

Abstract

The present application relates to the technical field of image processing and visual analysis, and discloses a kind of visual feedback parameter adjustment method of garment finishing.The present application obtains the surface image of garment after finishing, extracts four kinds of apparent characteristics, i.e.texture fluctuation response quantity, highlight concentration response quantity, structure edge continuity response quantity and area uniformity response quantity, and constructs apparent anomaly discrimination model to calculate the synergistic relationship of multiple characteristics, to obtain apparent discrimination value;Based on the discrimination value, the abnormal type of garment surface is determined, and the corresponding finishing parameter adjustment instruction is generated.The method realizes the identification of residual wrinkle, structural anomaly and over-ironing risk through coupling analysis of multi-dimensional visual features, and realizes the feedback regulation of finishing parameters based on the visual analysis result, thereby improving the finishing quality and stability.

Description

Technical Field

[0001] This invention relates to the field of image processing and visual analysis technology, and in particular to a method for adjusting visual feedback parameters in garment ironing. Background Technology

[0002] In the process of garment ironing, fabrics are usually treated with temperature, pressure and steam to eliminate wrinkles and improve surface smoothness. Existing technologies mostly rely on preset parameters or human experience for control, and lack the ability to conduct real-time quantitative analysis of the fabric's appearance. Although some solutions introduce image recognition technology to evaluate the ironing effect through texture, brightness or edge features, most of them are based on a single feature or simple threshold judgment, which is difficult to accurately reflect the real changes of the fabric during the ironing process.

[0003] Existing visual feedback solutions for garment ironing typically use surface anomalies in the ironing area as the basis for treating residual wrinkles. However, visual anomalies during ironing can originate from multiple sources, including actual residual wrinkles, steam-induced shine, short-term changes in reflectivity, structural undulations, and signs of over-ironing. Since different sources of anomalies correspond to different process responses, failure to differentiate between the sources can easily lead to incorrect adjustments of ironing parameters. This can result in areas unsuitable for further ironing being subjected to repeated heat or steam, leading to over-ironing, localized indentations, surface shine, or under-ironing. Therefore, there is an urgent need for a method to adjust garment ironing visual feedback parameters that can integrate multi-dimensional visual characteristics and establish a unified judgment criterion. Summary of the Invention

[0004] The technical problem to be solved by this invention is that existing methods lack the ability to model the collaborative relationship between multiple features, and the judgment process relies on multiple empirical thresholds, resulting in poor stability of the recognition results and difficulty in achieving effective feedback adjustment of ironing parameters. To this end, we propose a visual feedback parameter adjustment method for garment ironing.

[0005] To achieve the above objectives, this application adopts the following technical solution: a method for adjusting visual feedback parameters in garment ironing, comprising the following steps:

[0006] Step S1: Acquire the current frame image of the garment surface after ironing using a vision acquisition device, and perform grayscale preprocessing on the current frame image to obtain a grayscale image;

[0007] Step S2: Extract features from the grayscale image to obtain four normalized apparent response features: texture undulation response, which characterizes the degree of micro-wrinkle of the fabric; high brightness concentration response, which characterizes the degree of local reflective patch aggregation; structural edge continuity response, which characterizes the geometric coherence of the inherent seam and placket and crease contours; and regional uniformity response, which characterizes the macro-smoothness and uniformity of the surface. The values ​​of each feature are constrained within [0,1].

[0008] Step S3: Preset the appearance anomaly discrimination model, input the four normalized appearance response feature quantities into the appearance anomaly discrimination model, and calculate the discrimination value; The appearance anomaly discrimination model is constructed as a fractional structure with the combined effect of texture undulation response quantity and regional uniform response quantity as the numerator and the combined effect of highlight concentration response quantity and structural edge continuous response quantity as the denominator, and introduces power operation and exponential operation between feature quantities;

[0009] Step S4: Based on the comparison between the discriminant value and the preset benchmark threshold, and combined with the comparison between the highlighted concentrated response and the preset safety limit, determine the current abnormality type of the garment surface; where the preset benchmark threshold and the preset safety limit are both pre-calibrated constants, and the abnormality type includes at least residual wrinkle abnormalities, structural appearance abnormalities, transient appearance abnormalities, and overheating risk abnormalities.

[0010] Step S5: Based on the determined anomaly type, generate and execute a parameter adjustment instruction set for the ironing equipment actuator. The parameter adjustment instruction set is used to adjust at least one of the following working parameters: ironing temperature, ironing pressure, steam jet volume, and ironing mold dwell time.

[0011] Preferably, in step S1, when the visual acquisition device acquires the current frame image, the ironing mold of the ironing device has been raised and removed from the garment surface, and the residual steam on the garment surface has naturally dissipated to above the preset visibility threshold; the angle between the optical axis of the visual acquisition device and the normal of the garment surface is not greater than the preset tilt angle threshold, and diffuse reflection auxiliary light sources are symmetrically arranged on both sides of the visual acquisition device to eliminate the influence of directional specular reflection on image quality, thereby ensuring that the acquired image is not affected by instantaneous steam occlusion or specular reflection.

[0012] Preferably, in step S1, before performing grayscale preprocessing on the current frame image, the step of performing illumination normalization processing on the current frame image is further included: using a histogram equalization algorithm to eliminate the influence of uneven ambient illumination on the image brightness distribution, and performing Gaussian filtering on the equalized image to suppress high-frequency noise; the grayscale preprocessing uses a weighted average method to convert the color image into a grayscale image, and the weighting coefficients of each color channel are preset according to the hue characteristics of the clothing fabric to enhance the contrast between the fabric texture and the background.

[0013] Preferably, in step S2:

[0014] The method for extracting texture undulation response is as follows: a multi-directional filter bank is used to convolve the grayscale image, extract the maximum high-frequency response amplitude of each pixel in each direction, calculate the global average response energy, and use the S-shaped growth curve function to map the global average response energy to the closed interval [0,1].

[0015] The method for extracting the concentrated response of the highlight is as follows: statistically analyze the highlighted connected regions formed by pixels whose gray values ​​exceed the adaptive local threshold in the grayscale image, calculate the ratio of the area of ​​the largest highlighted connected region to the total area of ​​all highlighted connected regions, and then weight and fuse this ratio with the inverse of the centroid dispersion of the highlight pixel space. After normalization, the response is obtained in the closed interval [0,1].

[0016] The extraction method of continuous response quantity of structural edge is as follows: the binary edge map of grayscale image is extracted by edge detection operator, and isolated short line segments with length less than the preset pixel threshold are removed by morphological operation to obtain edge skeleton map. The ratio of the total length of edge skeleton map to the number of endpoints is calculated, and the ratio is mapped to the closed interval [0,1] by normalization function.

[0017] The method for extracting the uniform response of a region is as follows: the grayscale image is divided into a preset number of non-overlapping sub-regions, the standard deviation of pixel brightness in each sub-region is calculated, the mean and variance of the standard deviation of all sub-regions are statistically analyzed, and the joint statistic of the mean and variance is back-mapped to the closed interval [0,1] using the S-shaped growth curve function.

[0018] Preferably, the calculation and construction of the discriminant value in step S3 is specifically as follows:

[0019] The molecule is composed of the product of the first power term and the first exponent term. The first power term has the texture undulation response as the base and the region uniform response as the exponent. The first exponent term has the natural constant as the base and the difference between the texture undulation response and the highlight concentration response as the exponent.

[0020] The denominator is composed of a constant 1 and a second power term, with the second power term having the product of the highlight concentration response and the continuous response at the structure edge as the base and the continuous response at the structure edge as the exponent.

[0021] The discriminant value is equal to the quotient obtained by dividing the numerator by the denominator.

[0022] Preferably, the specific strategy for determining the anomaly type in step S4 is as follows:

[0023] A preset dead zone is introduced. The preset dead zone is a tolerance parameter used to define the neighborhood range of the preset benchmark threshold. The value of the preset dead zone is less than the smaller of the preset benchmark threshold and the preset safety limit. When the discrimination value is greater than the sum of the preset benchmark threshold and the preset dead zone, it is judged as a residual wrinkle anomaly.

[0024] When the discriminant value is less than the difference between the preset baseline threshold and the preset dead zone, it is judged as a structural appearance anomaly.

[0025] When the discrimination value falls within a closed numerical range centered on the preset benchmark threshold and bounded by the preset dead zone, and the concentrated response of the highlight exceeds the preset safety limit, it is judged as an abnormal risk of overheating.

[0026] When the discriminant value falls within a closed numerical range and the highlighted concentrated response is less than or equal to the preset safety limit, it is judged as a transient apparent anomaly.

[0027] Preferably, in step S5, the strategy for generating and executing the parameter adjustment instruction set is as follows:

[0028] If the abnormality is determined to be a residual wrinkle type, a first type of adjustment instruction is generated to control the increase of the heating temperature of the upper mold, the increase of the dehumidification vacuum degree of the lower mold, and the extension of the residence time of the hot mold in the current area;

[0029] If a transient apparent anomaly is determined, a second type of adjustment command is generated to control the suspension of steam supply, activation of auxiliary air cooling device, and increase the moving speed of the hot mold through the current area;

[0030] If the risk of overheating is determined to be abnormal, a third type of adjustment instruction is generated to control the cutting off of the heating power supply, drive the ironing mold to be lifted away from the fabric surface, and shut down the steam generating unit.

[0031] If a structural appearance abnormality is determined, a fourth type of adjustment instruction is generated to control the heat-pressing mold to perform pressure avoidance actions on the high-response coordinate area identified by the continuous response amount of the structural edge when executing the heat-pressing path, and to maintain the current heating temperature within the range not exceeding the preset structural protection temperature limit.

[0032] Preferably, after step S5 is completed, the following is also executed:

[0033] Step S6: After a preset delay, acquire the updated frame image of the garment surface after the ironing operation again, and repeat steps S1 to S5. If the discrimination values ​​obtained consecutively fall within the convergence closed interval [preset benchmark threshold - preset convergence tolerance, preset benchmark threshold + preset convergence tolerance] centered on the preset benchmark threshold and with the preset convergence tolerance as the radius, then it is determined that the ironing process has achieved the target effect and an end command is output; wherein the value of the preset convergence tolerance is less than the value of the preset dead zone.

[0034] Preferably, step S6 further includes setting a maximum loop iteration threshold; when the number of loop executions from step S1 to step S5 reaches the maximum loop iteration threshold, if the discrimination value still does not fall within the convergence closed interval, then convergence is determined to have failed, a manual intervention prompt signal is output and the automatic operation process is paused, and the current frame image and corresponding feature value are recorded for fault tracing.

[0035] Preferably, the preset baseline threshold, preset safety limit, preset dead zone, and adjustment amplitude parameters in the parameter adjustment instruction set used in steps S2 to S5 are all pre-mapped with the fabric type of the garment to be ironed. Before executing step S1, the fabric type of the garment to be ironed is obtained by reading the garment identification information or by a pre-identification algorithm, and the corresponding threshold parameter group and adjustment amplitude parameter group are retrieved from the preset mapping table according to the fabric type, so as to realize adaptive visual feedback parameter adjustment for different fabric characteristics.

[0036] The technical effects and advantages of this invention are as follows:

[0037] This invention simultaneously extracts texture undulation response, highlight concentration response, structural edge continuous response, and regional uniform response, and constructs an appearance anomaly discrimination model. Compared with existing methods based on single features or multiple thresholds, this invention can effectively distinguish residual wrinkles, structural appearance anomalies, transient appearance anomalies, and over-scalding risk anomalies, avoiding misadjustment of ironing parameters caused by confusion of anomaly types. At the same time, by directly mapping the discrimination results to adjustment commands for parameters such as ironing temperature, steam volume, and dwell time, and combining them with a feedback iteration mechanism, adaptive and complete control of the ironing process is achieved, thereby improving the stability of ironing quality and reducing the probability of over-scaling and under-scaling. Attached Figure Description

[0038] Fig. 1 This is an overall flowchart of the present invention;

[0039] Fig. 2 This is the timing control diagram of the present invention;

[0040] Fig. 3 This is a schematic diagram of the discriminant value change curve during the ironing process of the present invention. Detailed Implementation

[0041] It is readily understood that, based on the technical solution of this invention, those skilled in the art can propose various interchangeable structural methods and implementations without altering the essential spirit of the invention. Therefore, the following detailed embodiments and accompanying drawings are merely illustrative examples of the technical solution of this invention and should not be considered as the entirety of the invention or as limitations or restrictions on the technical solution of this invention.

[0042] Example 1

[0043] Reference Figs. 1-3 As shown in the figure, this embodiment describes in detail the complete implementation process of a visual feedback parameter adjustment method for garment ironing according to the present invention.

[0044] Step S1. Image Acquisition and Preprocessing:

[0045] In this embodiment, the visual acquisition device is an industrial camera, which is installed directly above the ironing mold of the ironing equipment. The angle between the camera's optical axis and the normal to the garment surface is controlled within a small angle. Diffuse auxiliary light sources are symmetrically arranged on both sides of the camera to eliminate the influence of directional specular reflection on image quality.

[0046] Before each image acquisition, the ironing mold of the ironing equipment is completely lifted off the garment surface, and after a short delay, the residual steam on the garment surface is allowed to dissipate naturally, ensuring that the acquired image is not affected by the instantaneous obstruction of steam or the reflection of the mirror; the camera acquires the current frame image of the garment surface after the ironing operation.

[0047] The current frame image is first subjected to illumination normalization: a histogram equalization algorithm is used to eliminate the influence of uneven ambient lighting on the image brightness distribution, followed by Gaussian filtering to suppress high-frequency noise; then grayscale preprocessing is performed to convert the color image into a grayscale image; for fabrics of different hues, different weighting coefficients can be selected during grayscale conversion to enhance the contrast between texture and background. After processing, a grayscale image is obtained, and each pixel value is normalized to the closed interval [0,1].

[0048] Step S2. Extraction of four apparent response features:

[0049] Four normalized apparent response features are extracted from the grayscale image obtained in step S1: texture undulation response. Highlighted concentrated response Continuous response at structural edges and regional uniform response The values ​​of each feature are all constrained within the closed interval [0,1].

[0050] Texture undulation response The extraction process is as follows: A multi-directional filter bank is used to convolve the grayscale image to extract the maximum high-frequency response amplitude of each pixel in each direction, forming a high-frequency response amplitude map; the global average response energy of this amplitude map is calculated, and the S-shaped growth curve function is used to map this global average response energy to the closed interval [0,1]. When the folds and undulations on the fabric surface are more significant, the high-frequency response energy is greater, and the high-frequency response energy obtained after mapping is... The closer the value is to 1, the smoother and flatter the surface. The value approaches 0.

[0051] Highlight Concentrated Response The extraction process is as follows: Pixels in the grayscale image whose brightness values ​​exceed the adaptive local threshold are counted, and these pixels are used to form a bright connected region; the ratio of the area of ​​the largest bright connected region to the total area of ​​all bright connected regions is calculated, and this ratio is combined with the concentration of the bright pixel spatial distribution. After normalization, the response value located in the closed interval [0,1] is obtained. This extraction method ensures that only locally dense reflective patches will cause [the problem]. The numerical value increases, while large-area uniform reflection is suppressed; The closer the value is to 1, the higher the concentration of bright patches.

[0052] Continuous response at structural edge The extraction process is as follows: An edge detection operator is used to extract the binary edge map of the grayscale image. Short, isolated line segments are removed through morphological operations to obtain the edge skeleton map. The ratio of the total length of the edge skeleton map to the number of endpoints is calculated, and this ratio is mapped to the closed interval [0,1] using a normalization function to obtain the result. When the edges are more continuous and the degree of fragmentation is lower, the ratio of total length to the number of endpoints is larger. The closer the value is to 1, the more severe the edge breakage and the presence of numerous isolated short line segments. The value approaches 0.

[0053] Uniform response of the region The extraction process is as follows: the grayscale image is divided into multiple non-overlapping sub-regions, and the standard deviation of pixel brightness in each sub-region is calculated. The mean and dispersion of the standard deviations of all sub-regions are statistically analyzed, and the above joint statistics are back-mapped to the closed interval [0,1] using an S-shaped growth curve function to obtain... When the surface is smoother and more uniform, the standard deviations of each sub-region are smaller and closer to each other, after mapping... The closer the value is to 1, the more likely it is that there are local inhomogeneities on the surface. The value decreased significantly.

[0054] Step S3. Calculation of the discriminant value:

[0055] The four normalized apparent response features extracted in step S2 are used as the basis for further analysis. , , and The input is fed into a preset apparent anomaly discrimination model, and the discrimination value is calculated according to the following formula. :

[0056] ;

[0057] In the formula, It is a natural constant.

[0058] The construction of the above-mentioned discriminant model has the following technical implications:

[0059] The numerator consists of the first power term. With the first exponent term Multiplication constitutes the first power term, which is the response of the texture undulation. Using the base as the uniform response quantity of the region It is an exponent. When the uniform response of the region... When it approaches 1, Approximately equal to The wrinkle feature contributes to the molecule according to its actual intensity; when the region has a uniform response... When the value decreases due to surface unevenness, due to the base value Located between 0 and 1, a decreasing exponent causes the power term to rise towards 1, thus automatically amplifying the contribution of the complex state of wrinkles accompanied by inhomogeneity to the discriminant value. The first exponent term is expressed as a natural constant. As the base, with texture undulation response amount Highlight Concentration Response The difference is the exponent. When Greater than When the exponent is positive, the exponent term is greater than 1, which amplifies the effect on the molecule; when Less than When the exponent is negative, the exponent term is less than 1, and it exerts a compressive effect on the molecules. Through this exponential construction, an asymmetric competitive response between wrinkle intensity and high brightness intensity is achieved.

[0060] The denominator consists of the constant 1 and the second power term. The terms are added together. The second power term is highlighted with a concentrated response value. Continuous response at structural edges The product of the two is used as the base, and the continuous response at the structural edge is used as the base. It is an exponential expression. When the continuous response at the structural edge... When the power term approaches 0, regardless of the base, the power term approaches 1, and the denominator remains near the basic level; when Approaching 1 and highlighting concentrated response quantity When the value is large, the base increases and the exponent approaches 1, causing the value of this term to increase rapidly, thus significantly increasing the denominator and suppressing the discriminant value. This construction achieves the function of "confirming and amplifying gloss anomalies at structural edges," meaning that the discriminant value is only strongly compressed when both the highlight and the structural edge are significant.

[0061] Step S4. Exception type determination:

[0062] Based on the discriminant value calculated in step S3 And the highlighted concentrated response extracted in step S2 Anomaly type determination is performed. The preset baseline threshold, preset safety limit, and preset dead zone used in the determination process are all pre-calibrated constants, and their specific values ​​can be calibrated according to the actual application scenario.

[0063] The determination strategy is as follows:

[0064] When discriminant value When the value exceeds the sum of the preset baseline threshold and the preset dead zone, it indicates that the texture fluctuation response is significant. With regional uniform response The combined intensity significantly suppresses the high-brightness concentrated response. Continuous response at structural edges The combined strength is then used to determine if there are residual wrinkles or other abnormalities on the surface of the garment.

[0065] When discriminant value When the difference between the preset baseline threshold and the preset dead zone is less than the value of the highlight concentration response, it indicates that the response value is concentrated. Continuous response at structural edges The combined strength significantly suppresses the texture undulation response. With regional uniform response The combined strength, and the continuous response at the structural edges. It plays a confirmatory role in the denominator, at which point it is determined that there is a structural appearance abnormality on the surface of the current garment.

[0066] When discriminant value The highlighted concentrated response value falls within a closed interval centered on a preset baseline threshold and bounded by a preset dead zone. When the value exceeds the preset safety limit, it indicates that the wrinkles and gloss are in a state of competitive equilibrium, but the concentration of local high gloss has exceeded the safety level. At this time, it is determined that there is an abnormal risk of overheating on the surface of the current garment.

[0067] When discriminant value The concentrated response value falls within the above closed interval and is highlighted. When the value is less than or equal to the preset safety limit, it indicates that wrinkles and gloss are in a state of competitive equilibrium and the concentration of high brightness has not exceeded the standard. At this time, it is determined that there is a transient appearance abnormality on the surface of the current garment. This abnormality is caused by instantaneous reflection of steam or moisture condensation artifacts, which are transient phenomena that can disappear on their own.

[0068] Step S5. Adjust ironing parameters:

[0069] Based on the exception type determined in step S4, generate and execute the corresponding parameter adjustment instruction set.

[0070] If the abnormality is determined to be a residual wrinkle, a first type of adjustment instruction is generated to control the increase of the heating temperature of the upper mold, the increase of the dehumidification vacuum of the lower mold, and the extension of the residence time of the hot stamping mold in the current area, so as to enhance the shaping effect on the wrinkled parts.

[0071] If a transient apparent anomaly is determined, a second type of adjustment command is generated to control the suspension of steam supply, the activation of auxiliary air cooling device, and the increase of the movement speed of the hot mold through the current area, so as to accelerate the evaporation of surface moisture and avoid water stains.

[0072] If an abnormal risk of overheating is detected, a third type of adjustment command is generated to control the cutting off of the heating power supply, drive the ironing mold to rise away from the fabric surface, and shut down the steam generating unit to prevent irreversible heat damage to the fabric.

[0073] If a structural appearance anomaly is determined, a fourth type of adjustment instruction is generated to control the heat-pressing mold during the execution of the heat-pressing path, specifically addressing the continuous response amount from the structural edges. The high-response coordinate area is marked to perform pressure avoidance action and maintain the current heating temperature within the range that does not exceed the preset structural protection temperature limit to avoid heat damage to the garment structure.

[0074] Step S6. Closed-loop convergence verification:

[0075] After step S5 is completed, after a preset delay, the updated frame image of the garment surface after the ironing operation is acquired again, and steps S1 to S5 are repeated. If the discrimination values ​​obtained consecutively are... If all values ​​fall within the convergence closed interval centered on the preset benchmark threshold and with the preset convergence tolerance as the radius, the ironing process is deemed to have achieved the target effect. An end command is then output to control the ironing equipment to enter the next work area or end the ironing task. The value of the preset convergence tolerance is less than the value of the preset dead zone to ensure that the convergence criterion is stricter than the anomaly criterion.

[0076] In addition, a maximum number of loop iterations is set; when the number of loop executions from step S1 to step S5 reaches this maximum number of loop iterations, if the discrimination value... If the convergence interval is still not found, the convergence is considered to have failed. A manual intervention prompt signal is output and the automatic operation process is paused. At the same time, the current frame image and the corresponding feature value are recorded for fault tracing.

[0077] Example 2

[0078] This embodiment, based on Embodiment 1, further describes how to achieve adaptive parameter adjustment for different fabric types without making any modifications to the discrimination model itself.

[0079] Before performing step S1, the fabric type information of the garment to be ironed is first obtained. The fabric type can be obtained by reading the fabric code information from the electronic tag attached to the garment, or by using an image classification algorithm to identify the fabric type after pre-scanning the garment surface.

[0080] For different fabric types, a mapping relationship between fabric type and judgment parameter set is established in advance. The judgment parameter set includes: the preset benchmark threshold, preset safety limit and preset dead zone used in step S4, and the adjustment range in the parameter adjustment instruction set used in step S5. For example, for pure cotton fabric, because it has good heat resistance and is prone to wrinkling, a relatively high heating temperature adjustment range and a long dwell time adjustment range can be set; for silk fabric, because it has poor heat resistance and is prone to aurora, a lower preset safety limit and a smaller heating temperature adjustment range can be set; for chemical fiber fabric, because it is prone to shrinkage when heated and prone to electrostatic adsorption, a lower structural protection temperature limit can be set.

[0081] Before executing step S1, the corresponding set of judgment parameters is retrieved from the preset mapping table according to the obtained fabric type, and the set of parameters is used for judgment and adjustment in subsequent steps S4 and S5. In this way, the present invention can achieve adaptive matching of different fabric characteristics without changing the structure and calculation formula of the core discrimination model.

[0082] Example 3

[0083] This embodiment describes how to generate a discriminant value using the method of the present invention. Quantitatively evaluate the completion of ironing.

[0084] In garment ironing production lines, it is often necessary to quantify and score the ironing quality of each garment in order to enable quality traceability and process optimization. In this embodiment, the discrimination values ​​obtained in step S6 are used to... The average value or final stable value is used as a quantitative indicator of the ironing completion.

[0085] When the final discriminant value The response value is stable near the preset benchmark threshold and the highlighted concentrated response value is stable. When the value is below the preset safety limit, it indicates that the ironing process has fully eliminated wrinkles without introducing the risk of over-ironing, and the ironing completion is rated as excellent; when the discrimination value is below the preset safety limit, it indicates that the ironing process has fully eliminated wrinkles without introducing the risk of over-ironing, and the ironing completion is rated as excellent. Although it converges to the convergence interval, the highlighted concentrated response is still high. If the ironing temperature exceeds the preset safety limit at certain stages, it indicates a short-term risk of overheating during the ironing process, but this risk is ultimately controlled. In this case, the ironing completion rating is good. When the discrimination value... If the ironing effect fails to converge after reaching the maximum number of iterations, or if it converges but the convergence value deviates significantly from the preset benchmark threshold, it indicates that the ironing effect is not up to standard. In this case, the ironing completion evaluation is poor, and a manual re-inspection process is triggered.

[0086] Using the above quantitative evaluation method, this embodiment will determine the discriminant value. The criteria for determining anomalies have been expanded to include continuous metrics for ironing quality, providing data support for the refined management and optimization of the ironing process.

[0087] The technical scope of this invention is not limited to the content described above. Those skilled in the art can make various modifications and variations to the above embodiments without departing from the technical concept of this invention, and all such modifications and variations should fall within the protection scope of this invention.

Claims

1. A method for adjusting visual feedback parameters in garment ironing, characterized in that, Includes the following steps: Step S1: Acquire the current frame image of the garment surface after ironing using a visual acquisition device, and perform grayscale preprocessing on the current frame image to obtain a grayscale image; Step S2: Extract features from the grayscale image to obtain four normalized apparent response features: texture undulation response, which characterizes the degree of micro-wrinkle of the fabric; high brightness concentration response, which characterizes the degree of local reflective patch aggregation; structural edge continuity response, which characterizes the geometric coherence of the inherent seam and placket and crease contours; and regional uniformity response, which characterizes the macro-smoothness and uniformity of the surface. The values ​​of each feature are constrained within [0,1]. Step S3: Preset the apparent anomaly discrimination model, input the four normalized apparent response feature quantities into the apparent anomaly discrimination model, and calculate the discrimination value; The apparent anomaly discrimination model is constructed as follows: a fractional structure with the combined effect of texture undulation response and regional uniform response as the numerator and the combined effect of high-brightness concentrated response and structural edge continuous response as the denominator, and power and exponential operations between feature quantities are introduced. Step S4: Based on the comparison relationship between the discrimination value and the preset benchmark threshold, and combined with the comparison relationship between the high-brightness concentrated response quantity and the preset safety limit, determine the current abnormality type of the garment surface; wherein, the preset benchmark threshold and the preset safety limit are both pre-calibrated constants, and the abnormality type includes at least residual wrinkle abnormalities, structural appearance abnormalities, transient appearance abnormalities, and overheating risk abnormalities. Step S5: Based on the determined anomaly type, generate and execute a parameter adjustment instruction set for the ironing equipment actuator. The parameter adjustment instruction set is used to adjust at least one of the following working parameters: ironing temperature, ironing pressure, steam injection volume, and ironing mold dwell time.

2. The method for adjusting visual feedback parameters in garment ironing according to claim 1, characterized in that, In step S1, when the visual acquisition device acquires the current frame image, the ironing mold of the ironing device has been raised and removed from the garment surface, and the residual steam on the garment surface has naturally dissipated to above the preset visibility threshold; the angle between the optical axis of the visual acquisition device and the normal of the garment surface is not greater than the preset tilt angle threshold, and diffuse reflection auxiliary light sources are symmetrically arranged on both sides of the visual acquisition device to eliminate the influence of directional specular reflection on image quality, thereby ensuring that the acquired image is not affected by instantaneous steam obstruction or specular reflection.

3. The method for adjusting visual feedback parameters in garment ironing according to claim 1, characterized in that, In step S1, before performing grayscale preprocessing on the current frame image, the step of performing illumination normalization processing on the current frame image is also included: using a histogram equalization algorithm to eliminate the influence of uneven ambient illumination on the image brightness distribution, and performing Gaussian filtering on the equalized image to suppress high-frequency noise; the grayscale preprocessing uses a weighted average method to convert the color image into a grayscale image, and the weighting coefficients of each color channel are preset according to the hue characteristics of the clothing fabric to enhance the contrast between the fabric texture and the background.

4. The method for adjusting visual feedback parameters in garment ironing according to claim 1, characterized in that, In step S2: The extraction method of the texture undulation response is as follows: the grayscale image is convolved using a multi-directional filter bank to extract the maximum high-frequency response amplitude of each pixel in each direction, the global average response energy is calculated, and the global average response energy is mapped to the closed interval [0,1] using the S-shaped growth curve function. The extraction method of the high-brightness concentrated response is as follows: count the high-brightness connected regions formed by pixels whose gray values ​​exceed the adaptive local threshold in the grayscale image, calculate the ratio of the area of ​​the largest high-brightness connected region to the total area of ​​all high-brightness connected regions, and then weight and fuse this ratio with the inverse of the centroid dispersion of the high-brightness pixel space. After normalization, the response is obtained in the closed interval [0,1]. The extraction method of the continuous response of the structure edge is as follows: the binary edge map of the grayscale image is extracted by using an edge detection operator, and the edge skeleton map is obtained after removing isolated short line segments with a length less than a preset pixel threshold by morphological operation. The ratio of the total length of the edge skeleton map to the number of endpoints is calculated, and the ratio is mapped to the closed interval [0,1] using a normalization function. The method for extracting the uniform response of the region is as follows: the grayscale image is divided into a preset number of non-overlapping sub-regions, the standard deviation of pixel brightness in each sub-region is calculated, the mean and variance of the standard deviation of all sub-regions are statistically analyzed, and the joint statistic of the mean and variance is back-mapped to the closed interval [0,1] using the S-shaped growth curve function.

5. The method for adjusting visual feedback parameters in garment ironing according to claim 1, characterized in that, The calculation and construction of the discriminant value in step S3 is specifically as follows: The molecule is composed of a first power term and a first exponent term multiplied together. The first power term is based on the texture undulation response amount and the region uniform response amount. The first exponent term is based on the natural constant and the difference between the texture undulation response amount and the highlight concentration response amount is used as the exponent. The denominator is formed by adding a constant and a second power term, where the second power term has the product of the high-concentration response and the continuous response at the structural edge as the base and the continuous response at the structural edge as the exponent. The discriminant value is equal to the quotient obtained by dividing the numerator by the denominator.

6. The method for adjusting visual feedback parameters in garment ironing according to claim 5, characterized in that, The specific strategy for determining the anomaly type in step S4 is as follows: A preset dead zone is introduced, which is a tolerance parameter used to define the neighborhood range of the preset benchmark threshold, and the value of the preset dead zone is less than the smaller of the preset benchmark threshold and the preset safety limit. When the discrimination value is greater than the sum of the preset baseline threshold and the preset dead zone, it is determined to be a residual wrinkle type anomaly; When the discrimination value is less than the difference between the preset benchmark threshold and the preset dead zone, it is determined to be a structural appearance anomaly. When the discrimination value falls within a closed numerical range centered on the preset benchmark threshold and bounded by the preset dead zone, and the high-brightness concentrated response is greater than the preset safety limit, it is determined to be an abnormal risk of overheating. When the discrimination value falls within the closed numerical range and the highlight concentration response is less than or equal to the preset safety limit, it is determined to be a transient apparent anomaly.

7. The method for adjusting visual feedback parameters in garment ironing according to claim 6, characterized in that, In step S5, the strategy for generating and executing the parameter adjustment instruction set is as follows: If the abnormality is determined to be a residual wrinkle type, a first type of adjustment instruction is generated to control the increase of the heating temperature of the upper mold, the increase of the dehumidification vacuum degree of the lower mold, and the extension of the residence time of the hot mold in the current area; If a transient apparent anomaly is determined, a second type of adjustment command is generated to control the suspension of steam supply, activation of auxiliary air cooling device, and increase the moving speed of the hot mold through the current area; If the risk of overheating is determined to be abnormal, a third type of adjustment instruction is generated to control the cutting off of the heating power supply, drive the ironing mold to be lifted away from the fabric surface, and shut down the steam generating unit. If a structural appearance abnormality is determined, a fourth type of adjustment instruction is generated to control the heat-pressing mold to perform a pressure avoidance action on the high-response coordinate area identified by the continuous response amount of the structure edge when executing the heat-pressing path, and to maintain the current heating temperature within the range not exceeding the preset structural protection temperature limit.

8. The method for adjusting visual feedback parameters in garment ironing according to claim 7, characterized in that, After step S5 is completed, the following is also executed: Step S6: After a preset delay, acquire the updated frame image of the garment surface after the ironing operation again, and repeat steps S1 to S5. If the discrimination value obtained in multiple consecutive operations falls within the convergence closed interval [preset reference threshold - preset convergence tolerance, preset reference threshold + preset convergence tolerance] centered on the preset reference threshold and with the preset convergence tolerance as the radius, then it is determined that the ironing process has achieved the target effect and an end command is output. The value of the preset convergence tolerance is less than the value of the preset dead zone.

9. The method for adjusting visual feedback parameters in garment ironing according to claim 8, characterized in that, In step S6, a maximum loop iteration threshold is set. When the number of loop executions from step S1 to step S5 reaches the maximum loop iteration threshold, if the discrimination value still does not fall within the convergence closed interval, convergence is determined to have failed, a manual intervention prompt signal is output and the automatic operation process is paused, and the current frame image and corresponding feature value are recorded for fault tracing.

10. The method for adjusting visual feedback parameters in garment ironing according to claim 9, characterized in that, The preset baseline threshold, preset safety limit, preset dead zone, and adjustment amplitude parameters in the parameter adjustment instruction set used in steps S2 to S5 are all pre-mapped with the fabric type of the garment to be ironed. Before executing step S1, the fabric type of the garment to be ironed is obtained by reading the garment identification information or a pre-identification algorithm, and the corresponding threshold parameter group and adjustment amplitude parameter group are retrieved from the preset mapping table according to the fabric type to achieve adaptive visual feedback parameter adjustment for different fabric characteristics.

Images