A Machine Vision-Based Method for Monitoring the Operation of Food Waste Disposal Equipment

Abstract

This invention relates to the field of operational monitoring technology for kitchen waste treatment equipment, and discloses a machine vision-based method for monitoring the operation of kitchen waste treatment equipment. This method uses the door sealing ring as the monitoring object, establishes visual sampling coordinates, divides the area into circumferential zones and radial layers, and collects dark field, light transmission reference, normal sealing benchmark, and operational images. After grayscale normalization, benchmark and operational data are generated. Then, the upper limit of normal light leakage in the circumferential direction and the normal sealing compression benchmark are constructed. The amount of light leakage, compression deviation, sealing anomaly score, and judgment boundary are calculated, and the sealing status, anomaly location, anomaly type, operational permit identifier, and monitoring record are output. This method transforms the continuous sealing ring into a zoned and locatable detection object, reducing the influence of lighting and equipment differences, and solving the problem that traditional door lock detection struggles to identify local light leakage, compression anomalies, and anomaly locations, facilitating operation control and maintenance troubleshooting.

Description

Technical Field

[0001] This invention relates to the field of operation monitoring technology for kitchen waste treatment equipment, specifically a method for monitoring the operation of kitchen waste treatment equipment based on machine vision. Background Technology

[0002] Food waste treatment equipment is typically used to crush, mix, dry, ferment, and deodorize water-containing organic waste such as kitchen scraps, fruit and vegetable waste, and leftover food. During operation, the treatment chamber may contain moisture, oil, odorous gases, liquid leachates, and splashing particles. Therefore, the airtightness of the doors directly affects the equipment's operational safety, odor suppression effectiveness, chamber pressure stability, and the hygiene of the surrounding environment. Existing food waste treatment equipment typically uses mechanical door locks, reed switches, microswitches, Hall effect sensors, and pressure switches to determine whether the doors are closed; some equipment also indirectly determines whether the equipment is in an abnormal operating state by using operating parameters such as motor current, temperature, and fan status.

[0003] Existing hatch detection methods primarily determine whether the hatch has reached a preset closed position, but struggle to assess whether the hatch sealing ring has achieved continuous and effective compression. When the sealing ring is partially aged, deformed, contaminated with oil, or traps vegetable scraps or small debris, the hatch latch may still be closed, and the mechanical switch may output a normal signal, but tiny gaps have already formed in the sealing ring. These tiny gaps are difficult to detect when the equipment is stationary, but during stirring, heating, drying, or deodorizing operations, they can easily create channels for odor leakage, moisture leakage, and waste splashing, affecting the equipment's sealing performance and operational reliability. Furthermore, existing visual inspection methods are mostly focused on waste type identification, feed quantity identification, overflow detection, or equipment appearance inspection, typically relying on overall image judgment and failing to provide precise monitoring of the continuous circumferential sealing state of the hatch sealing ring. Even when using a camera to observe the hatch area, it is easily affected by ambient light, material reflection, local shadows, lens installation deviations, and differences in the sealing ring's color, leading to unstable detection results. Existing methods lack the technical means to uniformly quantify the degree of micro-leakage, the degree of compression deviation, and the location of abnormalities after dividing the sealing ring into circumferential partitions and radial layers, making it difficult to output reproducible and locatable sealing abnormality results.

[0004] Therefore, this case aims to propose a machine vision-based method for monitoring the operation of kitchen waste disposal equipment. Taking the door sealing ring as the visual monitoring object, the sealing ring is transformed from a normal image area into structured visual data with circumferential orientation, radial hierarchy, and sampling order. Then, through normalization processing between dark field images, light transmission reference images, normal sealing reference images, and operating images, the changes in light leakage and compression of the sealing ring in different orientations are extracted. Furthermore, sealing anomaly scores, judgment boundaries, anomaly orientation sets, anomaly type labels, and operation permit identifiers are generated. Summary of the Invention

[0005] This invention provides a machine vision-based method for monitoring the operation of kitchen waste disposal equipment, which helps to solve the problems mentioned in the background art.

[0006] This invention provides the following technical solution: a method for monitoring the operation of kitchen waste disposal equipment based on machine vision, comprising: Select the door sealing ring area, establish visual sampling coordinates, divide the circumferential partitions and radial layers, and generate circumferential sampling angles and sealing ring sampling points; Acquire dark field images, light-transmitted reference images, normal sealed reference images, and cabin door sealing ring operation images; read the grayscale of the sealing ring sampling points; and generate normal reference normalized grayscale and operation normalized grayscale through normalization processing. The normalized grayscale of the normal reference is processed circumferentially to generate the upper limit of normal light leakage and the normal sealing and pressing reference. The normalized grayscale is compared with the upper limit of normal circumferential light leakage in layers to generate circumferential micro-leakage, average light leakage over the whole circle and local peak light leakage. The running pressure degree is generated based on the normalized grayscale and compared with the normal sealing pressure benchmark to generate the pressure deviation, the average pressure deviation of the whole circle and the local peak pressure deviation. The door sealing anomaly score is generated based on the mean and peak values ​​of light leakage and compression deviation, and the sealing judgment boundary is generated based on the normalized grayscale of the normal baseline, the upper limit of normal light leakage in the circumferential direction, and the normal sealing compression baseline. Based on the circumferential micro-leakage and compression deviation, a circumferential anomaly positioning quantity is generated. The hatch sealing anomaly score is compared with the sealing judgment boundary to generate a hatch sealing status identifier. In the non-sealed state, anomaly circumferential partition number, anomaly circumferential partition center angle, circumferential angular distance and anomaly orientation set are generated. In the sealed state, a number indicating that there is no anomaly circumferential partition and an empty anomaly orientation set are generated. Anomaly type labels and operation permit labels are generated based on the hatch sealing status indicator, local peak light leakage amount, and local peak compression deviation amount, and operation monitoring records are output.

[0007] Optionally, the step of selecting the hatch sealing ring area, establishing visual sampling coordinates, dividing the area into circumferential zones and radial layers, and generating circumferential sampling angles and sealing ring sampling points specifically includes: The door sealing ring area will be used as the monitoring target for the operation of the food waste disposal equipment. Set the center point of the sealing ring as the origin of the coordinate system, set the direction of the door hinge pointing to the latch as the positive direction of the horizontal coordinate, and set the direction of the image plane after rotating 90 degrees counterclockwise from the positive direction of the horizontal coordinate as the positive direction of the vertical coordinate. The sealing ring area is divided into seventy-two circumferential partitions and five radial layers. A set of indexes for circumferential partitions, a set of indexes for radial layers, and a set of indexes for normal sealing reference acquisition are established. The set of indexes for circumferential partitions contains consecutive integers from one to seventy-two, the set of indexes for radial layers contains consecutive integers from one to five, and the set of indexes for normal sealing reference acquisition contains consecutive integers from one to six. Each integer in the circumferential partition index set is used as the circumferential partition number, each integer in the radial hierarchical index set is used as the radial hierarchical number, and each integer in the normal closed benchmark acquisition index set is used as the normal closed benchmark acquisition number. Measure the inner boundary radius and outer boundary radius of the sealing ring, divide the radial width between the outer boundary radius and the inner boundary radius of the sealing ring into five equal-width radial intervals, and use the middle radius of each radial interval as the sampling radius of the corresponding radial layer; The full circle angle corresponding to twice pi is divided into seventy-two circumferential zones with equal angles, and the sampling angles corresponding to each circumferential zone are formed according to the order of the circumferential zone numbers; For each radial layer and each circumferential partition, corresponding sealing ring sampling points are generated according to the corresponding sampling radius, sampling angle, direction cosine of the sampling angle in the positive direction of the horizontal coordinate, and direction sine of the sampling angle in the positive direction of the vertical coordinate.

[0008] Optionally, the acquisition of dark-field images, transmitted light reference images, normal sealing reference images, and cabin door sealing ring operation images, reading the grayscale of the sealing ring sampling points, and generating normal reference normalized grayscale and operation normalized grayscale through normalization processing specifically includes: Set a fixed zero-prevention constant, the value of which is 10 to the power of negative 6; With the circumferential detection light source turned off, dark field images are acquired when the hatch is closed and the equipment is not feeding materials. The dark field images are then normalized to form a dark field normalized grayscale image. The grayscale value of each sealing ring sampling point in the dark field normalized grayscale image is read, and the upper limit of the read grayscale value is clipped. The upper limit of the clipping is the value after subtracting a fixed anti-zero constant from one, thus forming the dark field grayscale of the corresponding sealing ring sampling point. Turn on the circumferential detection light source and acquire a light transmission reference image when the hatch is open to the point where the sealing ring is unobstructed. Perform grayscale normalization processing on the light transmission reference image to form a light transmission reference normalized grayscale image. Read the grayscale value of each sealing ring sampling point in the light transmission reference normalized grayscale image and perform lower limit cropping on the read grayscale value. The lower limit of cropping is the value of the dark field grayscale of the corresponding sealing ring sampling point plus a fixed anti-zero constant, to form the light transmission reference grayscale of the corresponding sealing ring sampling point. With the hatch clean, the sealing ring free of debris, and the door lock fully engaged, six normal airtightness benchmark acquisitions are performed consecutively. Each acquisition forms a normal airtightness benchmark image. The gray values ​​of the sealing ring sampling points corresponding to each radial layer and each circumferential partition in each normal airtightness benchmark image are read to form the normal airtightness benchmark sampling gray values. After the food waste equipment enters the operation monitoring state, the operation images of the door sealing ring are collected according to the discrete sampling time. The gray values ​​of the corresponding sealing ring sampling points of each radial layer and each circumferential partition in each door sealing ring operation image are read to form the operation sampling gray value. For each discrete sampling moment, each radial layer and each circumferential partition, the dark field gray level of the corresponding sealing ring sampling point is subtracted from the running sampling gray level to form the running gray level difference. The dark field gray level is subtracted from the light-transmitting reference gray level of the corresponding sealing ring sampling point to form the reference gray level difference. The result of dividing the running gray level difference by the reference gray level difference is clipped from zero to one to form the running normalized gray level. For each normal sealed baseline acquisition, each radial layer, and each circumferential partition, the dark field gray level of the corresponding sealing ring sampling point is subtracted from the gray level of the normal sealed baseline sampling point to form a baseline gray level difference. The dark field gray level is subtracted from the light-transmitting reference gray level of the corresponding sealing ring sampling point to form a reference gray level difference. The result of dividing the baseline gray level difference by the reference gray level difference is clipped from zero to one to form the normal baseline normalized gray level.

[0009] Optionally, the circumferential processing of the normalized grayscale to generate the circumferential normal light leakage upper limit and the normal sealing pressure reference specifically includes: For each circumferential partition and each normal closed baseline acquisition, the normal baseline normalized gray values ​​corresponding to the five radial layers are summed, and the summation result is divided by five to form the circumferential normalized gray value mean of the corresponding circumferential partition under the normal closed baseline acquisition. For each circumferential zone, the circumferential normalized grayscale mean values ​​corresponding to the six normal sealed baseline acquisitions are compared, and the highest value is taken as the upper limit of normal circumferential light leakage of the corresponding circumferential zone under normal sealed conditions. For each circumferential zone, the normalized gray values ​​of the six normal sealing reference acquisitions and the five radial layer corresponding normal references are summed. The summation result is divided by 30 to form the mean normalized gray value of the circumferential comprehensive reference for the corresponding circumferential zone. The mean normalized gray value of the circumferential comprehensive reference for the corresponding circumferential zone is then subtracted by 1 to form the normal sealing and pressing reference for the corresponding circumferential zone. The upper limit of normal light leakage and the normal sealing and pressing benchmark of each circumferential zone are fixed as the calculation benchmark for this operation monitoring of the kitchen waste equipment.

[0010] Optionally, the step of comparing the normalized grayscale with the upper limit of normal circumferential light leakage in layers to generate circumferential micro-leakage, average light leakage over the entire circle, and local peak light leakage specifically includes: Construct positive out-of-limit processing rules, which include: when the dimensionless value of the input is greater than zero, output the dimensionless value of the input; when the dimensionless value of the input is less than or equal to zero, output zero. For each discrete sampling time, each circumferential partition, and each radial layer, the upper limit of normal circumferential light leakage for the corresponding circumferential partition is subtracted from the normalized grayscale, and positive over-limit processing is performed on the subtracted value to form the over-limit value of light leakage for the corresponding radial layer. For each discrete sampling time and each circumferential partition, the leakage light exceeding the limit value corresponding to the five radial layers is summed, and the summation result is divided by five to form the circumferential micro-leakage light amount of the corresponding circumferential partition at the corresponding discrete sampling time; For each discrete sampling time, the circumferential micro-leakage corresponding to the seventy-two circumferential partitions is summed, and the summation result is divided by seventy-two to form the whole-circumferential average leakage amount for the corresponding discrete sampling time. For each discrete sampling moment, the circumferential micro-leakage amounts corresponding to the seventy-two circumferential partitions are compared, and the one with the highest value is taken as the local peak leakage amount for the corresponding discrete sampling moment.

[0011] Optionally, the step of generating the operational pressing degree based on the normalized grayscale and comparing it with the normal sealing pressing benchmark to generate pressing deviation, average pressing deviation for the entire cycle, and local peak pressing deviation specifically includes: For each discrete sampling time and each circumferential partition, the running normalized grayscale corresponding to the five radial layers is summed, and the summation result is divided by five to form the running normalized grayscale mean of the corresponding circumferential partition. For each discrete sampling time and each circumferential partition, subtract the normalized grayscale mean of the corresponding circumferential partition from one to form the running compression degree of the corresponding circumferential partition at the corresponding discrete sampling time. For each discrete sampling time and each circumferential partition, the normal sealing and pressing benchmark of the corresponding circumferential partition is subtracted from the running pressing degree of the corresponding circumferential partition to form the pressing deviation of the corresponding circumferential partition at the corresponding discrete sampling time. For each discrete sampling moment, the compression deviations corresponding to the seventy-two circumferential partitions are summed, and the summation result is divided by seventy-two to form the average compression deviation of the entire circle at the corresponding discrete sampling moment; For each discrete sampling time, the compression deviations corresponding to the seventy-two circumferential partitions are compared, and the one with the highest value is taken as the local peak compression deviation for the corresponding discrete sampling time.

[0012] Optionally, the step of generating a door sealing anomaly score based on the mean and peak values ​​of light leakage and compression deviation, and generating a sealing judgment boundary based on the normalized grayscale of the normal baseline, the upper limit of normal circumferential light leakage, and the normal sealing compression baseline, specifically includes: For each discrete sampling moment, the average light leakage, local peak light leakage, average compression deviation, and local peak compression deviation are summed, and the sum is divided by four to form the hatch sealing anomaly score for the corresponding discrete sampling moment. For each normal sealed baseline acquisition, each circumferential partition and each radial layer, the upper limit of normal circumferential light leakage of the corresponding circumferential partition is subtracted from the normal baseline normalized grayscale, and the value after subtraction is subjected to positive over-limit processing to form the baseline light leakage over-limit value of the corresponding radial layer. For each normal closed baseline acquisition and each circumferential zone, the baseline light leakage exceeding the limit corresponding to the five radial layers is summed, and the summation result is divided by five to form the baseline light leakage amount of the corresponding circumferential zone under the normal closed baseline acquisition. For each normal sealed baseline acquisition and each circumferential partition, the normal baseline normalized grayscale values ​​corresponding to the five radial layers are summed, and the summation result is divided by five to form the single baseline normalized grayscale mean value of the corresponding circumferential partition. The single baseline normalized grayscale mean value of the corresponding circumferential partition is subtracted by one, and then the normal sealing and pressing baseline of the corresponding circumferential partition is subtracted to form the baseline pressing deviation of the corresponding circumferential partition under the normal sealed baseline acquisition. For each normal sealed baseline acquisition, the baseline leakage amounts corresponding to the seventy-two circumferential zones are summed and divided by seventy-two to form the baseline average leakage amount. The baseline leakage amounts corresponding to the seventy-two circumferential zones are compared and the highest value is taken to form the baseline peak leakage amount. The baseline compression deviation amounts corresponding to the seventy-two circumferential zones are summed and divided by seventy-two to form the baseline average compression deviation amount. The baseline compression deviation amounts corresponding to the seventy-two circumferential zones are compared and the highest value is taken to form the baseline peak compression deviation amount. For each normal airtight benchmark acquisition, the average light leakage, peak light leakage, average compression deviation, and peak compression deviation of the benchmark are summed, and the sum is divided by four to form the benchmark airtightness anomaly score corresponding to the normal airtight benchmark acquisition. The baseline airtightness abnormality scores corresponding to the six normal airtightness baseline collections are summed, and the summation result is divided by six to form the mean of the normal baseline scores; Calculate the absolute value of the difference between the baseline closure anomaly score and the mean normal baseline score corresponding to the six normal closure baseline collections. Sum the six absolute values ​​of the difference and divide the sum by six to form the average deviation of the normal baseline score. The highest value in the baseline airtightness anomaly score corresponding to the six normal airtightness benchmark acquisitions is added to the average deviation of the normal baseline score to form a candidate judgment value. The candidate judgment value is compared with the mean of the normal baseline score, and the highest value is selected. The highest value after comparison is then pruned with an upper limit of one to form the airtightness judgment boundary.

[0013] Optionally, the step of generating circumferential anomaly positioning data based on circumferential micro-leakage and compression deviation, comparing the hatch sealing anomaly score with the sealing judgment boundary to generate a hatch sealing status identifier, and generating anomaly circumferential partition number, anomaly circumferential partition center angle, circumferential angular distance, and anomaly orientation set in the non-sealed state, and generating a number indicating the absence of anomaly circumferential partition and an empty anomaly orientation set in the sealed state, specifically includes: For each discrete sampling time, the hatch sealing anomaly score is compared with the sealing judgment boundary. When the hatch sealing anomaly score is less than or equal to the sealing judgment boundary, the hatch sealing status flag at the corresponding discrete sampling time is set to one, and the hatch is recorded as being in a sealed state. For each discrete sampling time, the hatch sealing anomaly score is compared with the sealing judgment boundary. When the hatch sealing anomaly score is greater than the sealing judgment boundary, the hatch sealing status flag at the corresponding discrete sampling time is set to zero, and the hatch is recorded as being in an unsealed state. For each discrete sampling time and each circumferential partition, the circumferential micro-leakage amount and the compression deviation amount of the corresponding circumferential partition are summed, and the summation result is divided by two to form the circumferential anomaly location amount of the corresponding circumferential partition at the corresponding discrete sampling time. For each discrete sampling moment, when the door sealing status is marked as one, the abnormal circumferential partition number is set to zero, and it is recorded that there is no abnormal circumferential partition. For each discrete sampling moment, when the door sealing status indicator is zero, the circumferential abnormal positioning quantities corresponding to the seventy-two circumferential partitions are compared, and the circumferential partition number corresponding to the circumferential abnormal positioning quantity with the highest value is taken as the abnormal circumferential partition number. For each discrete sampling moment, when the door sealing status is marked as one, candidate angles are selected within the angle range of zero to two times pi, the circumferential angular distance corresponding to each candidate angle is set to zero, and the abnormal orientation set is set to an empty set. For each discrete sampling moment, when the door sealing status indicator is zero, the sampling angle corresponding to the abnormal circumferential partition number is taken as the center angle of the abnormal circumferential partition. Candidate angles are selected within the angle range of zero to twice pi. The absolute value of the angle difference between the candidate angle and the center angle of the abnormal circumferential partition is taken. The absolute value of the angle difference between the candidate angle and the center angle of the abnormal circumferential partition is then subtracted from the full circle angle corresponding to twice pi. The results of the absolute value of the angle difference between the candidate angle and the center angle of the abnormal circumferential partition are compared with the results of the full circle angle corresponding to twice pi minus the absolute value of the angle difference between the candidate angle and the center angle of the abnormal circumferential partition. The one with the lowest value is taken as the circumferential angular distance from the candidate angle to the center angle of the abnormal circumferential partition. For each discrete sampling moment, when the door sealing status is zero, candidate angles with circumferential angular distances less than the ratio of pi to 72 are selected from the angle range of zero to twice pi, and the selected candidate angles are combined into an abnormal orientation set.

[0014] Optionally, the step of generating anomaly type labels and operation permit labels based on the hatch sealing status indicator, local peak light leakage amount, and local peak compression deviation amount, and outputting operation monitoring records, specifically includes: For each discrete sampling time, when the door sealing status is marked as one, the abnormality type label is set to zero, and the corresponding discrete sampling time is recorded as normal door sealing. For each discrete sampling moment, when the door sealing status is zero and the local peak light leakage is greater than or equal to the local peak compression deviation, the anomaly type label is set to one, and the corresponding discrete sampling moment is recorded as a light leakage-dominated sealing anomaly. For each discrete sampling moment, when the door sealing status is zero and the local peak light leakage is less than the local peak compression deviation, the anomaly type label is set to two, and the corresponding discrete sampling moment is recorded as a compression deviation-dominated sealing anomaly. For each discrete sampling time, the operation permission flag is set to the same value as the hatch sealing status flag. The state where the hatch sealing status flag is one is used as the operation monitoring permission condition. When the operation permission flag is one, the record meets the operation monitoring permission condition. When the operation permission flag is zero, the record does not meet the operation monitoring permission condition. For each discrete sampling moment, the discrete sampling moment number, the hatch sealing anomaly score, the sealing judgment boundary, the hatch sealing status identifier, the anomaly circumferential partition number, the anomaly location set, and the anomaly type label are written into the operation monitoring record.

[0015] The present invention has the following beneficial effects: 1. First, select the door sealing ring area and establish visual sampling coordinates around the center point of the sealing ring, the direction from the door hinge to the latch, and the vertical direction within the image plane. Then, divide the sealing ring into circumferential partitions and radial layers, generating circumferential sampling angles and sealing ring sampling points. This process transforms the originally continuous, ring-shaped sealing area, which is easily affected by the shooting angle, into a sampling object with clear directional meaning and hierarchical structure. Each sampling point corresponds to the specific orientation and radial position of the sealing ring. When anomalies are subsequently detected, they can be directly traced to a local area of ​​the door sealing ring, rather than just giving a general anomaly result. Compared with common technologies such as door lock contact detection, single-point photoelectric detection, or whole-image brightness judgment, this solution establishes a monitoring foundation with spatial positioning capabilities. It can solve the problem of difficulty in timely positioning of local kitchen waste residue, local deformation, and insufficient local compression in the sealing ring, providing a stable data structure for subsequent light leakage, compression, and anomaly orientation analysis.

[0016] 2. Simultaneously, dark-field images, transmitted light reference images, normal sealing baseline images, and images of the door sealing ring in operation are introduced. Grayscale values ​​are read at the sealing ring sampling point level. Then, the grayscale values ​​of the normal sealing baseline sampling and the operation sampling are normalized based on the dark-field grayscale and transmitted light reference grayscale to form normalized grayscale values ​​and operation normalized grayscale values. Non-sealing factors such as camera sensitivity differences, ambient light interference, changes in the intensity of the detection light source, and differences in the reflectivity of the sealing ring material are removed from the monitoring data, making the data comparable between different sampling times and different sampling points. Food waste equipment is usually in a humid, oily, residue-laden, and complex lighting environment. If the original grayscale is used directly for judgment, false alarms or missed alarms are likely to occur. This solution establishes the grayscale scale of the sampling point itself through dark-field and transmitted light references. Compared with the traditional single image thresholding method, it is more adaptable to visual differences caused by fluctuations in on-site lighting and equipment aging, improving the stability and repeatability of sealing status judgment.

[0017] 3. Circumferential processing is performed using normalized grayscale as a baseline to generate the upper limit of normal light leakage and the normal sealing and pressing baseline for each circumferential zone. This scheme does not simply set a fixed threshold, but rather forms a circumferential baseline that matches the current sealing structure, camera arrangement, detection light source, and sealing ring condition of the equipment based on the baseline data of the equipment itself under normal sealing conditions. Even if different food waste disposal equipment has the same structure, it may exhibit different normal brightness distributions due to installation errors, batches of sealing ring materials, camera angles, and light source positions. By establishing a normal light leakage upper limit and pressing baseline for each circumferential zone separately, long-term misjudgments in certain directions caused by using a uniform threshold can be avoided. Compared with the empirical thresholds or manually calibrated thresholds commonly used in existing technologies, this scheme makes the judgment baseline adaptable to individual equipment, and can solve the problem of difficulty in accurate monitoring when the sealing ring itself has uneven normal brightness, local reflection differences, or initial assembly deviations.

[0018] 4. The normalized grayscale value is compared with the upper limit of normal circumferential light leakage in layers to generate circumferential micro-leakage, average light leakage across the entire ring, and local peak light leakage. This scheme transforms "light leakage" from a purely visual phenomenon into a quantifiable, comparable, and localizable monitoring indicator. Circumferential micro-leakage reflects the degree of local light leakage in a specific location, average light leakage across the entire ring reflects the overall sealing level, and local peak light leakage reflects the most severe local anomaly. It can identify both general light leakage caused by insufficient overall sealing ring compression and localized micro-leakage caused by residue inclusions or localized warping in a small section of the sealing ring. Traditional image detection typically calculates the brightness of the entire image or uses a single area threshold, which is easily affected by strong local reflections, stains, or background brightness. This scheme, through refined comparison of circumferential partitioning and radial layering, makes light leakage judgment directional, localized, and anti-interference capable, enabling earlier detection of sealing anomalies where weak light leakage has already occurred before obvious gaps have formed.

[0019] 5. This approach not only focuses on the intensity of light leakage but also generates the operational pressing degree based on normalized grayscale values. It compares this operational pressing degree with a normal sealing pressing benchmark to generate pressing deviation, average pressing deviation across the entire ring, and local peak pressing deviation. This scheme characterizes the door sealing state from another dimension, namely whether the sealing ring is uniformly and fully pressed. In some cases, although the sealing ring may not yet show obvious light leakage, there may be local insufficient pressing, uneven stress, or sealing ring fatigue deformation, which may be difficult to detect in time with simple light leakage detection. This scheme, through pressing consistency analysis, can identify sealing failure trends in advance, allowing the equipment to issue monitoring results before the anomaly escalates. Compared to existing technologies that only detect whether the door is closed or whether there is obvious light transmission, this scheme introduces a quantitative evaluation of sealing pressing quality, which can solve problems such as door locks being engaged but the sealing ring not fitting properly, uneven pressing force, and difficulty in identifying seal life degradation.

[0020] 6. The average and peak values ​​of light leakage and compression deviation are combined to generate a door sealing anomaly score. This score is then combined with the normalized grayscale of the normal baseline, the upper limit of normal circumferential light leakage, and the normal sealing compression baseline to generate a sealing judgment boundary. This scheme does not use a single indicator as the basis for judging whether something is abnormal, but considers both the overall degree of anomaly and the local abnormal peak value. This ensures that the score reflects both global sealing degradation and sensitively captures severe local anomalies. It avoids simply using average values ​​to mask local anomalies and also avoids misjudgments triggered by random noise from a single peak value. The sealing judgment boundary is derived from normal baseline data and fits the normal fluctuation range of the equipment itself. Compared with existing technologies that use fixed thresholds, manual experience thresholds, or single-indicator judgments, this scheme achieves a comprehensive judgment that better reflects the actual changes in the sealing state. It can solve the problem of threshold inaccuracies caused by pollution, light drift, and aging of sealing rings after long-term operation of food waste disposal equipment.

[0021] 7. Based on the circumferential micro-leakage and compression deviation, a circumferential anomaly location quantity is generated. After comparing the door sealing anomaly score with the sealing judgment boundary, a door sealing status indicator is generated. Simultaneously, different forms of anomaly location results are output according to the sealing or non-sealing state. In the non-sealing state, the abnormal circumferential partition number, abnormal circumferential partition center angle, circumferential angular distance, and anomaly location set are further generated. In the sealing state, the number indicating the absence of abnormal circumferential partitions and the empty anomaly location set are output. This not only determines whether the door is sealed but also indicates which circumferential area the anomaly is concentrated in, facilitating maintenance personnel to quickly check whether there is food residue trapped, damaged sealing rings, door misalignment, or abnormal latch compression at the corresponding location. Compared to existing technologies that only provide alarm or shutdown signals, this solution provides directional diagnostic information, reducing troubleshooting time and providing a basis for automatic prompting of cleaning locations, maintenance locations, or subsequent intelligent control of the equipment.

[0022] 8. Based on the door sealing status indicator, local peak light leakage, and local peak compression deviation, anomaly type labels and operation permit labels are generated, and operation monitoring records are output. This solution summarizes the results of the aforementioned visual sampling, normalization, benchmark construction, light leakage analysis, compression analysis, scoring judgment, and orientation positioning into data records that can be used for equipment operation control and maintenance traceability. The equipment can not only know whether operation is currently permitted, but also distinguish whether the anomaly mainly originates from light leakage or compression deviation, and record the discrete sampling time, score, boundary, status, orientation, and type, providing a basis for subsequent fault review, maintenance planning, and seal life assessment. Compared with existing technologies that rely solely on switch alarms or manual recording, this solution makes the operation monitoring results traceable, categorizable, and usable for operation decision-making, solving the problems of lack of cause classification, lack of orientation records, and lack of continuous monitoring basis after anomalies occur in food waste equipment. Attached Figure Description

[0023] Figure 1 This is a schematic diagram of the process of the present invention.

[0024] Figure 2 This is a schematic diagram of the visual sampling coordinates and sampling points in the closed loop inside the cabin according to the present invention.

[0025] Figure 3 This is a schematic diagram of the closed-loop visual data acquisition and grayscale normalization process of the present invention.

[0026] Figure 4 This is a schematic diagram of the process for generating the sealing anomaly scoring and judgment boundary of the present invention. Detailed Implementation

[0027] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0028] Example, refer to Figure 1 A machine vision-based method for monitoring the operation of kitchen waste disposal equipment includes: Select the door sealing ring area, establish visual sampling coordinates, divide the circumferential partitions and radial layers, and generate circumferential sampling angles and sealing ring sampling points; Acquire dark field images, light-transmitted reference images, normal sealed reference images, and cabin door sealing ring operation images; read the grayscale of the sealing ring sampling points; and generate normal reference normalized grayscale and operation normalized grayscale through normalization processing. The normalized grayscale of the normal reference is processed circumferentially to generate the upper limit of normal light leakage and the normal sealing and pressing reference. The normalized grayscale is compared with the upper limit of normal circumferential light leakage in layers to generate circumferential micro-leakage, average light leakage over the whole circle and local peak light leakage. The running pressure degree is generated based on the normalized grayscale and compared with the normal sealing pressure benchmark to generate the pressure deviation, the average pressure deviation of the whole circle and the local peak pressure deviation. The door sealing anomaly score is generated based on the mean and peak values ​​of light leakage and compression deviation, and the sealing judgment boundary is generated based on the normalized grayscale of the normal baseline, the upper limit of normal light leakage in the circumferential direction, and the normal sealing compression baseline. Based on the circumferential micro-leakage and compression deviation, a circumferential anomaly positioning quantity is generated. The hatch sealing anomaly score is compared with the sealing judgment boundary to generate a hatch sealing status identifier. In the non-sealed state, anomaly circumferential partition number, anomaly circumferential partition center angle, circumferential angular distance and anomaly orientation set are generated. In the sealed state, a number indicating that there is no anomaly circumferential partition and an empty anomaly orientation set are generated. Anomaly type labels and operation permit labels are generated based on the hatch sealing status indicator, local peak light leakage amount, and local peak compression deviation amount, and operation monitoring records are output.

[0029] A visual sampling system was established around the hatch sealing ring, and image acquisition, grayscale normalization, normal baseline construction, light leakage analysis, compression analysis, airtightness scoring, anomaly localization, and result output were completed sequentially. This scheme transforms the originally continuous sealing ring area into a segmentable, layerable, and traceable visual inspection object by selecting the hatch sealing ring region, establishing visual sampling coordinates, and dividing it into circumferential partitions and radial layers. This solves the problem that traditional door lock switch detection can only determine whether the hatch is closed and cannot reflect the local state of the sealing ring. Furthermore, by acquiring and normalizing dark-field images, transmitted light reference images, normal airtightness baseline images, and operating images, the impact of ambient light, camera sensitivity differences, and fluctuations in the detection light source on the monitoring results is reduced. By forming the upper limit of circumferential normal light leakage, the normal sealing compression baseline, the circumferential micro-light leakage amount, the compression deviation, and the airtightness judgment boundary, both light leakage anomalies and compression anomalies can be evaluated simultaneously, avoiding misjudgments caused by relying solely on a single brightness threshold. By outputting the abnormal circumferential partition number, abnormal location set, abnormal type label, and operation permit identifier, the monitoring results can be used for equipment operation control and maintenance prompts, which has the beneficial effects of improving the accuracy of abnormal identification, enhancing the location positioning capability, and facilitating subsequent maintenance and troubleshooting.

[0030] Reference Figure 2 The process of selecting the door sealing ring area, establishing visual sampling coordinates, dividing the area into circumferential zones and radial layers, and generating circumferential sampling angles and sealing ring sampling points specifically includes: The door sealing ring area will be used as the monitoring target for the operation of the food waste disposal equipment. Set the center point of the sealing ring as the origin of the coordinate system, set the direction of the door hinge pointing to the latch as the positive direction of the horizontal coordinate, and set the direction of the image plane after rotating 90 degrees counterclockwise from the positive direction of the horizontal coordinate as the positive direction of the vertical coordinate. The sealing ring area is divided into seventy-two circumferential partitions and five radial layers. A set of indexes for circumferential partitions, a set of indexes for radial layers, and a set of indexes for normal sealing reference acquisition are established. The set of indexes for circumferential partitions contains consecutive integers from one to seventy-two, the set of indexes for radial layers contains consecutive integers from one to five, and the set of indexes for normal sealing reference acquisition contains consecutive integers from one to six. Each integer in the circumferential partition index set is used as the circumferential partition number, each integer in the radial hierarchical index set is used as the radial hierarchical number, and each integer in the normal closed benchmark acquisition index set is used as the normal closed benchmark acquisition number. Measure the inner boundary radius and outer boundary radius of the sealing ring, divide the radial width between the outer boundary radius and the inner boundary radius of the sealing ring into five equal-width radial intervals, and use the middle radius of each radial interval as the sampling radius of the corresponding radial layer; The full circle angle corresponding to twice pi is divided into seventy-two circumferential zones with equal angles, and the sampling angles corresponding to each circumferential zone are formed according to the order of the circumferential zone numbers; For each radial layer and each circumferential partition, corresponding sealing ring sampling points are generated according to the corresponding sampling radius, sampling angle, direction cosine of the sampling angle in the positive direction of the horizontal coordinate, and direction sine of the sampling angle in the positive direction of the vertical coordinate.

[0031] The door sealing ring area will be used as the monitoring target for the operation of the food waste disposal equipment. With the center point of the sealing ring Using the origin of the coordinate system and the direction from the hatch hinge to the latch as the reference point... The positive direction of the axis, with respect to the image plane by Rotate the axis counterclockwise direction as Positive direction of the axis; The sealing ring area is divided into One ring partition and A radial hierarchy is established, and the index set is as follows: , , ;in, This represents the total number of circumferential partitions. This represents the total number of radial strata. A set of circumferential partition indexes; A radial hierarchical index set; This is a set of indexes for normal sealed baseline acquisition; For circumferential partition numbering; Radial layer numbering; The data collection number is based on the normal sealed baseline. The radius of the inner boundary of the sealing ring is denoted as . ; The outer boundary radius of the sealing ring is denoted as . ; The radial sampling radius is calculated as follows: ;in, For the first The sampling radius of each radial layer; The circumferential sampling angle is calculated as follows: ;in, For the first The sampling angle of each circumferential partition; The sampling points for the sealing ring are calculated as follows: ;in, For the first The radial layer, the first Each circumferential partition corresponds to a sealing ring sampling point; Sampling angle exist Direction cosine along the axial direction; Sampling angle exist The directional sine along the axial direction.

[0032] Reference Figure 3 The acquisition of dark-field images, transmitted light reference images, normal sealing reference images, and cabin door sealing ring operation images, reading the grayscale of the sealing ring sampling points, and generating normal reference normalized grayscale and operation normalized grayscale through normalization processing specifically includes: Set a fixed zero-prevention constant, the value of which is 10 to the power of negative 6; With the circumferential detection light source turned off, dark field images are acquired when the hatch is closed and the equipment is not feeding materials. The dark field images are then normalized to form a dark field normalized grayscale image. The grayscale value of each sealing ring sampling point in the dark field normalized grayscale image is read, and the upper limit of the read grayscale value is clipped. The upper limit of the clipping is the value after subtracting a fixed anti-zero constant from one, thus forming the dark field grayscale of the corresponding sealing ring sampling point. Turn on the circumferential detection light source and acquire a light transmission reference image when the hatch is open to the point where the sealing ring is unobstructed. Perform grayscale normalization processing on the light transmission reference image to form a light transmission reference normalized grayscale image. Read the grayscale value of each sealing ring sampling point in the light transmission reference normalized grayscale image and perform lower limit cropping on the read grayscale value. The lower limit of cropping is the value of the dark field grayscale of the corresponding sealing ring sampling point plus a fixed anti-zero constant, to form the light transmission reference grayscale of the corresponding sealing ring sampling point. With the hatch clean, the sealing ring free of debris, and the door lock fully engaged, six normal airtightness benchmark acquisitions are performed consecutively. Each acquisition forms a normal airtightness benchmark image. The gray values ​​of the sealing ring sampling points corresponding to each radial layer and each circumferential partition in each normal airtightness benchmark image are read to form the normal airtightness benchmark sampling gray values. After the food waste equipment enters the operation monitoring state, the operation images of the door sealing ring are collected according to the discrete sampling time. The gray values ​​of the corresponding sealing ring sampling points of each radial layer and each circumferential partition in each door sealing ring operation image are read to form the operation sampling gray value. For each discrete sampling moment, each radial layer and each circumferential partition, the dark field gray level of the corresponding sealing ring sampling point is subtracted from the running sampling gray level to form the running gray level difference. The dark field gray level is subtracted from the light-transmitting reference gray level of the corresponding sealing ring sampling point to form the reference gray level difference. The result of dividing the running gray level difference by the reference gray level difference is clipped from zero to one to form the running normalized gray level. For each normal sealed baseline acquisition, each radial layer, and each circumferential partition, the dark field gray level of the corresponding sealing ring sampling point is subtracted from the gray level of the normal sealed baseline sampling point to form a baseline gray level difference. The dark field gray level is subtracted from the light-transmitting reference gray level of the corresponding sealing ring sampling point to form a reference gray level difference. The result of dividing the baseline gray level difference by the reference gray level difference is clipped from zero to one to form the normal baseline normalized gray level.

[0033] Set the fixed zero-prevention constant as ; With the circumferential detection light source turned off, dark field images were acquired with the hatch closed and the equipment not feeding materials. Read sampling points The grayscale value is used to obtain the dark field grayscale value: ;in, For the first The radial layer, the first Dark field grayscale of each circumferential partition sampling point; For dark field normalized grayscale images at sampling points The grayscale value at that location; Turn on the circumferential detection light source and collect a transmitted reference image when the hatch is opened to the point where the sealing ring is unobstructed; The grayscale value of the sampling point is read to obtain the transmittance reference grayscale: ;in, For the first The radial layer, the first Transmittance reference grayscale of each circumferential partition sampling point; For the light-transmitting reference, the normalized grayscale image is at the sampling point. The grayscale value at that location; With the hatch clean, the sealing ring free of debris, and the door lock fully engaged, perform six consecutive baseline data acquisitions. The sampled grayscale value of the secondary benchmark acquisition is recorded as: ;in, For the first The first normal closed baseline data collection, the first The radial layer, the first The baseline gray level of each circumferential partition sampling point; For the first Subnormal closed baseline image at sampling point The grayscale value at that location; After the food waste disposal equipment enters the operation monitoring state, it is numbered according to the sampling time. Acquire images of the hatch sealing ring, and then... The sampled gray level at each sampling time is recorded as: ;in, Number the discrete sampling time points; For the first The discrete sampling time, the first The radial layer, the first The grayscale of each circumferential partition sampling point is sampled. For the first The running image at each discrete sampling time point The grayscale value at that location; The normalized grayscale value is calculated as follows: ;in, For the first The discrete sampling time, the first The radial layer, the first Normalized grayscale of each circumferential partition; The normalized grayscale value is calculated as follows: ;in, For the first The first normal closed baseline data collection, the first The radial layer, the first Normalized grayscale of each circumferential partition.

[0034] The process of performing circumferential processing on the normalized grayscale of the normal reference to generate the upper limit of normal light leakage and the normal sealing and pressing reference specifically includes: For each circumferential partition and each normal closed baseline acquisition, the normal baseline normalized gray values ​​corresponding to the five radial layers are summed, and the summation result is divided by five to form the circumferential normalized gray value mean of the corresponding circumferential partition under the normal closed baseline acquisition. For each circumferential zone, the circumferential normalized grayscale mean values ​​corresponding to the six normal sealed baseline acquisitions are compared, and the highest value is taken as the upper limit of normal circumferential light leakage of the corresponding circumferential zone under normal sealed conditions. For each circumferential zone, the normalized gray values ​​of the six normal sealing reference acquisitions and the five radial layer corresponding normal references are summed. The summation result is divided by 30 to form the mean normalized gray value of the circumferential comprehensive reference for the corresponding circumferential zone. The mean normalized gray value of the circumferential comprehensive reference for the corresponding circumferential zone is then subtracted by 1 to form the normal sealing and pressing reference for the corresponding circumferential zone. The upper limit of normal light leakage and the normal sealing and pressing benchmark of each circumferential zone are fixed as the calculation benchmark for this operation monitoring of the kitchen waste equipment.

[0035] For each circumferential partition Calculate the upper limit of the circumferential normalized grayscale under normal sealed conditions, specifically as follows: ;in, For the first The upper limit of normal circumferential light leakage in a circumferential zone under normal sealed conditions; For each circumferential partition The calculation of the normal sealing pressure reference is as follows: ;in, For the first Normal sealing and pressing reference for each circumferential zone; Will and This will be used as the calculation benchmark for the current equipment operation monitoring.

[0036] The process of comparing the normalized grayscale with the upper limit of normal circumferential light leakage in layers to generate circumferential micro-leakage, average light leakage over the entire circle, and local peak light leakage specifically includes: Construct positive out-of-limit processing rules, which include: when the dimensionless value of the input is greater than zero, output the dimensionless value of the input; when the dimensionless value of the input is less than or equal to zero, output zero. For each discrete sampling time, each circumferential partition, and each radial layer, the upper limit of normal circumferential light leakage for the corresponding circumferential partition is subtracted from the normalized grayscale, and positive over-limit processing is performed on the subtracted value to form the over-limit value of light leakage for the corresponding radial layer. For each discrete sampling time and each circumferential partition, the leakage light exceeding the limit value corresponding to the five radial layers is summed, and the summation result is divided by five to form the circumferential micro-leakage light amount of the corresponding circumferential partition at the corresponding discrete sampling time; For each discrete sampling time, the circumferential micro-leakage corresponding to the seventy-two circumferential partitions is summed, and the summation result is divided by seventy-two to form the whole-circumferential average leakage amount for the corresponding discrete sampling time. For each discrete sampling moment, the circumferential micro-leakage amounts corresponding to the seventy-two circumferential partitions are compared, and the one with the highest value is taken as the local peak leakage amount for the corresponding discrete sampling moment.

[0037] For any dimensionless numerical value The positive transfinite function is constructed as follows: ; For the The sampling time, the first The circumferential partitions are used to calculate the circumferential micro-leakage: ;in, For the first The discrete sampling time, the first The amount of light leakage in each circumferential zone; For the At each sampling time point, the average light leakage over the entire cycle is calculated as follows: ;in, For the first The average light leakage over the entire cycle at each discrete sampling time; For the At each sampling time, the local peak light leakage is calculated as follows: ;in, For the first The local peak light leakage at each discrete sampling time.

[0038] The process of generating the operational pressing degree based on the normalized grayscale and comparing it with the normal sealing pressing benchmark to generate pressing deviation, average pressing deviation for the entire cycle, and local peak pressing deviation specifically includes: For each discrete sampling time and each circumferential partition, the running normalized grayscale corresponding to the five radial layers is summed, and the summation result is divided by five to form the running normalized grayscale mean of the corresponding circumferential partition. For each discrete sampling time and each circumferential partition, subtract the normalized grayscale mean of the corresponding circumferential partition from one to form the running compression degree of the corresponding circumferential partition at the corresponding discrete sampling time. For each discrete sampling time and each circumferential partition, the normal sealing and pressing benchmark of the corresponding circumferential partition is subtracted from the running pressing degree of the corresponding circumferential partition to form the pressing deviation of the corresponding circumferential partition at the corresponding discrete sampling time. For each discrete sampling moment, the compression deviations corresponding to the seventy-two circumferential partitions are summed, and the summation result is divided by seventy-two to form the average compression deviation of the entire circle at the corresponding discrete sampling moment; For each discrete sampling time, the compression deviations corresponding to the seventy-two circumferential partitions are compared, and the one with the highest value is taken as the local peak compression deviation for the corresponding discrete sampling time.

[0039] For the The sampling time, the first The degree of compression during operation is calculated for each of the three circumferential partitions: ;in, For the first The discrete sampling time, the first The degree of compression in the operation of each circumferential partition; For the The sampling time, the first For each circumferential partition, the compression deviation is calculated as follows: ;in, For the first The discrete sampling time, the first The compression deviation of each circumferential zone; For the At each sampling time point, the average compression deviation for the entire cycle is calculated as follows: ;in, For the first The average compression deviation of the entire circle at each discrete sampling time; For the At each sampling time, the local peak compression deviation is calculated as follows: ;in, For the first The local peak compression deviation at each discrete sampling time.

[0040] Reference Figure 4 The step of generating a door sealing anomaly score based on the mean and peak values ​​of light leakage and compression deviation, and generating a sealing judgment boundary based on the normalized grayscale of the normal baseline, the upper limit of normal circumferential light leakage, and the normal sealing compression baseline, specifically includes: For each discrete sampling moment, the average light leakage, local peak light leakage, average compression deviation, and local peak compression deviation are summed, and the sum is divided by four to form the hatch sealing anomaly score for the corresponding discrete sampling moment. For each normal sealed baseline acquisition, each circumferential partition and each radial layer, the upper limit of normal circumferential light leakage of the corresponding circumferential partition is subtracted from the normal baseline normalized grayscale, and the value after subtraction is subjected to positive over-limit processing to form the baseline light leakage over-limit value of the corresponding radial layer. For each normal closed baseline acquisition and each circumferential zone, the baseline light leakage exceeding the limit corresponding to the five radial layers is summed, and the summation result is divided by five to form the baseline light leakage amount of the corresponding circumferential zone under the normal closed baseline acquisition. For each normal sealed baseline acquisition and each circumferential partition, the normal baseline normalized grayscale values ​​corresponding to the five radial layers are summed, and the summation result is divided by five to form the single baseline normalized grayscale mean value of the corresponding circumferential partition. The single baseline normalized grayscale mean value of the corresponding circumferential partition is subtracted by one, and then the normal sealing and pressing baseline of the corresponding circumferential partition is subtracted to form the baseline pressing deviation of the corresponding circumferential partition under the normal sealed baseline acquisition. For each normal sealed baseline acquisition, the baseline leakage amounts corresponding to the seventy-two circumferential zones are summed and divided by seventy-two to form the baseline average leakage amount. The baseline leakage amounts corresponding to the seventy-two circumferential zones are compared and the highest value is taken to form the baseline peak leakage amount. The baseline compression deviation amounts corresponding to the seventy-two circumferential zones are summed and divided by seventy-two to form the baseline average compression deviation amount. The baseline compression deviation amounts corresponding to the seventy-two circumferential zones are compared and the highest value is taken to form the baseline peak compression deviation amount. For each normal airtight benchmark acquisition, the average light leakage, peak light leakage, average compression deviation, and peak compression deviation of the benchmark are summed, and the sum is divided by four to form the benchmark airtightness anomaly score corresponding to the normal airtight benchmark acquisition. The baseline airtightness abnormality scores corresponding to the six normal airtightness baseline collections are summed, and the summation result is divided by six to form the mean of the normal baseline scores; Calculate the absolute value of the difference between the baseline closure anomaly score and the mean normal baseline score corresponding to the six normal closure baseline collections. Sum the six absolute values ​​of the difference and divide the sum by six to form the average deviation of the normal baseline score. The highest value in the baseline airtightness anomaly score corresponding to the six normal airtightness benchmark acquisitions is added to the average deviation of the normal baseline score to form a candidate judgment value. The candidate judgment value is compared with the mean of the normal baseline score, and the highest value is selected. The highest value after comparison is then pruned with an upper limit of one to form the airtightness judgment boundary.

[0041] For the At each sampling time, the hatch sealing anomaly score is constructed as follows: ;in, For the first Door sealing anomaly score at each discrete sampling time; For the The first normal closed baseline data collection, the first The baseline light leakage is calculated for each of the three annular partitions: ;in, For the first The first normal closed baseline data collection, the first The baseline light leakage of each circumferential zone; For the The first normal closed baseline data collection, the first For each circumferential partition, the baseline compression deviation is calculated as follows: ;in, For the first The first normal closed baseline data collection, the first The reference compression deviation of each circumferential zone; For the Based on the sub-normal airtightness baseline data collection, an airtightness anomaly score was constructed as follows: ;in, For the first Benchmark sealing anomaly score collected from subnormal sealing benchmark; Calculate the mean of the normal benchmark score as ; Calculate the mean deviation of the normal baseline score as follows: ; Construct the closure determination boundary as .

[0042] The process involves generating a circumferential anomaly location based on circumferential micro-leakage and compression deviation, comparing the hatch sealing anomaly score with the sealing judgment boundary to generate a hatch sealing status identifier, and generating anomaly circumferential partition numbers, anomaly circumferential partition center angles, circumferential angular distances, and anomaly orientation sets in the non-sealed state, and generating numbers indicating the absence of anomaly circumferential partitions and empty anomaly orientation sets in the sealed state. Specifically, this includes: For each discrete sampling time, the hatch sealing anomaly score is compared with the sealing judgment boundary. When the hatch sealing anomaly score is less than or equal to the sealing judgment boundary, the hatch sealing status flag at the corresponding discrete sampling time is set to one, and the hatch is recorded as being in a sealed state. For each discrete sampling time, the hatch sealing anomaly score is compared with the sealing judgment boundary. When the hatch sealing anomaly score is greater than the sealing judgment boundary, the hatch sealing status flag at the corresponding discrete sampling time is set to zero, and the hatch is recorded as being in an unsealed state. For each discrete sampling time and each circumferential partition, the circumferential micro-leakage amount and the compression deviation amount of the corresponding circumferential partition are summed, and the summation result is divided by two to form the circumferential anomaly location amount of the corresponding circumferential partition at the corresponding discrete sampling time. For each discrete sampling moment, when the door sealing status is marked as one, the abnormal circumferential partition number is set to zero, and it is recorded that there is no abnormal circumferential partition. For each discrete sampling moment, when the door sealing status indicator is zero, the circumferential abnormal positioning quantities corresponding to the seventy-two circumferential partitions are compared, and the circumferential partition number corresponding to the circumferential abnormal positioning quantity with the highest value is taken as the abnormal circumferential partition number. For each discrete sampling moment, when the door sealing status is marked as one, candidate angles are selected within the angle range of zero to two times pi, the circumferential angular distance corresponding to each candidate angle is set to zero, and the abnormal orientation set is set to an empty set. For each discrete sampling moment, when the door sealing status indicator is zero, the sampling angle corresponding to the abnormal circumferential partition number is taken as the center angle of the abnormal circumferential partition. Candidate angles are selected within the angle range of zero to twice pi. The absolute value of the angle difference between the candidate angle and the center angle of the abnormal circumferential partition is taken. The absolute value of the angle difference between the candidate angle and the center angle of the abnormal circumferential partition is then subtracted from the full circle angle corresponding to twice pi. The results of the absolute value of the angle difference between the candidate angle and the center angle of the abnormal circumferential partition are compared with the results of the full circle angle corresponding to twice pi minus the absolute value of the angle difference between the candidate angle and the center angle of the abnormal circumferential partition. The one with the lowest value is taken as the circumferential angular distance from the candidate angle to the center angle of the abnormal circumferential partition. For each discrete sampling moment, when the door sealing status is zero, candidate angles with circumferential angular distances less than the ratio of pi to 72 are selected from the angle range of zero to twice pi, and the selected candidate angles are combined into an abnormal orientation set.

[0043] For the At each sampling time, the generated hatch sealing status identifier is: ;in, For the first The door sealing status indicator at each discrete sampling time. Indicates the first At each discrete sampling moment, the hatch is in a sealed state. Indicates the first At each discrete sampling moment, the hatch is in a non-sealed state; For the The sampling time, the first The number of circumferential partitions is used to construct the circumferential anomaly localization quantity: ;in, For the first The discrete sampling time, the first The circumferential anomaly location quantity of each circumferential partition; For the At each sampling time, the abnormal circumferential partition number is determined as follows: ;in, For the first The abnormal cyclic partition number at each discrete sampling time. Indicates the first There are no abnormal circumferential partitions at any of the discrete sampling times; For circumferential partition comparison index; For the first The discrete sampling time, the first The circumferential anomaly location quantity of each circumferential partition; For the At each sampling time, the circumferential angular distance is calculated as follows: ;in, For the first At each discrete sampling time, the angle variable Circumferential angular distance to the center corner of the abnormal circumferential partition; The input variable for the function represents the candidate angles in the set of abnormal orientations; For the first The sampling angle of each circumferential partition; The set of abnormal locations is generated as follows: ;in, For the first A set of abnormal locations at discrete sampling times; It is an empty set.

[0044] The system generates anomaly type labels and operation permit labels based on the hatch sealing status indicator, local peak light leakage, and local peak compression deviation, and outputs operation monitoring records, specifically including: For each discrete sampling time, when the door sealing status is marked as one, the abnormality type label is set to zero, and the corresponding discrete sampling time is recorded as normal door sealing. For each discrete sampling moment, when the door sealing status is zero and the local peak light leakage is greater than or equal to the local peak compression deviation, the anomaly type label is set to one, and the corresponding discrete sampling moment is recorded as a light leakage-dominated sealing anomaly. For each discrete sampling moment, when the door sealing status is zero and the local peak light leakage is less than the local peak compression deviation, the anomaly type label is set to two, and the corresponding discrete sampling moment is recorded as a compression deviation-dominated sealing anomaly. For each discrete sampling time, the operation permission flag is set to the same value as the hatch sealing status flag. The state where the hatch sealing status flag is one is used as the operation monitoring permission condition. When the operation permission flag is one, the record meets the operation monitoring permission condition. When the operation permission flag is zero, the record does not meet the operation monitoring permission condition. For each discrete sampling moment, the discrete sampling moment number, the hatch sealing anomaly score, the sealing judgment boundary, the hatch sealing status identifier, the anomaly circumferential partition number, the anomaly location set, and the anomaly type label are written into the operation monitoring record.

[0045] For the At each sampling time, the generated anomaly type label is: ;in, For the first Anomaly type label for each discrete sampling time. Indicates the first At each discrete sampling time, the hatch was properly sealed. Indicates the first Each discrete sampling moment is a light leakage-dominated type of sealed anomaly. Indicates the first Each discrete sampling moment represents a compression deviation-dominant type of sealing anomaly. For the At each sampling time, the generated runtime license identifier is: ;in, For the first The runtime permission identifier for each discrete sampling time. Indicates the first Each discrete sampling moment meets the operational monitoring permit conditions. Indicates the first The discrete sampling time does not meet the operational monitoring permit conditions; The first The operation monitoring results at each sampling time point are recorded as follows: ;in, For the first Operation monitoring records at discrete sampling times.

[0046] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0047] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for monitoring the operation of kitchen waste disposal equipment based on machine vision, characterized in that, include: Select the door sealing ring area, establish visual sampling coordinates, divide the circumferential partitions and radial layers, and generate circumferential sampling angles and sealing ring sampling points; Acquire dark field images, light-transmitted reference images, normal sealed reference images, and cabin door sealing ring operation images; read the grayscale of the sealing ring sampling points; and generate normal reference normalized grayscale and operation normalized grayscale through normalization processing. The normalized grayscale of the normal baseline is processed circumferentially to generate the upper limit of normal light leakage and the normal sealing and pressing baseline, specifically including: For each circumferential partition and each normal closed baseline acquisition, the normal baseline normalized gray values ​​corresponding to the five radial layers are summed, and the summation result is divided by five to form the circumferential normalized gray value mean of the corresponding circumferential partition under the normal closed baseline acquisition. For each circumferential zone, the circumferential normalized grayscale mean values ​​corresponding to the six normal sealed baseline acquisitions are compared, and the highest value is taken as the upper limit of normal circumferential light leakage of the corresponding circumferential zone under normal sealed conditions. For each circumferential zone, the normalized gray values ​​of the six normal sealing reference acquisitions and the five radial layer corresponding normal references are summed. The summation result is divided by 30 to form the mean normalized gray value of the circumferential comprehensive reference for the corresponding circumferential zone. The mean normalized gray value of the circumferential comprehensive reference for the corresponding circumferential zone is then subtracted by 1 to form the normal sealing and pressing reference for the corresponding circumferential zone. The upper limit of normal light leakage and the normal sealing and pressing benchmark of each circumferential zone are fixed as the calculation benchmark for this operation monitoring of the kitchen waste equipment; The normalized grayscale value is compared with the upper limit of normal circumferential light leakage in layers to generate circumferential micro-leakage, average light leakage over the entire circle, and local peak light leakage, specifically including: Construct positive out-of-limit processing rules, which include: when the dimensionless value of the input is greater than zero, output the dimensionless value of the input; when the dimensionless value of the input is less than or equal to zero, output zero. For each discrete sampling time, each circumferential partition, and each radial layer, the upper limit of normal circumferential light leakage for the corresponding circumferential partition is subtracted from the normalized grayscale, and positive over-limit processing is performed on the subtracted value to form the over-limit value of light leakage for the corresponding radial layer. For each discrete sampling time and each circumferential partition, the leakage light exceeding the limit value corresponding to the five radial layers is summed, and the summation result is divided by five to form the circumferential micro-leakage light amount of the corresponding circumferential partition at the corresponding discrete sampling time; For each discrete sampling time, the circumferential micro-leakage corresponding to the seventy-two circumferential partitions is summed, and the summation result is divided by seventy-two to form the whole-circumferential average leakage amount for the corresponding discrete sampling time. For each discrete sampling time, the circumferential micro-leakage amount corresponding to the seventy-two circumferential partitions is compared, and the one with the highest value is taken as the local peak leakage amount for the corresponding discrete sampling time. The operational pressing degree is generated based on the normalized grayscale values ​​and compared with the normal sealing pressing benchmark to generate pressing deviation, average pressing deviation over the entire cycle, and local peak pressing deviation, specifically including: For each discrete sampling time and each circumferential partition, the running normalized grayscale corresponding to the five radial layers is summed, and the summation result is divided by five to form the running normalized grayscale mean of the corresponding circumferential partition. For each discrete sampling time and each circumferential partition, subtract the normalized grayscale mean of the corresponding circumferential partition from one to form the running compression degree of the corresponding circumferential partition at the corresponding discrete sampling time. For each discrete sampling time and each circumferential partition, the normal sealing and pressing benchmark of the corresponding circumferential partition is subtracted from the running pressing degree of the corresponding circumferential partition to form the pressing deviation of the corresponding circumferential partition at the corresponding discrete sampling time. For each discrete sampling moment, the compression deviations corresponding to the seventy-two circumferential partitions are summed, and the summation result is divided by seventy-two to form the average compression deviation of the entire circle at the corresponding discrete sampling moment; For each discrete sampling moment, the compression deviations corresponding to the seventy-two circumferential partitions are compared, and the one with the highest value is taken as the local peak compression deviation for the corresponding discrete sampling moment. The door sealing anomaly score is generated based on the mean and peak values ​​of light leakage and compression deviation, and the sealing judgment boundary is generated based on the normalized grayscale of the normal baseline, the upper limit of normal light leakage in the circumferential direction, and the normal sealing compression baseline. Based on the circumferential micro-leakage and compression deviation, a circumferential anomaly positioning quantity is generated. The hatch sealing anomaly score is compared with the sealing judgment boundary to generate a hatch sealing status identifier. In the non-sealed state, anomaly circumferential partition number, anomaly circumferential partition center angle, circumferential angular distance and anomaly orientation set are generated. In the sealed state, a number indicating that there is no anomaly circumferential partition and an empty anomaly orientation set are generated. Anomaly type labels and operation permit labels are generated based on the hatch sealing status indicator, local peak light leakage amount, and local peak compression deviation amount, and operation monitoring records are output.

2. The method for monitoring the operation of kitchen waste disposal equipment based on machine vision according to claim 1, characterized in that, The process of selecting the door sealing ring area, establishing visual sampling coordinates, dividing the area into circumferential zones and radial layers, and generating circumferential sampling angles and sealing ring sampling points specifically includes: The door sealing ring area will be used as the monitoring target for the operation of the food waste disposal equipment. Set the center point of the sealing ring as the origin of the coordinate system, set the direction of the door hinge pointing to the latch as the positive direction of the horizontal coordinate, and set the direction of the image plane after rotating 90 degrees counterclockwise from the positive direction of the horizontal coordinate as the positive direction of the vertical coordinate. The sealing ring area is divided into seventy-two circumferential partitions and five radial layers. A set of indexes for circumferential partitions, a set of indexes for radial layers, and a set of indexes for normal sealing reference acquisition are established. The set of indexes for circumferential partitions contains consecutive integers from one to seventy-two, the set of indexes for radial layers contains consecutive integers from one to five, and the set of indexes for normal sealing reference acquisition contains consecutive integers from one to six. Each integer in the circumferential partition index set is used as the circumferential partition number, each integer in the radial hierarchical index set is used as the radial hierarchical number, and each integer in the normal closed benchmark acquisition index set is used as the normal closed benchmark acquisition number. Measure the inner boundary radius and outer boundary radius of the sealing ring, divide the radial width between the outer boundary radius and the inner boundary radius of the sealing ring into five equal-width radial intervals, and use the middle radius of each radial interval as the sampling radius of the corresponding radial layer; The full circle angle corresponding to twice pi is divided into seventy-two circumferential zones with equal angles, and the sampling angles corresponding to each circumferential zone are formed according to the order of the circumferential zone numbers; For each radial layer and each circumferential partition, corresponding sealing ring sampling points are generated according to the corresponding sampling radius, sampling angle, direction cosine of the sampling angle in the positive direction of the horizontal coordinate, and direction sine of the sampling angle in the positive direction of the vertical coordinate.

3. The method for monitoring the operation of kitchen waste disposal equipment based on machine vision according to claim 2, characterized in that, The acquisition of dark-field images, transmitted light reference images, normal sealing reference images, and cabin door sealing ring operation images, reading the grayscale of the sealing ring sampling points, and generating normal reference normalized grayscale and operation normalized grayscale through normalization processing, specifically includes: Set a fixed zero-prevention constant, the value of which is 10 to the power of negative 6; With the circumferential detection light source turned off, dark field images are acquired when the hatch is closed and the equipment is not feeding materials. The dark field images are then normalized to form a dark field normalized grayscale image. The grayscale value of each sealing ring sampling point in the dark field normalized grayscale image is read, and the upper limit of the read grayscale value is clipped. The upper limit of the clipping is the value after subtracting a fixed anti-zero constant from one, thus forming the dark field grayscale of the corresponding sealing ring sampling point. Turn on the circumferential detection light source and acquire a light transmission reference image when the hatch is open to the point where the sealing ring is unobstructed. Perform grayscale normalization processing on the light transmission reference image to form a light transmission reference normalized grayscale image. Read the grayscale value of each sealing ring sampling point in the light transmission reference normalized grayscale image and perform lower limit cropping on the read grayscale value. The lower limit of cropping is the value of the dark field grayscale of the corresponding sealing ring sampling point plus a fixed anti-zero constant, to form the light transmission reference grayscale of the corresponding sealing ring sampling point. With the hatch clean, the sealing ring free of debris, and the door lock fully engaged, six normal airtightness benchmark acquisitions are performed consecutively. Each acquisition forms a normal airtightness benchmark image. The gray values ​​of the sealing ring sampling points corresponding to each radial layer and each circumferential partition in each normal airtightness benchmark image are read to form the normal airtightness benchmark sampling gray values. After the food waste equipment enters the operation monitoring state, the operation images of the door sealing ring are collected according to the discrete sampling time. The gray values ​​of the corresponding sealing ring sampling points of each radial layer and each circumferential partition in each door sealing ring operation image are read to form the operation sampling gray value. For each discrete sampling moment, each radial layer and each circumferential partition, the dark field gray level of the corresponding sealing ring sampling point is subtracted from the running sampling gray level to form the running gray level difference. The dark field gray level is subtracted from the light-transmitting reference gray level of the corresponding sealing ring sampling point to form the reference gray level difference. The result of dividing the running gray level difference by the reference gray level difference is clipped from zero to one to form the running normalized gray level. For each normal sealed baseline acquisition, each radial layer, and each circumferential partition, the dark field gray level of the corresponding sealing ring sampling point is subtracted from the gray level of the normal sealed baseline sampling point to form a baseline gray level difference. The dark field gray level is subtracted from the light-transmitting reference gray level of the corresponding sealing ring sampling point to form a reference gray level difference. The result of dividing the baseline gray level difference by the reference gray level difference is clipped from zero to one to form the normal baseline normalized gray level.

4. The method for monitoring the operation of kitchen waste disposal equipment based on machine vision according to claim 3, characterized in that, The process involves generating a door sealing anomaly score based on the mean and peak values ​​of light leakage and compression deviation, and generating a sealing judgment boundary based on the normalized grayscale of the normal baseline, the upper limit of normal circumferential light leakage, and the normal sealing compression baseline. Specifically, this includes: For each discrete sampling moment, the average light leakage, local peak light leakage, average compression deviation, and local peak compression deviation are summed, and the sum is divided by four to form the hatch sealing anomaly score for the corresponding discrete sampling moment. For each normal sealed baseline acquisition, each circumferential partition and each radial layer, the upper limit of normal circumferential light leakage of the corresponding circumferential partition is subtracted from the normal baseline normalized grayscale, and the value after subtraction is subjected to positive over-limit processing to form the baseline light leakage over-limit value of the corresponding radial layer. For each normal closed baseline acquisition and each circumferential zone, the baseline light leakage exceeding the limit corresponding to the five radial layers is summed, and the summation result is divided by five to form the baseline light leakage amount of the corresponding circumferential zone under the normal closed baseline acquisition. For each normal sealed baseline acquisition and each circumferential partition, the normal baseline normalized grayscale values ​​corresponding to the five radial layers are summed, and the summation result is divided by five to form the single baseline normalized grayscale mean value of the corresponding circumferential partition. The single baseline normalized grayscale mean value of the corresponding circumferential partition is subtracted by one, and then the normal sealing and pressing baseline of the corresponding circumferential partition is subtracted to form the baseline pressing deviation of the corresponding circumferential partition under the normal sealed baseline acquisition. For each normal sealed baseline acquisition, the baseline leakage amounts corresponding to the seventy-two circumferential zones are summed and divided by seventy-two to form the baseline average leakage amount. The baseline leakage amounts corresponding to the seventy-two circumferential zones are compared and the highest value is taken to form the baseline peak leakage amount. The baseline compression deviation amounts corresponding to the seventy-two circumferential zones are summed and divided by seventy-two to form the baseline average compression deviation amount. The baseline compression deviation amounts corresponding to the seventy-two circumferential zones are compared and the highest value is taken to form the baseline peak compression deviation amount. For each normal airtight benchmark acquisition, the average light leakage, peak light leakage, average compression deviation, and peak compression deviation of the benchmark are summed, and the sum is divided by four to form the benchmark airtightness anomaly score corresponding to the normal airtight benchmark acquisition. The baseline airtightness abnormality scores corresponding to the six normal airtightness baseline collections are summed, and the summation result is divided by six to form the mean of the normal baseline scores; Calculate the absolute value of the difference between the baseline closure anomaly score and the mean normal baseline score corresponding to the six normal closure baseline collections. Sum the six absolute values ​​of the difference and divide the sum by six to form the average deviation of the normal baseline score. The highest value in the baseline airtightness anomaly score corresponding to the six normal airtightness benchmark acquisitions is added to the average deviation of the normal baseline score to form a candidate judgment value. The candidate judgment value is compared with the mean of the normal baseline score, and the highest value is selected. The highest value after comparison is then pruned with an upper limit of one to form the airtightness judgment boundary.

5. The method for monitoring the operation of kitchen waste disposal equipment based on machine vision according to claim 4, characterized in that, The process involves generating a circumferential anomaly location based on circumferential micro-leakage and compression deviation, comparing the hatch sealing anomaly score with the sealing judgment boundary to generate a hatch sealing status identifier, and generating anomaly circumferential partition numbers, anomaly circumferential partition center angles, circumferential angular distances, and anomaly orientation sets in the non-sealed state, and generating numbers indicating the absence of anomaly circumferential partitions and empty anomaly orientation sets in the sealed state. Specifically, this includes: For each discrete sampling time, the hatch sealing anomaly score is compared with the sealing judgment boundary. When the hatch sealing anomaly score is less than or equal to the sealing judgment boundary, the hatch sealing status flag at the corresponding discrete sampling time is set to one, and the hatch is recorded as being in a sealed state. For each discrete sampling time, the hatch sealing anomaly score is compared with the sealing judgment boundary. When the hatch sealing anomaly score is greater than the sealing judgment boundary, the hatch sealing status flag at the corresponding discrete sampling time is set to zero, and the hatch is recorded as being in an unsealed state. For each discrete sampling time and each circumferential partition, the circumferential micro-leakage amount and the compression deviation amount of the corresponding circumferential partition are summed, and the summation result is divided by two to form the circumferential anomaly location amount of the corresponding circumferential partition at the corresponding discrete sampling time. For each discrete sampling moment, when the door sealing status is marked as one, the abnormal circumferential partition number is set to zero, and it is recorded that there is no abnormal circumferential partition. For each discrete sampling moment, when the door sealing status indicator is zero, the circumferential abnormal positioning quantities corresponding to the seventy-two circumferential partitions are compared, and the circumferential partition number corresponding to the circumferential abnormal positioning quantity with the highest value is taken as the abnormal circumferential partition number. For each discrete sampling moment, when the door sealing status is marked as one, candidate angles are selected within the angle range of zero to two times pi, the circumferential angular distance corresponding to each candidate angle is set to zero, and the abnormal orientation set is set to an empty set. For each discrete sampling moment, when the door sealing status indicator is zero, the sampling angle corresponding to the abnormal circumferential partition number is taken as the center angle of the abnormal circumferential partition. Candidate angles are selected within the angle range of zero to twice pi. The absolute value of the angle difference between the candidate angle and the center angle of the abnormal circumferential partition is taken. The absolute value of the angle difference between the candidate angle and the center angle of the abnormal circumferential partition is then subtracted from the full circle angle corresponding to twice pi. The results of the absolute value of the angle difference between the candidate angle and the center angle of the abnormal circumferential partition are compared with the results of the full circle angle corresponding to twice pi minus the absolute value of the angle difference between the candidate angle and the center angle of the abnormal circumferential partition. The one with the lowest value is taken as the circumferential angular distance from the candidate angle to the center angle of the abnormal circumferential partition. For each discrete sampling moment, when the door sealing status is zero, candidate angles with circumferential angular distances less than the ratio of pi to 72 are selected from the angle range of zero to twice pi, and the selected candidate angles are combined into an abnormal orientation set.

6. The method for monitoring the operation of kitchen waste disposal equipment based on machine vision according to claim 5, characterized in that, The system generates anomaly type labels and operation permit labels based on the hatch sealing status indicator, local peak light leakage, and local peak compression deviation, and outputs operation monitoring records, specifically including: For each discrete sampling time, when the door sealing status is marked as one, the abnormality type label is set to zero, and the corresponding discrete sampling time is recorded as normal door sealing. For each discrete sampling moment, when the door sealing status is zero and the local peak light leakage is greater than or equal to the local peak compression deviation, the anomaly type label is set to one, and the corresponding discrete sampling moment is recorded as a light leakage-dominated sealing anomaly. For each discrete sampling moment, when the door sealing status is zero and the local peak light leakage is less than the local peak compression deviation, the anomaly type label is set to two, and the corresponding discrete sampling moment is recorded as a compression deviation-dominated sealing anomaly. For each discrete sampling time, the operation permission flag is set to the same value as the hatch sealing status flag. The state where the hatch sealing status flag is one is used as the operation monitoring permission condition. When the operation permission flag is one, the record meets the operation monitoring permission condition. When the operation permission flag is zero, the record does not meet the operation monitoring permission condition. For each discrete sampling moment, the discrete sampling moment number, the hatch sealing anomaly score, the sealing judgment boundary, the hatch sealing status identifier, the anomaly circumferential partition number, the anomaly location set, and the anomaly type label are written into the operation monitoring record.

Images