Search

CN-121998974-A - Image processing method for investigation of heating and ventilation equipment

CN121998974ACN 121998974 ACN121998974 ACN 121998974ACN-121998974-A

Abstract

The invention discloses an image processing method for investigation of heating and ventilation equipment, and in particular relates to the technical field of image processing of investigation of heating and ventilation equipment, which comprises the steps of acquiring a field image shot by target heating and ventilation equipment, executing equipment positioning processing on the field image, determining an equipment area corresponding to the target heating and ventilation equipment, extracting a state object area in the equipment area, and outputting a state object area set; by positioning a state object area in an image of the heating and ventilation equipment and positioning a peripheral evidence area according to a state formation mechanism corresponding to the object type, the attachment change, the abrasion change, the corrosion change, the connection change or the blocking change in the peripheral evidence area are identified and integrated, and the states of filter screen filth blockage, valve clamping stagnation, part aging and the like are judged according to the peripheral change evidence.

Inventors

  • YE MINGSHU
  • CHENG KUI
  • DENG DONGMING

Assignees

  • 厦门金名节能科技有限公司

Dates

Publication Date
20260508
Application Date
20260407

Claims (9)

  1. 1. An image processing method for a heating ventilation equipment survey, comprising: S1, acquiring a field image shot by a target heating and ventilation device, performing device positioning processing on the field image, determining a device region corresponding to the target heating and ventilation device, extracting a state object region in the device region, and outputting a state object region set; the state object region represents a candidate object region in which object validity determination processing is performed on a candidate object region set, and in which object morphological conditions and positional relationship conditions are satisfied; S2, performing object type identification processing on each state object region in the state object region set, determining the state object type corresponding to each state object region, positioning a peripheral evidence region bearing residual traces in a peripheral adjacent range according to a state forming mechanism corresponding to each state object type, and outputting a peripheral evidence region set; s3, performing image recognition processing on each peripheral evidence region set, recognizing adhesion change, abrasion change, rust change, connection change or blocking change in the peripheral evidence regions, merging the peripheral evidence regions into corresponding state object regions according to the adjacent relation, and outputting the peripheral evidence change set; S4, performing state judgment processing on each peripheral change evidence set according to the corresponding state object type, judging the dirty blocking state of the filter screen or the clamping state of the valve or the aging state of the electrical component based on the peripheral change evidence set, and outputting an object state result; In S4, it includes: s4-1, acquiring a peripheral change evidence set and an object type result corresponding to each state object area, respectively calculating change strength, change distribution range and change continuity degree for adhesion change, abrasion change, corrosion change, connection change or blocking change in each peripheral change evidence set, and outputting a change judgment feature set corresponding to each state object area; S4-2, invoking a state judgment rule corresponding to the object type identifier according to the object type result, executing dirty blockage matching processing on a change judgment feature set corresponding to the filter screen, executing clamping stagnation matching processing on a change judgment feature set corresponding to the valve, executing aging matching processing on a change judgment feature set corresponding to the electric component, and outputting a state candidate result corresponding to each state object area; S4-3, performing conflict checking processing on each state candidate result, judging whether each change judging feature in the same state object area meets the preset co-occurrence condition and the rejection condition, reserving the state candidate result meeting the preset co-occurrence condition and not meeting the rejection condition as an object state result, and outputting the object state result; S5, performing equipment aggregation processing on the object state results, aggregating the object state results belonging to the same target heating and ventilation equipment to form a investigation state result corresponding to the target heating and ventilation equipment, and outputting an operation state identification result of the target heating and ventilation equipment according to the investigation state result.
  2. 2. An image processing method for a hvac survey according to claim 1, characterized in that: The step S1 includes: s1-1, performing equipment contour recognition and structure boundary recognition processing on a field image, determining an outer contour range and an inner structure distribution range of a target heating and ventilation device in the field image, and outputting an equipment boundary result; S1-2, performing region constraint decomposition processing on a device boundary result, decomposing the device boundary result into a plurality of candidate object regions according to the installation position relationship and the structure adjacent relationship of a preset state object in the target heating and ventilation device, and outputting a candidate object region set; S1-3, performing object validity judgment processing on the candidate object region set, reserving the candidate object region meeting the object form condition and the position relation condition as a state object region, and outputting the state object region set.
  3. 3. An image processing method for a hvac survey according to claim 2, characterized in that: the step S2 includes: s2-1, acquiring a target state object area in a state object area set, executing object type identification processing on the target state object area, generating an object type identifier corresponding to the target state object area, and outputting an object type result; S2-2, calling states corresponding to object type identifiers one by one according to object type results to form a mechanism mapping relation, solving a residual trace bearing element set corresponding to the object type identifiers, and generating candidate peripheral regions in the peripheral adjacent range of the target state object region by taking each residual trace bearing element as a constraint, and outputting the candidate peripheral region set; S2-3, performing region screening processing on each candidate peripheral region in the candidate peripheral region set, calculating the region distance, the relative orientation and the boundary connection relation between each candidate peripheral region and the target state object region, reserving the candidate peripheral region, the region distance of which falls into a preset adjacent distance range, the relative orientation accords with the preset orientation rule corresponding to the object type, and the boundary connection relation accords with the structure connection rule corresponding to the object type, as a peripheral evidence region, and outputting the peripheral evidence region set.
  4. 4. An image processing method for a hvac survey according to claim 3, characterized in that: the process of outputting the candidate surrounding area set in S2-2 further includes: S2-21, obtaining an object type result, reading a state formation mechanism mapping relation corresponding to object type identifiers one by one, performing element calculation on the state formation mechanism mapping relation to obtain a residual trace bearing element set, generating a constraint set containing space orientation constraint, structural connection constraint and visibility constraint for each bearing element in the residual trace bearing element set, and outputting a bearing element constraint set; S2-22, acquiring a target state object region and limiting the peripheral adjacent range of the target state object region, extracting boundary and structure primitives of an image region in the peripheral adjacent range to form a candidate region diagram, respectively calculating azimuth deviation, connection inconsistency and visibility missing quantity of each candidate region in the candidate region diagram to form constraint violation costs, determining a search expansion sequence from low to high with constraint violation costs when the constraint search is carried out on the candidate region diagram, generating candidate peripheral regions item by item on the premise of meeting space azimuth constraint and structure connection constraint, and outputting an initial candidate peripheral region set ordered according to the constraint violation costs; S2-23, performing iterative confidence updating processing on the initial candidate surrounding area set, converting the satisfaction degree between each initial candidate surrounding area and the bearing element constraint set into confidence degrees, multiplexing the previous round of confidence degrees in iteration, performing pruning on the candidate area map to control the consumption of computing resources, and outputting the candidate surrounding area set when the confidence ordering of two continuous rounds of iteration is kept unchanged and the candidate set is not added and deleted any more.
  5. 5. An image processing method for a hvac survey according to claim 4, wherein: the step S3 includes: s3-1, acquiring a peripheral evidence region set and performing image standardization processing on each peripheral evidence region, wherein the image standardization processing comprises brightness normalization on uneven illumination, suppression separation on reflective highlighting, definition compensation on motion blur and outputting the standardized peripheral evidence region set; S3-2, performing change type identification processing on each peripheral evidence region in the standardized peripheral evidence region set, wherein the change type identification processing comprises the steps of respectively calculating the increment of texture density of the attachment change, the edge continuity breaking quantity of the abrasion change, the color diffusion connecting flux of the rust change, the structural alignment offset of the connection change and the form passing notch quantity of the blocking change, and outputting a peripheral change candidate set according to the meeting relation between each change quantity and the corresponding change type discrimination condition.
  6. 6. An image processing method for a hvac survey according to claim 5, wherein: in S3, further comprising: s3-3, performing adjacent merging processing on the peripheral change candidate set, namely calculating the area spacing and the relative orientation between each peripheral change candidate and the corresponding state object area, checking the structural adjacent relation of the peripheral change candidate set, reserving the peripheral change candidates which simultaneously meet the condition that the area spacing is in an adjacent range, the relative orientation accords with the object type orientation rule and the structural adjacent relation is established, merging the peripheral change candidates into the corresponding state object area, and outputting a peripheral change evidence set.
  7. 7. An image processing method for a hvac survey according to claim 6, characterized in that: The process of outputting the state candidate result corresponding to each state object region in S4-2 further includes: S4-21, acquiring an object type result and a change judgment feature set, calling a corresponding state judgment rule according to an object type identifier, disassembling the state judgment rule into a feature satisfaction condition, a feature rejection condition and a feature co-occurrence condition, and outputting a rule constraint set corresponding to each state object region; S4-22, executing candidate solving processing on the change judging feature set of each state object area, respectively calculating the support degree, the conflict degree and the missing degree of each change judging feature to each state candidate category in the rule constraint set, executing sorting on each state candidate category according to the sequence of the support degree from high to low, the conflict degree from low to high and the missing degree from low to high, keeping the state candidate category with the support degree sorted forward and meeting the conditions that the conflict degree is not higher than other state candidate categories and the missing degree is not higher than other state candidate categories in the corresponding sorting relation at the same time as an initial state candidate result, and outputting an initial state candidate result set.
  8. 8. An image processing method for a hvac survey according to claim 7, wherein: The process of outputting the state candidate result corresponding to each state object region in S4-2 further includes: S4-23, performing iterative correction processing on the initial state candidate result set, and taking the support degree, the conflict degree and the missing degree in each initial state candidate result as confidence update input; And performing confidence correction on each initial state candidate result by combining the co-occurrence consistency between the peripheral change evidences adjacent to the current state object region, and keeping the initial state candidate result with unchanged continuous two-round confidence ordering and no longer changed state candidate category as a state candidate result to output the state candidate result corresponding to each state object region.
  9. 9. An image processing method for a hvac survey according to claim 8, characterized in that: In S5, it includes: S5-1, acquiring state results of each object and corresponding state object region position information, performing in-device classification processing on the state results of each object according to a position attribution relation and an object type attribution relation in a device region, and outputting a device state unit set; S5-2, performing association aggregation processing on the equipment state unit sets, calculating co-occurrence relations, conflict relations and supplementary relations among the equipment state units, performing combination reconstruction on the equipment state units according to the co-occurrence relations, the conflict relations and the supplementary relations, and outputting a investigation state result set; S5-3, executing operation state identification processing on the investigation state result set, determining the operation state category corresponding to the target heating and ventilation equipment according to the combined result of the equipment state units in the investigation state result set, and outputting the operation state identification result of the target heating and ventilation equipment.

Description

Image processing method for investigation of heating and ventilation equipment Technical Field The invention relates to the technical field of image processing for investigation of heating and ventilation equipment, in particular to an image processing method for investigation of heating and ventilation equipment. Background In the existing investigation and energy-saving diagnosis work of heating and ventilation equipment, the main current practice in the industry is mainly to solve the problem that the state of field equipment is difficult to obtain quickly, usually, a worker uses a mobile phone or a handheld terminal to shoot equipment pictures, then performs target positioning and image recognition on a filter screen, a valve, a dial and electric components in the pictures, and further gives out the conclusions of whether the filter screen is dirty and blocked, whether the valve is blocked, whether dial parameters are abnormal, whether the components are aged or not and the like; Taking the air conditioner room inspection of a market or an office building as an example, staff often needs to rapidly complete the shooting of a plurality of devices along a narrow channel under the condition of no shutdown, and the on-site hard constraint that the light is insufficient, the light reflection is obvious, the shielding is frequent, the shooting angle is difficult to ensure and the disassembly and the verification are impossible exists, and the image result formed by one shooting is required to be directly used for the subsequent operation judgment and the energy-saving strategy generation; Under the constraint, the main stream method can stably expose observable defects, namely, although the appearance of a valve body, a filter screen surface or a part can be shot in a photo, the identification result often shows that the conclusion of the same object in different photos is inconsistent, or the filter screen appears blackish but cannot be stably judged as long-term dirty blockage, or the valve appears at a certain angle but cannot be stably judged as stuck, or the part appears older but cannot be stably judged as high in aging degree, because key evidence of the states is not always concentrated on the object body completely, but more on the connection parts, limit contact parts, frame attachment distribution parts, air duct inlet adjacent parts, the change of the associated areas such as a fastener and a nameplate and the like, and the existing method usually only directly identifies the object body, and lacks identification and merging of the peripheral evidence, so that state judgment is unstable and difficult to recheck; the application aims to solve the technical problem of how to realize stable identification of states such as filter screen filth blockage, valve clamping stagnation, part aging and the like by utilizing visible change evidence in a peripheral region of an object under the conditions of limited shooting and incapability of disassembling and verifying in a investigation image of heating and ventilation equipment. Disclosure of Invention In order to overcome the defects in the prior art, the embodiment of the invention provides an image processing method for investigation of heating and ventilation equipment, which is characterized in that a state object area in an image of the heating and ventilation equipment is positioned, a peripheral evidence area is positioned according to a state forming mechanism corresponding to an object type, and the attachment change, the abrasion change, the rust change, the connection change or the blocking change in the peripheral evidence area are identified and integrated, so that the states of filter screen filth blockage, valve clamping stagnation, part aging and the like are judged according to the peripheral change evidence. In order to achieve the above purpose, the invention provides a technical scheme that the image processing method for investigation of heating ventilation equipment comprises the following steps: S1, acquiring a field image shot by a target heating and ventilation device, performing device positioning processing on the field image, determining a device region corresponding to the target heating and ventilation device, extracting a state object region in the device region, and outputting a state object region set; the state object region represents a candidate object region in which object validity determination processing is performed on a candidate object region set, and in which object morphological conditions and positional relationship conditions are satisfied; S2, performing object type identification processing on each state object region in the state object region set, determining the state object type corresponding to each state object region, positioning a peripheral evidence region bearing residual traces in a peripheral adjacent range according to a state forming mechanism corresponding to each state object type, and outpu