A Vision Inspection Method for Frame Quality of Heavy-Duty Special Vehicles

Technical Field

The present invention belongs to the technical field of vehicle quality inspection, in particular to a vision inspection method for frame quality of heavy-duty special vehicles.

Background Art

After cargo transportation, heavy-duty special vehicles need return-to-facility maintenance with priority inspection on frame quality. Conventional frame inspection requires disassembly of vehicle body and other auxiliary components before frame detection, which complicates inspection workflows and raises labor costs for subsequent reassembly. Moreover, existing inspection schemes fail to implement risk prediction, posing hidden threats to safe operation of heavy-duty special vehicles.

China Patent Publication No. CN106767550A discloses an on-line inspection system for heavy truck frame assemblies, which consists of a travelling support set for frame conveying, wireless digital scribers measuring flatness, straightness and coaxiality of left/right longitudinal girders, CCD digital cameras for frame image acquisition, a data analysis platform generating 3D coordinate data of scribe contact points, a display screen and an inspection rack mounted at rivet trimming station for fixing display and CCD cameras.

It can be concluded that existing vehicle inspection technologies cannot realize efficient and precise detection and fail to guarantee driving safety of vehicles.

Summary of the Invention

Therefore, the objective of the present invention is to provide a vision inspection method for frame quality of heavy-duty special vehicles to overcome defects of low efficiency and insufficient safety assurance in conventional vehicle quality inspection technologies.

To fulfill the above objective, the proposed vision inspection method comprises the following steps:

1. Scan all key positions on frames of returned heavy-duty special vehicles to acquire practical inspection images;

2. Perform feature extraction on each practical inspection image to obtain practical inspection information, followed by sequential defect analysis and defect marking;

3. Implement primary quality assessment for single-point quality of each key position and overall frame quality based on defect marking results;

4. Number and integrate all practical inspection information in chronological historical serial number order to build a practical inspection database;

5. Compare the practical inspection database with historical inspection database to conduct risk analysis and risk marking;

6. Implement secondary quality assessment for single-point quality of each key position and overall frame quality according to risk marking results;

7. Carry out correlation marking for all key positions by combining primary and secondary assessment results, and output final inspection results of respective key positions according to correlation marking.

Further Improvement on Feature Extraction

The feature extraction procedure from practical inspection images includes:

1. Identify practical overexposed regions and view-limited regions within practical inspection images;

2. Judge whether both overexposed regions and view-limited regions satisfy feature extraction criteria;

3. If neither region meets the criteria, determine the corresponding practical inspection image satisfies single-trigger rescan condition;

4. Count the quantity of images complying with single-trigger rescan condition and select either single-point rescan mode or full-frame rescan mode accordingly.

The feature extraction criteria are defined to quantify scanning quality of collected inspection images.

Further Improvement on Defect Analysis & Marking

Procedures of sequential defect analysis and defect marking on practical inspection information:

1. Conduct defect analysis for each piece of practical inspection information to identify abnormalities; if abnormal defects exist, confirm defect location and defect classification;

2. Evaluate actual defect severity based on defect location and classification data;

3. Define defect grades including Grade-I Defect and Grade-II Defect according to severity, and complete defect marking corresponding to determined grades.

Abnormality judgment via defect analysis is realized by edge analysis, deformation analysis and texture analysis on inspection data. An abnormality is confirmed once any single analysis output fails qualification threshold:

• Edge analysis: evaluates integrity of key positions;

• Deformation analysis: evaluates load-bearing stability of key positions;

• Texture analysis: evaluates connection tightness of key positions.

Defect severity evaluation steps:
Determine total defect count, individual defect coverage and corresponding defect types from edge, deformation and texture analysis outcomes; calculate practical defect score using defect coverage, defect category and defect quantity to quantify defect severity.

Defect location parameters: individual defect coverage + total defect quantity
Defect classification parameters: categorized defect types

Further Improvement on Primary Quality Assessment

Single-point primary quality judgment for each key position:

• Grade-I Defect → Primary Pass for single-point quality;

• Grade-II Defect → Primary Fail for single-point quality.

Overall frame primary quality assessment:
Count total number of primary-passed key positions (Pass Count 1) and primary-failed key positions (Fail Count 1):

1. Overall frame is unqualified if Pass Count 1 ≥ preset First Standard Quantity;

2. Overall frame is unqualified if Fail Count 1 ≥ 1.

Further Improvement on Risk Analysis and Risk Marking via Database Comparison

1. Query historical defect score of each key position from historical inspection database;

2. Compare historical and practical defect scores for risk analysis to assign risk level (High Risk / Low Risk);

3. Complete targeted risk marking by determined risk level.

Further Improvement on Secondary Quality Assessment Based on Risk Marking

Single-point secondary quality judgment for each key position:

• Low Risk → Secondary Pass for single-point quality;

• High Risk → Secondary Fail for single-point quality.

Count total number of secondary-passed key positions (Pass Count 2) and secondary-failed key positions (Fail Count 2):

1. Overall frame is unqualified if Pass Count 2 ≥ preset Second Standard Quantity;

2. Overall frame is unqualified if Fail Count 2 ≥ 1.

Further Improvement on Correlation Marking and Final Result Output

Correlation marking rules combining primary and secondary assessment outcomes:

1. Both primary and secondary pass → Final Qualified;

2. Exactly one fail among primary/secondary assessment → Class-I Unqualified;

3. Both primary and secondary fail → Class-II Unqualified.

Corresponding maintenance decisions:

• Final Qualified: No Maintenance Required;

• Class-I Unqualified: Immediate Maintenance;

• Class-II Unqualified: Immediate Replacement.

Beneficial Effects Compared with Prior Art

1. Full-coverage scanning and feature extraction realize comprehensive inspection on all frame key positions for reliable overall quality evaluation; automated processing eliminates human error and boosts inspection accuracy & efficiency;

2. Cross-reference between real-time and historical inspection databases enables quantitative risk prediction, identifies potential safety hazards and guides preventive maintenance to reduce accident probability; changing trends of frame performance are tracked to support follow-up quality optimization;

3. Dual assessment (primary + secondary) with hierarchical marking outputs differentiated maintenance schemes for targeted repair, cuts unnecessary maintenance expense and optimizes operating cost;

4. Pre-judgment on overexposure and view restriction avoids invalid image data, triggers selective partial or full rescan to save scanning resources and improve data availability;

5. Triple-dimensional defect detection (edge / deformation / texture) comprehensively checks integrity, bearing performance and connection stability, realizes graded defect management to avoid over-maintenance or insufficient repair and curbs expansion of latent failures;

6. Quantified scoring of defects and historical trend-based risk grading enable predictive maintenance, prioritize overhaul on high-risk locations and enhance structural safety of heavy-duty vehicle frames;

7. Complete inspection data archiving forms traceable quality records for long-term maintenance reference and continuous process improvement.

Brief Description of Drawings

FIG.1: Flowchart of the vision inspection method for heavy-duty special vehicle frame quality;
FIG.2: Flowchart of risk analysis and risk marking procedure;
FIG.3: Flowchart of secondary quality assessment procedure;
FIG.4: Flowchart of correlation marking procedure for final result classification.

Detailed Description of Preferred Embodiments

The invention is described in detail below with reference to attached drawings and preferred embodiments. The described embodiments serve only for explanation rather than limitation of the invention scope. Terms including upper, lower, left, right, inner and outer refer to positional relations shown in attached drawings for descriptive convenience instead of mandatory fixed structural layout of devices.

Step S100

Scan all key positions on returned heavy-duty vehicle frames to acquire practical inspection images.

Step S200 Feature Extraction, Defect Analysis and Marking

S211~S214 Image Quality Screening and Rescan Trigger Rules

1. Calculate first area ratio = Overexposed Area / Total Image Area; overexposed region fails criteria if ratio ≥10%;

2. Calculate second area ratio = View-Limited Area / Total Image Area; view-limited region fails criteria if ratio ≥10%;

3. Single-trigger rescan condition is satisfied only when both two ratios exceed 10%;

4. Compute third ratio = Quantity of Single-Condition Images / Total Images:

￮ Ratio ≥20% → Full-frame Global Rescan Mode;

￮ Ratio ＜20% → Selective Single-Point Rescan Mode only for non-compliant key positions.

S221~S223 Defect Scoring & Grade Classification

• Edge anomaly: actual damage range ≥ standard damage threshold (Integrity Defect);

• Deformation anomaly: actual deformation range ≥ standard deformation threshold (Stability Defect);

• Texture anomaly: actual crack range ≥ standard crack threshold (Connection Defect);

Defect Score Formula:
Defect Score = (Damage Range × Integrity Coeff + Deformation Range × Stability Coeff + Crack Range × Connection Coeff) × Defect Count Coefficient
Preset fixed coefficients: Integrity Coeff=0.6, Stability Coeff=0.7, Connection Coeff=0.7; Defect Count Coefficient positively correlates with total detected defects (e.g. 0.8 when defect quantity=2).

• Defect Score ≥ Standard Threshold (0.3~0.5, adjustable by vehicle service life and factory standard) → Grade-I Defect (Severe Hazard);

• Defect Score ＜ Standard Threshold → Grade-II Defect (Potential Risk).

Step S300 Primary Quality Determination

Grade-I → Single-point Primary Pass; Grade-II → Single-point Primary Fail. Count Pass Count 1 and Fail Count 1 for overall frame qualification judgment per preset First Standard Quantity.

Step S400 Build Practical Inspection Database

Index and archive all valid inspection records sequentially by original historical serial numbers.

Step S500 Historical Data Comparison & Risk Rating

Score Difference = Practical Defect Score − Historical Defect Score

• Score Difference ≥ preset First Evaluation Threshold → High Risk;

• Score Difference ＜ preset First Evaluation Threshold → Low Risk.
Complete corresponding risk marking after level confirmation.

Step S600 Secondary Quality Determination Based on Risk Label

Low Risk → Secondary Pass; High Risk → Secondary Fail. Count Pass Count 2 and Fail Count 2 for overall frame qualification judgment against preset Second Standard Quantity; any single high-risk key point renders the whole frame unqualified.

Step S700 Correlation Marking and Final Maintenance Output

• Dual Pass: Final Qualified → No Maintenance Required;

• One Single Fail: Class-I Unqualified → Immediate Maintenance;

• Dual Fail: Class-II Unqualified → Immediate Replacement.

All coefficient values in calculation formulas are positive unless otherwise specified; weighting coefficients are configured to unify calculation dimension and adjust scoring amplitude, adjustable flexibly per practical production standard.

While preferred embodiments of the invention have been described with reference to attached drawings, the protection scope of the invention is not limited to such specific embodiments. Equivalent modifications or substitutions made on technical features without departing from the core principle of the invention shall all fall within the appended claim scope of the present invention.

The foregoing descriptions are preferred embodiments of the invention and shall not be construed as restriction. All modifications, equivalent substitutions and improvements made within the spirit and principle of the invention are covered by its protection scope.