Target Detection Using Dilated Convolution for Robustness
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional target detection methods suffer from poor robustness and long processing times when determining the position of target objects in images.
Innovation Solution
A target detection method involving the extraction of first and second image features, followed by dilated convolution and classification and regression to determine candidate position parameters, which improves robustness and reduces detection time by expanding the receptive field.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional target detection method is used to search for locating points, then the position of target object can be determined, but the robustness is poor and detection time is long
Solution Approach 1:
The patent changes the parameter of receptive field size by applying dilated convolution with different dilation rates to image features at different stages. This allows the network to capture multi-scale contextual information, improving detection robustness while maintaining efficient processing speed through parameterized feature transformation rather than exhaustive searching
Solution Approach 2:
The patent introduces an intermediary mechanism by using dilated convolution as a bridge between raw image features and target position determination. This intermediary layer extracts enriched features with expanded receptive fields, enabling more robust detection without the time cost of conventional point-by-point searching
2Reliability
If dilated convolution is applied to expand receptive field, then detection robustness improves, but computational complexity increases
Solution Approach 1:
The patent segments the computational process by applying dilated convolution only to specific intermediate image features rather than the entire detection pipeline. This selective application expands the receptive field for critical features while avoiding unnecessary computational overhead in other stages, resolving the contradiction between robustness and complexity
Solution Approach 2:
The patent applies local quality by using different dilation rates at different network stages, tailoring the receptive field expansion to the specific needs of features at each level. This localized approach ensures computational resources are focused where they provide maximum robustness improvement, rather than uniformly increasing complexity throughout the system
3Measurement precision
If multiple image features are extracted and processed, then detection accuracy improves, but processing time increases
Solution Approach 1:
The patent maintains continuity of useful action by processing multiple image features through dilated convolution in a continuous, integrated manner rather than sequentially analyzing each feature independently. This allows parallel feature enrichment that improves detection accuracy while minimizing the time penalty through efficient batch processing of multiple features simultaneously
Data Source
AI summary
This application relates to a target detection method performed at a computer device. The method includes: obtaining a to-be-detected image; extracting a first image feature and a second image feature corresponding to the to-be-detected image; performing dilated convolution to the second image feature, to obtain a third image feature corresponding to the to-be-detected image; performing classification and regression to the first image feature and the third image feature, to determine candidate position parameters corresponding to a target object in the to-be-detected image and degrees of confidence corresponding to the candidate position parameters; and selecting a valid position parameter from the candidate position parameters according to their corresponding degrees of confidence, and determining a position of the target object in the to-be-detected image according to the valid position parameter. The solutions in this application can improve robustness and consume less time.