Object Tracking With Dynamic Thresholds for Close-Range Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Vehicles struggle to track targets at close range due to image recognition techniques becoming ineffective when targets come into close proximity, leading to potential collisions.
Innovation Solution
A computer-implemented method that determines the probability of object existence and adjusts this probability based on close-range criteria, using a tracking device with modules to extend tracking duration even when detection fails.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional image recognition techniques are used to track targets, then tracking works at normal distances, but tracking fails when targets come into close proximity to the cameras
Solution Approach 1:
The system dynamically adjusts the probability threshold for target detection based on the determined distance to the target. When the target is in close proximity, a modified lower threshold is applied, allowing the tracking to continue despite degraded image quality. This dynamic adaptation resolves the contradiction by making the detection criteria flexible rather than fixed.
Solution Approach 2:
The invention changes the detection parameter (probability threshold) based on the distance parameter. By modifying the threshold value according to the target's distance from the camera, the system maintains tracking reliability across different ranges, particularly improving performance at close distances where conventional fixed thresholds fail.
2Reliability
If image recognition is applied to close-range targets, then collision avoidance is improved, but image quality becomes insufficient due to out-of-focus or partial capture
Solution Approach 1:
The system applies a cushioning effect by introducing a distance-based probability adjustment mechanism that compensates for the expected degradation in image quality at close ranges. Before the detection fails completely, the system proactively adjusts the threshold to maintain tracking, cushioning against the anticipated loss of measurement precision.
Solution Approach 2:
The invention accepts that conventional image recognition will fail at close ranges and uses a simpler, more robust probability-based tracking approach for these specific conditions. This disposable approach abandons the need for high-precision image recognition when distance criteria indicate close proximity, relying instead on the adjusted probability model that is less sensitive to image quality degradation.
3Measurement precision
If standard probability thresholds are used for object detection, then detection is reliable at normal distances, but detection fails when objects are in close proximity
Solution Approach 1:
The detection system transitions from static to dynamic by continuously adjusting the probability threshold based on real-time distance measurements. This dynamic adaptation allows the same detection system to maintain high precision across varying distances, resolving the contradiction between normal-distance reliability and close-range detection capability.
Solution Approach 2:
The invention creates a universal detection mechanism that functions effectively across both normal and close distances by incorporating distance-based threshold adjustment. The single detection system performs multiple functions: using standard thresholds at normal distances and modified thresholds at close distances, eliminating the need for separate detection systems for different ranges.
Data Source
Figure 1~3
Figure 4A~4B
Figure 5
AI summary
According to various embodiments, there is provided a computer-implemented method for tracking an object. The method includes receiving a real-time video feed. The method further includes, for each frame of the video feed: determining a probability of existence of the object in the frame, determining whether a probability of existence of the object in the frame falls below a probability threshold, determining whether the object fulfils a set of close-range criteria based on determining that the probability of existence is lower than the probability threshold, and generating a adjusted probability of existence based on the determination of whether the object fulfils the set of close-range criteria.