Radar Object Classification via Composite Plot Spectrum
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Low-resolution radar systems with high sensitivity face a high number of incorrect detections, leading to unreliable object tracking due to false positives from unwanted objects, which conventional methods fail to adequately address.
Innovation Solution
The method involves forming a composite 'plot spectrum' by adding Doppler filter amplitudes from adjacent bursts and filling gaps with noise values, allowing for improved object classification by reducing spectral components from neighboring objects, and using a modified K-Nearest-Neighbor classification with individual normalization and feature determination for enhanced accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If high sensitivity is used in radar systems, then detection capability is improved, but false detections increase
Solution Approach 1:
The radar signal processing is segmented into multiple coherent integration intervals (bursts), with each burst processed independently to generate individual range reports. These segmented detections are then compiled by a cluster algorithm to form plots, allowing false detections to be identified and excluded through geometric association criteria across multiple segments rather than relying on a single sensitive detection.
Solution Approach 2:
A cluster algorithm acts as an intermediary between the sensitive radar detection system and the tracking system. This intermediary processes individual range reports from multiple bursts, applies geometric association criteria, and generates validated plots for tracking. The intermediary filters out false detections while preserving true targets, resolving the contradiction between high sensitivity and false alarm rates.
2Device complexity
If low-resolution radar systems are used, then system complexity is reduced, but object classification accuracy deteriorates
Solution Approach 1:
The system compensates for low spatial resolution by introducing the time dimension through multiple coherent integration intervals. By processing detections across multiple bursts and applying cluster algorithms that consider geometric associations in range-Doppler-azimuth space, the system achieves improved object classification accuracy without increasing the physical resolution capabilities of the radar hardware.
Solution Approach 2:
The cluster algorithm performs preliminary processing and validation of range reports before they are used for tracking. By pre-processing detections from multiple bursts and applying geometric association criteria in advance, the system prepares high-quality plot data that enables accurate object classification even with low-resolution input data from simple radar systems.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach significantly reduces false detections and improves object classification quality even in low-resolution radar systems, ensuring accurate tracking by distinguishing desired from undesired objects.
Implementation Method 1
In the 'Doppler' dimension, a defined resolution by the radar is also required; the individual resolution cells are referred to below as 'Doppler filters'. The contents of the Doppler filter are the levels per range cell, sorted according to the frequency shift of the radar echo signal. The Doppler filters correspond to different radial velocities of a reflecting object.
Data Source
Figure 1~2
Figure 3
Figure 4
AI summary
The invention relates to a method for classifying radar objects, wherein: - in an input channel within a plurality of successive coherent integration intervals (bursts) a received signal is processed and object detection is carried out using Doppler filter amplitudes, generating range reports; - a plurality of range reports are compiled by a clustering algorithm to form a plot; - Doppler filter amplitudes of the two azimuthally adjacent bursts from the same distance cell are added to each range report; - a subset of the range reports is formed bythat, starting with the range report at the location of a plot centroid, neighboring range reports from a defined number of bursts are sorted by azimuth position in a matrix of dimension azimuth x distance, with the plot centroid at the center; the filter amplitudes associated with the range reports are arranged from the area of the plot centroid ± a defined number of bursts according to the burst sequence; a composite plot spectrum is formed as a vector of range reports from the aforementioned matrix, initiated with the distance cell of the plot centroid; subsequently, for all burst positions for which no range report exists in the aforementioned distance cell, the nearest range report is inserted in adjacent distance cells; remaining gaps are filled with the adjacent filter amplitudes of the used range reports; and finally, any remaining gaps are filled with a suitable noise value.