US 8,948,522 B2 · Adaptive threshold for object detection

Abstract

Systems and methods for developing and using adaptive threshold values for different input images for object detection are disclosed. In embodiments, detector response histogram-based systems and methods train models for predicting optimal threshold values for different images. In embodiments, when training the model, an optimal threshold value for an image is defined as the value that maximizes the reduction of false positive image patches while preserving as many true positive image patches as possible. Once trained, the model may be used to set different threshold values for different images by inputting a detector response histogram for the image patches of an image into the model to determine a threshold value for detection.

Drawings

Drawings are temporarily unavailable. Open the PDF at USPTO

Drawings

Claims13

Dependent claims

A computer-implemented method to detect an object in an input image, the method comprising:

submitting said input image to a classifier for initial object detection, said classifier being already trained to detect said object, said classifier identifying multiple image regions within said input image and assigning each image region a detector response value indicative of a likelihood of the object, being within the image region, each image region whose detector response value is not less than a first threshold value being classified as containing the object and being designated a candidate patch;
collecting candidate patches to form a set of candidate patches from the input image, each candidate patch having its corresponding detector response value associated with it;
using a trained model to determine a second threshold value based on the detector response values candidate of the candidate patches in the set of candidate patches;
re-classifying all the candidate patches using the second threshold value, wherein each candidate patch whose associated detector response value is not less than the second threshold value is classified as containing the object, and each candidate patch whose associated detector response value is less than the second threshold value is re-classified as not containing the object.

1 more element

A non-transitory computer-readable medium comprising one or more sets of instructions which, when executed by one or more processors, causes the one or more processors to perform a method for detecting an object in an input image comprising the steps of:

submitting said input image to a classifier for initial object detection, said classifier being already trained to detect said object, said classifier identifying multiple image regions within said input image and assigning each image region a detector response value indicative of a likelihood of the object being within the image region, each image region whose detector response value is not less than a first threshold value being classified as containing the object and being designated a candidate patch;
collecting candidate patches to form a set of candidate image patches from the input image, each candidate image patch having its corresponding detector response value associated with it;
computing a normalized histogram of the associated detector response values of the candidate image patches in the set of candidate image patches;
using an adaptive threshold model to determine a second threshold value based on normalized histogram;
re-classifying one or more selected candidate image patches using the second threshold value, wherein each selected candidate image patch whose associated detector response value is not less than the second threshold is classified as containing the object, and each selected candidate image patch whose associated detector response value is less than the second threshold value is re-classified as not containing the object.

2 more elements

Dependencies are inferred from the claim text.

Description4,354 words

Summary1,102 words

#1. Field of Invention

#The present patent document is directed towards systems and methods for object detection. More particularly, the present patent document is directed towards systems and methods that provide adaptive threshold values for different input images.

#2. Description of the Related Art

Brief description of the drawings10 figures

#Reference will be made to embodiments of the invention, examples of which may be illustrated in the accompanying figures, in which like parts may be referred to by like or similar numerals. These figures are intended to be illustrative, not limiting. Although the invention is generally described in the context of these embodiments, it should be understood that it is not intended to limit the scope of the invention to these particular embodiments.

#FIG. 1 depicts a typical ROC curve for image detection, in which a single threshold value on the curve is selected for all input images.

#FIG. 2 depicts a method for training a model for adaptively determining a detection threshold value according to embodiments of the present invention.

Detailed description2,936 words

#In the following description, for purposes of explanation, specific details are set forth in order to provide an understanding of the invention. It will be apparent, however, to one skilled in the art that the invention can be practiced without these details. Furthermore, one skilled in the art will recognize that embodiments of the present invention, described below, may be implemented in a variety of ways, including software, hardware, or firmware, or combinations thereof. Accordingly, the figures described herein are illustrative of specific embodiments of the invention and are meant to avoid obscuring the invention.

#It shall be noted that some of the steps presented herein may be performed in different order but achieve the same or similar results. One skilled in the art shall recognize that certain steps may be performed before, after, or at the same time as other steps. Accordingly, the listing of steps, including in the claims, does not imply a required sequential order.

#Components, or modules, shown in block diagrams are illustrative of exemplary embodiments of the invention and are meant to avoid obscuring the invention. It shall also be understood that throughout this discussion that components may be described as separate functional units, which may comprise sub-units, but those skilled in the art will recognize that various components, or portions thereof, may be divided into separate components or may be integrated together, including integrated within a single system or component. It should be noted that functions or operations discussed herein may be implemented as components or modules.

Facts

Application

13/198,412Filed 4 Aug 2011

Published

US 2013/0034263 A17 Feb 2013

Granted

3 Feb 2015

Term ends

27 Sep 2033

Estimated · how

20 years from earliest known nonprovisional domestic filing plus recorded patent term adjustment, including 785 days of adjustment. Maintenance-fee lapses, terminal disclaimers and statutory extensions have not been checked. The continuity chain may be incomplete; this is not an enforceability determination.

Classification6 codes

G06V 10/7715
Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or…:…
G06F 18/2134
Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods: based on separation criteria, e.g.…
G06F 18/2411
Relating to the classification model, e.g. parametric or non-parametric approaches: based on the proximity to a decision surface, e.g. support…
G06V 10/50
Extraction of image or video features: by performing operations within image blocks; by using histograms, e.g. histogram of oriented gradients…

Timeline

4 Aug 2011Application filedUS 13/198,412
4 Aug 2011Assigned to Epson Research and Development, Inc.assignment of assignor's interest
30 Sep 2011Assigned to Seiko Epson Corporationassignment of assignor's interest
7 Feb 2013PublishedUS 2013/0034263 A1
3 Feb 2015GrantedUS 8,948,522 B2
12 Oct 2026TodayWithin its term
27 Sep 2033Term endsEstimated, includes 785 days of adjustment

Adaptive threshold for object detection

Life of this patent

Adaptive threshold for object detection

Life of this patent

Life of this patent

Abstract

Drawings

Drawings

Claims13

Description4,354 words

Classification6 codes

Life of this patent

Description4,354 words

Drawings